Terms such as conceptual thinking, logical thinking, critical thinking, systems thinking, and creative thinking highlight different functions. They do not name isolated mental organs, and they do not form one perfectly exclusive taxonomy.

A useful classification of thinking should tell us which cognitive problem each mode addresses, how it interacts with the others, and which errors it can correct.

The focus here is how conceptual and logical thinking function alongside other modes of inquiry. Concepts and Logic examine those subjects in their own right; this essay maps their roles in a broader activity rather than proposing a new unified discipline.

Conceptual thinking organizes experience

Conceptual thinking is the activity of forming, applying, comparing, and revising concepts so that experience can be classified and understood.

The phrase does not name a single discipline with one fixed theory. In philosophy, psychology, and education it overlaps with concept formation, categorization, abstraction, and conceptual reasoning. Used functionally, it includes at least six operations.

Distinction

Thought first separates what should not be collapsed: need from desire, intention from purpose, product from commodity, fact from opinion, and rule from principle.

Abstraction

Abstraction ignores some differences in order to preserve a structure relevant to the inquiry. Across many transactions, for example, one may isolate agents, objects, terms, and exchange.

Abstraction is selective rather than simply reductive. Which details can be ignored depends on the question.

Classification

Classification places a new case under a category or motivates a new category. A paid online service may be a product, a commodity, a service, or an object with several roles at once.

Boundary work

Conceptual thinking examines central cases, exclusions, and difficult border cases. A category that classifies only familiar examples has not yet demonstrated general usefulness.

Relation

Concepts become informative when connected:

situation → need → goal → intention → action
capability → product → commodity → exchange → outcome
evidence → judgment → inference → decision → feedback

Revision

Counterexamples, new evidence, and new practices can force a concept to change. A category protected from every correction becomes a label for defending an existing view rather than a tool for inquiry.

Logical thinking organizes commitments

Logical thinking turns judgments into arguments and examines whether conclusions follow from premises.

It asks:

• Are the claims mutually consistent?
• Which premises does the conclusion require?
• Is a step missing?
• Does the conclusion exceed the premises?
• Is there a counterexample in which the premises hold and the conclusion fails?
• Has a key term changed meaning during the argument?

Consider:

Successful products satisfy a need.
This product satisfies a need.
Therefore this product will succeed.

The argument affirms the consequent. Satisfying a need may be necessary for success without being sufficient. Logical analysis exposes the unsupported move.

Conceptual and logical thinking are not identical

Conceptual thinking asks:

What are we talking about, and how should it be distinguished and classified?

Logical thinking asks:

Given these claims, what follows?

They interact in a feedback loop:

experience
→ conceptual formation and classification
→ judgments
→ logical inference
→ new judgments
→ empirical test
→ revision of concepts and premises

Without concepts, logic lacks meaningful content to organize. Without logic, conceptual connections can remain arbitrary or contradictory.

Yet conceptual thinking is wider than inference. Creating a new distinction, redrawing a boundary, and redescribing a problem are not merely deductions from fixed premises.

Conceptual thinking supplies objects, classifications, and meanings for inference. Logical thinking constrains the judgments and transitions built from them.

Conceptual and abstract thinking

Abstract thinking extracts general structures from particular cases. Conceptual thinking also includes applying established concepts, comparing them, managing border cases, and revising a conceptual system.

Abstraction is therefore one major operation within conceptual thought. It is not a guarantee of truth. A thinker can abstract the wrong common feature or ignore a difference that later proves causally or morally important.

An abstraction earns its place by supporting successful classification, explanation, inference, or action across new cases.

Perception and intuition

Perception makes features of a situation available. Intuition often appears as rapid pattern recognition.

An experienced physician may notice an abnormal pattern before articulating it. A designer may immediately sense a hierarchy problem in an interface. Such judgments can result from compressed experience rather than mysterious access to truth.

Intuition is valuable for detecting signals and generating hypotheses. It is also vulnerable to biased samples, affect, and misplaced familiarity, so it needs later checking.

Embodied approaches to cognition further challenge the idea that thinking is only abstract symbol manipulation inside the head. Cognitive activity may depend on bodily capacities, environmental structure, and action.Stanford Encyclopedia of Philosophy: Embodied Cognition

Causal thinking

Causal thinking asks what produces or changes an outcome.

It examines mechanisms, temporal order, counterfactual dependence, confounding, selection, and intervention:

• Why did users abandon checkout?
• Would changing price alter conversion?
• Would the outcome have occurred without the policy?
• Is the observed variable a cause, an effect, a common consequence, or merely a predictor?

Logic can test the form of a causal argument. It cannot identify real causes from form alone. Causal claims need evidence and a defensible model of the system.

Probabilistic thinking

Probabilistic thinking represents uncertainty. Instead of asking only whether a proposition is true, it asks how strongly current information supports alternative possibilities.

It requires a thinker to:

• separate possibility, probability, and necessity;
• attend to base rates and conditioning information;
• update with new evidence;
• combine likelihood with consequence;
• distinguish one realized outcome from a long-run pattern or model.

Logic governs valid transformations within probabilistic reasoning, while probability supplies graded support that deduction alone does not express.

Critical thinking

Critical thinking is disciplined evaluation, not habitual opposition. It examines concepts, evidence, sources, inferences, counterexamples, alternative explanations, and the influence of perspective.

It asks whether:

• key terms are clear;
• evidence is relevant and reliable;
• a source is credible for the claim at issue;
• the inference is valid or adequately supported;
• contrary cases have been ignored;
• confidence exceeds the evidence;
• the thinker would revise the conclusion under specified conditions.

Its scope is wider than formal logic because natural-language interpretation, evidential quality, and intellectual self-correction also matter.Internet Encyclopedia of Philosophy: Critical Thinking

Creative thinking

Creative thinking expands the space of candidate concepts, explanations, and actions. It uses association, analogy, recombination, reframing, cross-domain transfer, and counterfactual imagination.

Generation and evaluation perform different work:

creative thinking produces candidates
→ critical thinking tests them
→ discovered weaknesses motivate new candidates

Generation without evaluation produces many unusable ideas. Evaluation without generation can make inquiry converge before serious alternatives exist.

Systems thinking

Systems thinking situates an object within interacting relationships. It attends to:

• feedback loops;
• delays between action and effect;
• local optimization and global outcomes;
• boundary choices;
• adaptation by other parts of the system;
• unintended consequences.

A lower price may attract more users while changing support costs, customer composition, perceived quality, and future positioning. A single-variable explanation misses these interactions.

Systems and causal thinking overlap. Systems thinking places greater emphasis on feedback, boundaries, multi-variable interaction, and behavior over time.

Dialectical thinking

Dialectical thinking has different meanings across philosophical traditions. In a broad functional sense, it draws attention to relation, change, internal tension, and the possibility that the same object behaves differently under different conditions.

This orientation can correct static and isolated analysis. It becomes empty, however, when “everything is dialectical” replaces explicit concepts, mechanisms, evidence, and conditions.

Strategic and game-theoretic thinking

Strategic thinking concerns long-term direction under limited resources and changing conditions. It coordinates goals, tradeoffs, capability building, timing, and sequences of action.

Game-theoretic thinking adds strategic interdependence: an agent’s outcome depends on what others do, while others anticipate and respond to that agent.

Both require some systems awareness. Strategic thinking emphasizes direction and resource allocation; game-theoretic thinking emphasizes mutual prediction, response, and equilibrium among agents.

Decision thinking

Decision thinking converts understanding into a commitment to act. It integrates:

• facts and information quality;
• probabilities of outcomes;
• causal and controllable conditions;
• goals and values;
• cost, risk, and opportunity cost;
• reversibility;
• authority and responsibility.

Many “thinking models” serve decisions, but not every model is a decision model. Classification models identify kinds, causal models explain changes, predictive models estimate future states, and decision models connect possible states and actions to preferences or values.

Metacognition

Metacognition monitors and regulates cognition itself:

• Which premises am I using?
• Do I understand this concept or merely repeat its label?
• Does my confidence exceed the evidence?
• Am I searching only for confirmation?
• What observation would change my mind?
• Should I continue analysis or move into action and feedback?

Metacognition does not answer the original question directly. It helps select, interrupt, and revise the process used to answer it.

These modes occupy different levels

A functional map makes the relationships clearer:

The table assigns each mode a prominent function, not an exclusive territory. Concepts operate throughout the process. Logic enters probability and causal analysis. Critical thinking can inspect every stage.

A cycle from world to action

The modes can be connected without reducing them to one method:

world and perception: what is happening?
↓
concepts and representation: what is it?
↓
logic, evidence, and probability: what is supported?
↓
causal and systems inquiry: why did it happen, and what could change it?
↓
creative and critical work: what alternatives survive examination?
↓
values, strategy, and decision: what should be pursued and chosen?
↓
action and outcome: what changed in the world?
↓
feedback and metacognition: what must be revised?
└────────────────────→ return to concepts, premises, and models

Real thought moves backward and sideways through this cycle. New evidence can change a concept. Resource limits can reshape a goal. An action can disconfirm the causal story that originally justified it.

Different failures need different repairs

A conceptual error uses an inadequate category. Saying that customers buy only functions may exclude trust, experience, identity, and risk reduction.

A logical error makes an unsupported transition. Inferring guaranteed success from need satisfaction affirms the consequent.

A factual error begins from a premise that does not match the world, even if the inference is valid.

A probabilistic error treats a low-probability realized event as proof that the earlier probability estimate was necessarily irrational.

A causal error treats association or predictive usefulness as evidence that intervention on the variable will change the result.

A practical or normative error can occur when the facts and inference are correct but the goal is unjustified or the cost unacceptable.

Diagnosis comes before repair. The thinker must know whether to revise a category, inference, dataset, causal model, or objective.

Conclusion

Conceptual and logical thinking are foundational, but neither is the whole of thought.

Conceptual thinking builds and revises the classifications through which a world becomes intelligible. Logical thinking constrains the commitments formed from those classifications. Probabilistic, causal, creative, critical, systems, strategic, decision, and metacognitive thinking address uncertainty, explanation, alternatives, evaluation, interaction, action, and correction.

Good thinking does not apply one favored method everywhere. It identifies the level of the present problem, uses the method suited to it, and allows outcomes to revise the concepts, premises, and models that guided the action.

A concept is not the object itself. It is part of the way an object becomes intelligible as something.

This essay asks what a concept is, how its boundaries form, and how counterexamples can revise it. Words, Concepts, and Objects examines expression and reference; Conceptual and Logical Thinking follows the use of concepts in judgment.

This gives concepts a double role. They reduce complexity enough for thought and communication, but every reduction highlights some differences and ignores others. Concepts make knowledge possible, and poorly formed concepts make systematic error possible.

Concepts, words, and objects are different

A word is a public expression. A concept is the content or capacity involved in understanding and using such expressions. An object is what the expression and concept may concern.

The three levels do not map one-to-one.

One word can express several concepts. Product may mean a manufactured item, the output of an operation, a market offering, or a managed digital service. Several expressions can also approach the same conceptual content. And a concept can represent something that does not exist, such as a unicorn, an ideal circle, or a fictional institution.

Consulting a dictionary is therefore only a beginning. Conceptual analysis also asks what is included, what is excluded, which contrasts matter, and whether the same expression changes meaning during an argument.

Three philosophical accounts of concepts

Contemporary philosophy does not offer one uncontested ontology of concepts. Three families of views organize much of the debate.

Concepts as mental representations

On a representational view, concepts are components of mental states that carry content. They help explain how a person can think about cats in their absence, combine CAT with other concepts, and draw new conclusions.

A representation need not be a vivid inner picture. It may be symbolic, prototype-like, schematic, or distributed across a cognitive system. What matters is that it represents something and participates in cognition. Mental representation is consequently a central theoretical construct in cognitive science.Stanford Encyclopedia of Philosophy: Mental Representation

Concepts as abilities

An ability view emphasizes what a competent thinker can do. Possessing the concept CAT may involve discriminating cats from relevant non-cats, understanding claims about cats, and drawing appropriate inferences.

This account explains why repeating a definition is insufficient evidence of understanding. Concept possession appears in recognition, application, explanation, and inference.

Concepts as abstract objects

An abstract-object view treats concepts as shareable contents rather than private mental episodes. Two people can think about the same concept even though their neural and psychological states differ. Mathematical, legal, and scientific concepts can remain available within a public practice after particular individuals forget them.

The Stanford Encyclopedia of Philosophy presents mental representations, abilities, and abstract objects as the three leading options. They emphasize psychological realization, competent use, and public content respectively.Stanford Encyclopedia of Philosophy: Concepts

These positions need not be casually collapsed into one theory. They answer different questions: what realizes a concept in a mind, what possessing it enables an agent to do, and what makes conceptual content shareable.

Intension, extension, and boundary

The intension of a concept concerns the properties, conditions, and relations used to characterize it. The extension concerns the things to which it applies.

For a concept such as COMMODITY, the intension might include availability for exchange under economic conditions. Its extension may include food, clothing, subscriptions, and standardized services.

Changing the intension commonly changes the extension. Requiring physical form would remove software from the extension. Removing exchange conditions might make almost every useful object a commodity.

The boundary between inclusion and exclusion is often where the real analysis begins:

• Is free software a product?
• Is personal data a commodity?
• Is an informal promise a contract?
• Is a simulated agent an entity?

Boundary cases reveal assumptions hidden by central examples.

Definitions and prototypes do different work

Some concepts are governed by explicit criteria. Others are organized partly around prototypes and family resemblances.

A sparrow is a prototypical bird. A penguin is still a bird despite lacking the prototype’s ability to fly. Prototype-based recognition is fast and useful, but a prototype is not automatically a definition.

This distinction matters in product design, law, and AI. A model trained on typical cases can fail on valid but unusual cases. A policy based only on familiar examples can exclude people or situations that satisfy the actual standard.

Definitions also differ in purpose. A lexical definition reports established usage. A stipulative definition introduces a usage for a particular inquiry. A theoretical definition situates something inside an explanatory theory. A legal or institutional definition may help constitute a status.

Asking “What is the correct definition?” is incomplete until the purpose and domain are specified.

Concepts form networks

Concepts rarely function as isolated entries. They occupy inferential and practical networks:

need → goal → intention → action
need → product → commodity → exchange
fact → judgment → inference → decision
value → principle → rule → constraint

To understand a concept is partly to understand what follows from applying it, what would count against applying it, and how it differs from neighboring categories.

This is why a topic word can serve as an index entry without yet being a developed concept. A useful concept record needs distinctions, criteria, examples, counterexamples, relations, and revision conditions.

Concepts represent more than objects

Concepts can concern different ontological and grammatical categories:

Treating every concept as a name for a thing produces category mistakes. Causation is not another physical object sitting beside causes and effects. Probability is not located somewhere in space. Responsibility may depend on relations among agents, norms, knowledge, and control.

How concepts are formed

No single process explains every concept. Concept formation can involve:

• perceptual discrimination among recurring features;
• abstraction from particular instances;
• classification under learned categories;
• linguistic correction within a social practice;
• embodied interaction and practical feedback;
• institutional rules that create statuses;
• scientific theories that reorganize ordinary categories.

Some concepts are learned through repeated examples. Some are explicitly defined. Some are created by institutions. Others change when a theory explains why familiar classifications were misleading.

Scientific concepts illustrate the difference between preserving a word and preserving a concept. Everyday language may keep the word heat while physics gives it a more precise theoretical role. Conceptual continuity cannot be inferred from verbal continuity alone.

Why conceptual disputes persist

At least four kinds of disagreement are easily conflated.

First, speakers may attach different criteria to the same word. One person may define a product by production, another by user value, and another by market offering.

Second, the category itself may have vague boundaries. Game, art, intelligence, and consciousness resist simple criteria that cover every accepted case.

Third, descriptive and evaluative content may be mixed. Normal, successful, progressive, and civilized can describe patterns while also conveying approval.

Fourth, disciplines may construct different concepts for different explanatory purposes. Entity does different work in metaphysics, databases, natural-language processing, and law.

Good analysis does not erase these differences. It states which concept is being used, in which domain, for which purpose, and with which exclusions.

Reification and other conceptual errors

Conceptual mistakes take recurring forms:

• word-object confusion: assuming that every noun names a separate entity;
• equivocation: shifting a term’s meaning during an argument;
• category mistake: asking of one kind of thing a question appropriate to another;
• reification: treating an abstraction such as “the market” as a unified intentional agent;
• unwarranted essentialism: assuming every useful category has one timeless hidden essence;
• prototype substitution: treating a familiar example as the full criterion;
• descriptive-normative collapse: inferring what ought to be from what is common;
• level confusion: mixing an object, a model of the object, and an evaluation of the model.

These errors cannot always be repaired by gathering more data. Sometimes the categories used to organize the data must be repaired first.

A method for analyzing a concept

A reusable inquiry can ask:

1. Which expressions are used for the concept?
2. What kinds of entities, events, properties, or relations does it concern?
3. How do ordinary and specialized uses differ?
4. What criteria make up its intension?
5. What central cases belong to its extension?
6. Which boundary cases are difficult?
7. Which neighboring concepts must be separated?
8. Which counterexamples challenge the current account?
9. What cognitive or practical work does the concept perform?
10. What evidence or practice would justify revising it?

Conceptual analysis is not the search for a sentence immune to change. It builds a public, usable distinction and specifies how reality can correct it.

Conclusion

Concepts let finite thinkers move beyond isolated experiences. They make recognition, communication, judgment, and inquiry possible.

Every concept also selects. It makes some differences visible and leaves others in the background. Responsible concept use therefore requires criteria, boundaries, contrasts, counterexamples, and revision.

To possess a concept is not merely to know a word. It is to identify, distinguish, apply, infer, and revise with it.

The ambiguity matters because these attitudes answer different questions:

What do I think will happen?
What do I want to happen?
What is someone supposed to do?
What do I believe my action can bring about?

When these questions are compressed into one word, desire can masquerade as evidence, prediction can sound like obligation, and another person’s demand can be treated as a fact about the future.

A working definition

An expectation can be defined as:

An expectation is an attitude toward an outcome that remains unresolved for an agent, representing that outcome as likely, anticipated, required, or connected to action.

The outcome is usually future-directed, but it need not concern an event that has not yet occurred. Someone may say, “I expect the package has already arrived,” when the delivery is complete but its status remains unknown to them. The important condition is epistemic openness: the agent does not yet know the outcome.

Every expectation therefore contains at least:

an agent
+ an unresolved outcome
+ a way of representing that outcome
+ some attitude toward its occurrence

The last element determines which kind of expectation is involved.

Expectation, expectancy, hope, and anticipation

English separates several ideas that ordinary conversation often blends.

Hope has both a cognitive and a conative side. A standard philosophical account treats hope as involving a desired outcome together with belief that the outcome remains possible. A person can therefore hope for an outcome that they regard as very unlikely. That same person would not honestly say that they expect it.Stanford Encyclopedia of Philosophy: Hope

Expectation usually places more weight on belief. Hope places more weight on desirability. Anticipation adds an affective and attentional orientation toward an approaching event.

Predictive expectations

A predictive expectation represents what an agent thinks will happen.

Given the current release data, I expect adoption to grow next month.

This is an epistemic judgment. It should answer to evidence, and it should change when the evidence changes.

A predictive expectation may be precise:

There is a 70% chance that demand will exceed capacity.

It may also be qualitative:

Demand will probably remain stable.

Neither statement guarantees the outcome. Both summarize what the agent currently takes the evidence to support.

This kind of expectation should be kept separate from desire:

I want A to happen.
I currently expect B to happen.

There is no contradiction in holding both attitudes. Rational agency often begins by acknowledging that the desired outcome is not the most likely one.

Expectancy as an action–outcome belief

In psychology, expectancy often refers to a belief that an action can produce a particular outcome. The APA Dictionary describes it both as a mental set shaping how a person approaches a situation and, in motivation theory, as a belief that one’s actions can attain an outcome.APA Dictionary of Psychology: Expectancy

This introduces agency:

If I perform action A under conditions C,
how likely is outcome O?

Such beliefs influence effort and persistence. If someone believes that preparation can change performance, preparation becomes instrumentally intelligible. If they believe that nothing they do can affect the result, motivation may collapse even when the desired outcome remains valuable.

Expectancy is still not intention. Believing that an action would work does not mean that the agent has decided to perform it.

Normative expectations

An expectation can also express a standard rather than a forecast:

The organization expects employees to protect confidential information.

This sentence need not predict universal compliance. It states what employees are required or supposed to do.

Normative expectations organize families, workplaces, institutions, and social roles. They also raise questions of authority:

• Who sets the expectation?
• On what grounds?
• Who bears the cost of compliance?
• What follows from refusal?
• Can the expectation be contested or renegotiated?

The fact that a standard is widely expected does not establish that it is justified. A social regularity, a role demand, and a moral obligation are different things.

Expectations can alter outcomes

Expectations shape attention, interpretation, effort, and interaction. This can create a feedback loop:

expectation
→ attention and behavior
→ changes in the situation or in other people
→ observed outcome
→ reinforcement or revision of the expectation

If someone expects to fail, they may prepare less, withdraw earlier, or interpret ambiguous feedback as confirmation. The resulting behavior can make failure more likely. In social interaction, a perceiver’s expectation may elicit behavior from another person that appears to confirm the original belief, a process described as behavioral confirmation.APA Dictionary of Psychology: Behavioral Confirmation

This does not mean that thought directly controls reality. Expectations influence outcomes only through causal pathways such as attention, effort, communication, coordination, and the reactions of other people. External constraints and chance remain real.

Expectation is not intention

Expectation belongs to a larger chain of agency:

“I expect the article to reach more readers” expresses a belief or hope about an outcome. “I intend to revise the title and opening” introduces an action commitment. “I will revise the title today” is a more specific decision.

The full loop is:

expectation
→ intention
→ decision
→ action
→ outcome
→ revised expectation

The arrows are not automatic. An expectation may never become an intention. An intention may not survive execution. Action does not control every cause of the final result.

Expectation gaps and reference points

Expectations also become standards against which outcomes are experienced.

expectation gap = observed outcome − reference expectation

The same outcome can feel favorable when it exceeds the reference point and disappointing when it falls below it. The outcome has not changed; the comparison has.

Managing expectations should therefore not mean adopting permanent pessimism. A better practice is to separate four questions:

1. Desired outcome: What would I like to happen?
2. Predictive judgment: What does the current evidence make likely?
3. Controllable action: Which conditions can I influence?
4. Risk boundary: What can I tolerate if the desired result does not occur?

This preserves ambition without converting desire into prediction.

How to examine an expectation

When a person or institution says, “I expect…,” ask:

1. Who holds the expectation?
2. What outcome remains unresolved?
3. Is this a prediction, a hope, a standard, or an action–outcome belief?
4. If it is predictive, what evidence supports it?
5. If it is normative, who has authority and who bears the cost?
6. Which causal conditions can the agent influence?
7. Has the expectation become an intention, decision, or plan?
8. What observation would require revision?

A final definition

Expectation is an agent’s orientation toward an unresolved outcome: a representation of what is likely, desired, normal, required, or achievable through action.

The concept spans three domains that must remain distinguishable:

epistemic: what is likely to happen
evaluative: what would be good to happen
normative: what is supposed to happen

Once these are separated, expectations can be tested as beliefs, discussed as values, challenged as standards, and converted into action without being mistaken for guarantees.

References

• Stanford Encyclopedia of Philosophy: Hope
• APA Dictionary of Psychology: Expectancy
• APA Dictionary of Psychology: Behavioral Confirmation

The word system becomes useful when parts are related in ways that produce persistent behavior at the level of a whole.

A system is a bounded set of interdependent elements whose organization and interactions produce behavior over time within an environment.

This definition contains several commitments. A system has elements, but it cannot be understood from an inventory alone. It has a boundary, though that boundary depends partly on the question being asked. It has a state that can change. It interacts with an environment. Its overall behavior depends on relations among parts, not merely on the parts considered separately.

The word and its central idea

English system comes through Latin systema from Greek systēma, an organized whole composed of parts. Its roots carry the idea of things standing together rather than existing as an unrelated assortment.Etymonline: system

The term now covers very different objects:

• the solar system;
• a nervous system;
• an ecosystem;
• a legal system;
• an organization;
• a payment system;
• a software system;
• an AI system.

These examples do not share one material composition or one kind of purpose. What they share is an analytical form: distinguishable elements participate in relations that sustain some pattern of behavior at the level of a whole.

What must a system contain?

The smallest useful account of a system normally identifies:

elements
+ relations
+ a boundary

An account of how the system operates also needs:

state
+ rules or mechanisms of change
+ an environment
+ inputs and outputs
+ time

For an engineered system, further questions become central:

intended purpose
+ constraints
+ performance criteria
+ authority and responsibility

Engineering standards commonly define a system as interacting elements organized to achieve one or more stated purposes.ISO/IEC/IEEE 42020:2019

That purpose-centered definition is appropriate for engineered systems. It should not be projected onto every natural system. A climate system and a river system display organized behavior without needing intentions of their own.

A system is not an inventory

Suppose an online publishing operation contains an author, articles, Markdown files, a Git repository, a static-site generator, a deployment service, a domain, search engines, and readers.

The list tells us what might be present. It does not yet explain the system. The explanation begins with relations:

author ──writes──> article
article ──stored as──> Markdown
Git ──records──> revision
site generator ──transforms──> web page
deployment service ──publishes──> site
reader ──visits──> page
search engine ──indexes──> content

The system exists as an organized pattern of dependencies, transformations, permissions, and flows. If those relations disappear, the same objects may remain, but the publishing capability does not.

This is why a system is not simply the sum of its components. Organization is causally relevant.

How is a system different from a collection?

A collection is defined mainly by membership. A system is defined by interdependence and organization.

One hundred chairs in a warehouse form a collection. Players, coaches, rules, training, communication, and matches may form a team system. A contact list is a collection of records. Communication, roles, authority, and feedback can turn a group of people into an operating organization.

Boundaries make system analysis possible

Everything is connected to something else. If every remote influence must be included, the system expands until it becomes indistinguishable from the world. A useful analysis therefore defines a boundary.

A boundary answers:

Which elements and relations belong to the system under study, and which belong to its environment?

Consider a coffee shop. An analysis of waiting time may include customers, ordering, baristas, equipment, and queue rules. An analysis of profitability may add rent, suppliers, delivery platforms, and pricing. A food-safety analysis may include storage, temperature control, cleaning, and regulation.

The coffee shop has not become three different objects. The analytical boundary changes because the question changes.

INCOSE describes the environment as the part of the outside world that significantly interacts with and affects a system, often as the source of inputs and destination of outputs. Defining the boundary and environment is therefore a basic step in systems thinking.INCOSE: Systems Thinking 101

The boundary is selected, but it is not arbitrary. A model cannot exclude a major causal influence merely because including it would make the diagram untidy.

Open systems exchange with an environment

An environment may provide information, energy, materials, users, prices, legal constraints, threats, and disturbances. A system may return products, decisions, services, waste, risk, and social effects.

environment
    ↓ input
system state and operation
    ↓ output
changed environment
    ↓ new input
continued operation

Most systems studied in biology, society, organizations, and computing are open in this sense. They depend on continued exchange.

A closed system is often an analytical idealization. It means that particular exchanges can be ignored for a particular purpose, not that the object has no relation to anything outside it.

Systems have state and history

A system is not only an architecture diagram. It occupies states, changes state, and carries effects from its history.

A website may be healthy, building, partially deployed, unavailable because of DNS, or updated in a repository while production still serves an earlier release. The same components and nominal connections can therefore produce different outcomes at different moments.

A simple dynamic description is:

current state + input + transition rules
                    ↓
             next state + output

Time delays matter. A page may be live before a search engine indexes it. A product may improve before public expectations change. A policy may appear effective in the short term while accumulating a long-term cost.

Ignoring delay turns a dynamic system into a misleading snapshot.

Feedback changes subsequent behavior

Feedback occurs when an output or consequence returns as an influence on later system behavior.

Negative feedback counteracts a deviation. A thermostat detects a falling temperature, turns on heating, and later turns it off as the room returns to range. Negative feedback often supports stability.

Positive feedback reinforces a change. More engagement may produce more recommendation exposure, which produces more engagement. Positive feedback may create growth, lock-in, polarization, or collapse.

The words positive and negative do not mean good and bad. They describe whether the loop amplifies or counteracts change.

Delayed feedback can cause overshoot and oscillation. Missing feedback can allow a system to optimize a proxy long after the proxy has stopped representing the intended result.

Emergence comes from organization

“The whole is greater than the sum of its parts” is a familiar systems slogan. It is suggestive but imprecise.

What the inventory omits is organization: spatial arrangement, causal interaction, timing, feedback, and constraints. Those relations allow a whole to display properties that isolated parts do not display.

• A single vehicle does not constitute a traffic jam; many mutually constraining vehicles can.
• A single market participant does not determine a market price; structured interaction among many participants may produce one.
• A single neuron does not perform the full cognitive work of a nervous system; organized neural activity supports higher-level capacities.
• A molecule does not have the thermodynamic temperature of a macroscopic body; temperature characterizes a collective state.

Systems biology likewise studies components in the context of their interactions and the constraints imposed by the whole.Stanford Encyclopedia of Philosophy: Philosophy of Systems and Synthetic Biology

Emergence should not be used as a label for mystery. An emergent property still calls for an explanation of arrangement, interaction, scale, constraint, and time.

Does every system have a goal?

No. Purpose, function, and intention must be distinguished.

An engineered payment system is built for stated purposes. An organization may pursue several partly conflicting objectives. A heart performs a biological function that can be explained through physiology and evolution without attributing intention to the organ. A solar system displays lawful behavior without pursuing a goal.

Even in designed systems, four things can diverge:

stated purpose
designer intention
metric actually optimized
outcome actually produced

A recommendation service may claim to improve user satisfaction while optimizing clicks. Higher click-through rates may coexist with lower trust, worse information quality, or compulsive use.

A system should therefore be evaluated by its operation and effects, not only by its declared purpose.

System, structure, mechanism, process, and model

These terms answer different questions.

A publishing system has a structure of repositories, builders, and servers; a mechanism that converts Markdown into HTML; a process of drafting, review, commit, build, and deployment; and a function of making content reliably accessible.

The system is the object under study. A system model is a selective representation of that object. A diagram may omit details usefully, but success in the diagram does not guarantee success in the world.

Systems, entities, and ontology

An entity analysis asks which particular object is being referred to. A system analysis asks how entities and relations are organized into behavior over time. An ontology specifies which kinds of entities and relations a domain recognizes.

entity: an identifiable object
relation: a way objects are connected
system: organized objects and relations operating over time
ontology: an account of the recognized kinds and relations

A system can itself be treated as an entity. A payment system may have an owner, version, status, service boundary, and lifecycle. At a lower level, the same system contains account, order, fraud-control, channel, and settlement subsystems.

Something can therefore be a system at one level and a component of a larger system at another. The relevant level depends on the question.

Related discussions appear in “What Is an Entity? Identity, Reference, and AI Systems” and “What Is Ontology? From What Exists to What AI Can Represent”.

Subsystems and systems of systems

Complex systems are often nested:

component
    ↓
subsystem
    ↓
system
    ↓
larger system

A payment interface may belong to an order-and-payment subsystem, which belongs to an e-commerce platform, which participates in a broader commercial and logistics environment.

A system of systems joins systems that can still operate and be managed with substantial independence. Urban mobility, for example, may involve roads, buses, rail, navigation, ticketing, traffic control, taxis, and ride-hailing platforms. No single component completely determines the whole, yet their interactions shape citywide movement.

This creates coordination problems that cannot be solved by optimizing one subsystem alone.

Are systems discovered or constructed?

Both descriptions capture part of the truth.

Real interactions, dependencies, feedback loops, and constraints are not invented merely by drawing a boundary. Traffic congestion, ecological exchange, institutional authority, and software calls have real effects.

Yet the choice of boundary, scale, state variables, and level of abstraction depends on what the investigator needs to explain. The same person can participate in biological, family, legal, organizational, economic, and information systems.

A useful position is:

Interactions are constrained by reality; system boundaries and levels are selected for inquiry and action.

The model must remain answerable to observed behavior. A convenient boundary that excludes decisive effects is a bad boundary.

System boundaries also carry values and power

Boundary choices determine what becomes visible.

When a system defines who counts as a user, which outcomes count as benefits, which harms count as externalities, which metric receives optimization pressure, and who may change the rules, it embeds practical and political judgments.

A delivery platform that measures only completed orders per hour may improve its internal efficiency while moving safety risk, waiting pressure, and road danger onto workers and the public.

The system did not eliminate the cost. Its measurement boundary excluded the cost.

Systems analysis therefore asks more than whether an operation is efficient:

• Efficient for whom?
• Which outcomes are measured?
• Who absorbs failure and delay?
• Who has authority to change the objective or boundary?
• Which effects appear only in a larger system?

What is an AI system?

An AI system is not synonymous with an AI model.

A deployed language-model application may include:

users and operators
+ interface
+ prompts and context
+ one or more models
+ retrieval and knowledge sources
+ memory and state
+ tools and external services
+ orchestration and control flow
+ identity and permissions
+ runtime infrastructure
+ logging and evaluation
+ human review
+ outcome feedback

The model performs part of the inference. The surrounding system decides what reaches the model, which external state is available, which tools may run, whose authority applies, how outputs are checked, and what changes in the world.

The OECD definition describes an AI system as a machine-based system that infers from inputs how to produce predictions, content, recommendations, or decisions that can influence physical or virtual environments. Its explanatory material describes a model as a core component of such a system.OECD: Updated Definition of an AI System

NIST emphasizes that AI systems are sociotechnical: their benefits and risks emerge from technical components together with operators, users, other systems, and the social context of deployment.NIST AI Risk Management Framework

A model can answer while the system still fails

A language model may generate a correct article draft. Publishing that article reliably requires a larger system to:

• resolve the intended article and language versions;
• read repository rules;
• modify the correct files;
• preserve identity and authorization boundaries;
• validate the build;
• commit and push with the correct repository identity;
• wait for deployment;
• verify the live pages, canonical URLs, language links, and sitemap.

Failure can occur at any boundary. The model may be correct while retrieval supplies stale facts. The plan may be correct while a tool targets the wrong entity. The code may build locally while deployment fails. The page may be live while search metadata is missing.

Therefore:

Model capability does not by itself establish system reliability.

Evaluation must match the system boundary. A model benchmark measures a model under specified conditions. It does not automatically measure the complete product, workflow, organization, or real-world outcome.

A practical method for analyzing a system

1. State the question. Are you trying to explain, design, improve, control, or evaluate the system?
2. Draw the boundary. What belongs inside, what belongs to the environment, and why?
3. Identify elements. Include people, software, physical objects, rules, institutions, data, and resources.
4. Map relations and flows. Track information, material, money, authority, and dependency.
5. Describe states and transitions. What states can occur, and what changes one state into another?
6. Find feedback loops. Which consequences reinforce or counteract earlier behavior?
7. Examine delays. When do outputs and consequences become observable?
8. Separate goals from metrics. What outcome is intended, and what variable actually receives optimization pressure?
9. Locate constraints and authority. Who can change which element, rule, or boundary?
10. Inspect excluded effects. Which people, costs, and risks appear only when the boundary expands?

A final definition

A system can be defined as:

A system is a bounded whole in which interdependent elements operate through some organization and rules, change state over time, interact with an environment, and thereby produce behavior or capacities at the level of the whole.

For designed systems, that account must also include purpose, constraints, metrics, authority, and responsibility.

A system is not merely many things placed together, and it is not merely an architecture diagram. It exists as organized interaction: changes in one part affect others, relations alter outcomes, and the condition of the whole constrains what its parts can do.

Identifying entities tells us what is present. Understanding a system tells us how those entities operate together and why the observed result occurs.

References

• Etymonline: system
• ISO/IEC/IEEE 42020:2019
• INCOSE: Systems Thinking 101
• Stanford Encyclopedia of Philosophy: Philosophy of Systems and Synthetic Biology
• OECD: Updated Definition of an AI System
• NIST AI Risk Management Framework

The word is broad because it does not describe a particular material composition. It describes a role within thought, language, or a model:

An entity is something treated as having a distinguishable identity, so that it can be referred to, described, related to other things, tracked over time, or acted upon.

An entity need not be tangible. It need not exist independently. It need not even exist in the actual world. What it needs, within a particular domain, is enough identity for the system to treat it as the same subject across multiple statements or operations.

Where does the word “entity” come from?

English entity comes through Medieval Latin entitas, formed from ens, a being or something that is, which in turn is related to the Latin verb esse, to be. The word therefore carries an ontological background: it concerns something considered as a being or item of existence.American Heritage Dictionary: entity

That history explains why entity can be used so widely. It may refer to a physical object, a person, an institution, an event, a number, a proposition, or another item admitted by a theory. In philosophy, thing, being, entity, and object may all compete for the role of a maximally general term for whatever a system acknowledges.Stanford Encyclopedia of Philosophy: Object

No single list of entities is philosophically neutral. A physicalist ontology, a mathematical ontology, a legal ontology, and a fictional world may recognize different kinds of things. Calling something an entity is therefore both a semantic move and, in many contexts, an ontological commitment.

An entity is not merely a physical object

A physical object usually has material structure and some spatial boundary. An entity need not.

A corporation has no single body. It depends on law, records, roles, property, contracts, and continued institutional recognition. A meeting is not a durable object, but it can have participants, a time, a location, an agenda, and an outcome. An account exists only within a platform and its rules, yet the platform must still distinguish one account from another.

All three can function as entities because each can be identified and become the subject of further claims.

The English words entity and substance should also be separated. An entity is any item treated as a being or object of reference. A substance, in major philosophical traditions, is more specifically something taken to exist relatively independently or to bear properties. Events, relations, numbers, and properties may count as entities without counting as substances in that stronger sense.Stanford Encyclopedia of Philosophy: Substance

Reference does not establish real existence

Sherlock Holmes can be named, described, compared with other characters, and placed in a network of fictional relations. He is an entity in literary discourse and may be an entity in a knowledge base. None of this makes him a historical person.

At least three questions must therefore remain separate:

An entity can exist at the first two levels without satisfying the third. Plans, hypothetical products, possible events, mistaken identities, and fictional characters all demonstrate why representation and reality must not be collapsed.

Identity is the central problem

Finding a noun is easy. Determining what makes something the same entity is harder.

A person may change names and addresses while remaining the same person. A corporation may replace every employee and continue as the same legal organization. An article may be revised many times while retaining one publication history. A product may keep its commercial name while its capabilities, components, or terms change substantially.

Different types of entities require different identity criteria:

A name is not an identity. Two people can share a name, and one person can use several names. A set of properties is not automatically an identity either. Properties change, records conflict, and two objects may resemble one another closely without being the same object.

A useful entity representation usually needs more than a label:

stable identifier
+ entity type
+ attributes
+ relationships
+ temporal state
+ provenance
+ confidence in the identity match

How does ontology relate to entities?

An entity does not become a useful modeling unit in isolation. A system must decide what kinds of entities it recognizes, which attributes they may have, which relations may connect them, and what counts as persistence or change.

Those decisions form part of an ontology.

An entity is a particular item recognized within a domain; an ontology states which kinds of items the domain recognizes and how they may be organized.

“Shanghai” may be stored as a string in a simple customer table. In a geographical knowledge system, it may be represented as an entity with coordinates, administrative status, districts, and relationships to other places. The difference depends on whether the system needs to identify Shanghai independently and reason about it.

Philosophical ontology asks what exists and how different kinds of beings exist. Computational ontology turns a domain commitment into an explicit model of classes, entities, properties, relations, and constraints. A fuller account appears in “What Is Ontology? From What Exists to Models AI Can Use”. Here ontology matters because it supplies the type system and identity conditions within which entities can be distinguished.

What does “entity” mean in AI?

In AI, an entity is usually an object distinguished from text, images, records, or an environment because the system needs to understand, retrieve, remember, reason about, or act on it.

The term changes meaning across tasks. Named-entity recognition, entity linking, entity resolution, knowledge graphs, computer vision, databases, and AI agents do not operate at exactly the same level.

Named-entity recognition finds mentions, not verified objects

Consider the sentence:

Apple plans to announce a new phone in Shanghai tomorrow.

A named-entity recognition system may label:

• Apple as an organization;
• Shanghai as a location;
• tomorrow as a date.

At this stage, the system has identified mentions: spans of language that appear to refer to named or otherwise categorized entities. In engineering terms, a conventional NER component predicts labeled token spans. spaCy, for example, describes its entity recognizer as identifying non-overlapping labeled spans.spaCy: EntityRecognizer

The output does not yet prove that Apple refers to Apple Inc. rather than a different organization, a title, or an annotation mistake. Nor does it establish that the announced phone is a particular product with a known identity.

The distinction is fundamental:

entity mention ≠ entity identity
entity name ≠ unique referent
entity type ≠ proof of existence

Entity linking connects a mention to a canonical entity

Entity linking attempts to determine which known entity a mention refers to.

“Apple” in a sentence
        ↓ disambiguation
Apple Inc.
        ↓ normalization
knowledge-base ID: company/apple-inc

This requires at least two forms of reasoning:

• Disambiguation: Which entity with this name fits the context?
• Coreference and normalization: Do Apple, Apple Inc., and the Cupertino company refer to the same entity here?

Linking converts a piece of language into an addressable object in a knowledge system. It is still fallible. The selected knowledge-base entry may be wrong, duplicated, incomplete, or out of date.

Entity resolution asks whether records describe the same thing

Entity linking usually begins with language and a knowledge base. Entity resolution often begins with records.

A customer system may contain:

Li Ming, Shanghai, 138****1234
Ming Li, 上海市, 138****1234
李明, Pudong, 138****1234

Do these records describe one person, two people, or three? Shared fields provide evidence, but they do not make the answer automatic. Phone numbers can be reassigned, addresses can be shared, and names can collide.

Entity resolution may involve:

• deduplicating records;
• matching aliases and transliterations;
• detecting that one entity has split or merged in a source system;
• preserving conflicting claims rather than forcing a premature merge;
• recording why two records were considered the same.

This is an identity decision under uncertainty. A useful system keeps the evidence and confidence behind the match instead of treating similarity as certainty.

Knowledge graphs place entities in a network of claims

In a knowledge graph, entities are commonly represented as nodes connected by typed relations:

Apple Inc. ──headquartered in──> Cupertino
Apple Inc. ──released──> iPhone
iPhone ──instance of──> smartphone product

This representation distinguishes:

• entities such as Apple Inc., Cupertino, and an iPhone model;
• classes such as company, city, and product;
• attributes such as dates and names;
• relations such as headquartered in and released.

RDF represents claims as subject–predicate–object triples. Its notion of a resource is deliberately broad: a resource may denote a physical thing, a document, an abstract concept, or another item in the universe of discourse.W3C: RDF 1.1 Concepts and Abstract Syntax

An ontology supplies general rules, while a knowledge graph contains particular claims:

ontology: a company may release a product
claim: Apple Inc. released a particular iPhone model

A graph node is not automatically a verified real-world object. It may represent a class, a fictional entity, a planned object, an uncertain hypothesis, or an erroneous record. Provenance and status remain necessary.

Concepts, classes, entities, identifiers, and records

Several layers are easily confused:

Company may be a concept and a class. Apple Inc. may be an entity. Apple may be a name or mention. company_001 may be an internal identifier. A row in a database may contain claims about the company. None of these layers is identical to the organization itself.

One entity may have many names and records. One record may accidentally combine facts about several entities. A unique database key guarantees uniqueness inside a table; it does not prove that the row corresponds correctly to one real-world thing.

When should something be modeled as an entity?

Not every noun phrase needs its own entity. Six questions help:

1. Does it need a distinct identity? Must the system distinguish this item from similar items?
2. Will it be referred to repeatedly? Will multiple documents, records, or tasks mention it?
3. Does it have its own attributes? Must the system store a status, date, location, or owner?
4. Does it participate in relationships? Must it be connected to other objects?
5. Does it persist through time? Must the system recognize it after some properties change?
6. Can the system act on it? Will it be queried, edited, sent, authorized, purchased, deleted, or monitored?

If most answers are yes, an entity is often appropriate. Otherwise a literal value, label, or temporary span may be enough.

Entity modeling is not free. Every entity type introduces identity rules, lifecycle questions, merge and split behavior, provenance requirements, and access-control consequences.

Can events, states, and intentions become entities?

Grammatical categories do not map directly onto model categories. Nouns do not always denote entities, and verbs do not always remain mere relations.

“Maya signed Contract C36” can be represented as a simple relation:

Maya ──signed──> Contract C36

If the system must record the date, location, version, witnesses, method, and legal status, the signing can become an event entity:

Signing Event E1024
├── signer: Maya
├── object: Contract C36
├── time: 2026-09-24
├── place: Shanghai
└── status: effective

An intention can likewise be modeled as a mental-state entity when the system needs to record whose intention it is, what outcome it concerns, when it was inferred, which evidence supports the inference, and how uncertain the interpretation remains.

Reification makes an event, relation, or state available for further description. It is useful when the system needs that detail, but it should not be mistaken for a discovery that the world naturally divides itself in exactly that form.

How do large language models handle entities?

A standard large language model receives tokens and computes distributed representations shaped by training data and current context. It can learn that Apple often occurs near company, product, iPhone, and Cupertino, and it can frequently disambiguate the word from the fruit.

This competence does not imply that the model contains a single, explicit, canonical Apple Inc. record comparable to a carefully maintained knowledge base. Entity information may be distributed across parameters and reconstructed probabilistically in context.

Consequently, a language model may:

• resolve an entity correctly in one context and confuse it in another;
• conflate a company, its brand, its products, and its website;
• recall an old property without knowing that it has changed;
• invent a plausible person, paper, organization, or product;
• answer without a stable source for the entity claim.

For tasks that require reliable action, model output is usually only one layer:

language model
+ domain ontology
+ entity IDs and resolution
+ current state and provenance
+ time-aware records
+ permissions and action rules

The language model interprets open-ended language. The ontology defines possible types and relations. Entity services determine identity. Data sources establish current state. Authorization determines which operations are permitted.

Why do AI agents need explicit entity identity?

Consider the instruction:

Update yesterday’s article about needs on Moonment.

The request contains several unresolved references:

The agent may understand the general intention while still acting on the wrong file, project, account, language version, or deployment.

A safer path is:

natural-language request
        ↓
mention and reference detection
        ↓
type assignment and coreference
        ↓
entity linking and resolution
        ↓
relation, event, and intent interpretation
        ↓
state, provenance, and authorization checks
        ↓
action on the resolved entity

Intent specifies the change the user appears to seek. Entity resolution determines what the intended action applies to. Authorization determines whether the action may be performed. None can substitute for the others.

Can AI establish that an entity really exists?

Language alone cannot establish existence.

An AI system can estimate that a phrase is probably a person’s name, that context probably indicates a company, or that a mention probably links to a known record. Real-world verification requires additional evidence, such as:

• an authoritative registry or primary source;
• a current database record;
• a file that actually exists in the relevant filesystem;
• a live website or API;
• an authenticated identity and permission system;
• a sensor observation or human confirmation.

The following claims are distinct:

the text mentions something
≠ the model identified it correctly
≠ the system linked the correct record
≠ the record is accurate and current
≠ the corresponding thing exists now
≠ the user is authorized to act on it

Model confidence is not proof of existence. It describes a model’s judgment under particular inputs and assumptions. It does not replace provenance or verification.

A checklist for entity design in AI systems

1. Which entity types does the system recognize, and why?
2. Are mentions, names, classes, entities, identifiers, and records kept separate?
3. What establishes identity for each entity type?
4. How are aliases, namesakes, renaming, and duplicate records handled?
5. Do attributes and relations carry time and provenance?
6. Are events and changing states being flattened into misleading static fields?
7. Are extraction, linking, resolution, and real-world verification separate stages?
8. Can the system distinguish fictional, planned, hypothetical, and actual entities?
9. What happens to historical claims when entities merge, split, or change type?
10. Before acting, how does the system verify identity, current state, and permission?

An entity-rich system can still be unreliable. Without identity criteria, provenance, temporal state, and authorization, it merely attaches confident-looking labels to uncertain referents.

Entities connect language to action

Entities form an interface between language, knowledge, and operations:

objects, events, and states in a domain
                ↓
names and descriptions in language
                ↓
entity, attribute, and relation recognition
                ↓
links to records and external evidence
                ↓
interpretation of needs and intentions
                ↓
authorized action and recorded outcomes

An entity answers, “Which particular thing are we talking about?” An attribute answers, “What is it like?” A relation answers, “How is it connected?” An event answers, “What happened?” An intention answers, “What outcome is an agent trying to bring about?”

The most serious entity error in AI is not failing to label a noun. It is treating an ambiguous expression as though it already named one verified, current, and actionable object.

References

• American Heritage Dictionary: entity
• Stanford Encyclopedia of Philosophy: Object
• Stanford Encyclopedia of Philosophy: Substance
• Stanford Encyclopedia of Philosophy: Logic and Ontology
• spaCy: EntityRecognizer
• spaCy: Linguistic Features—Named Entity Recognition
• W3C: RDF 1.1 Concepts and Abstract Syntax
• W3C: OWL 2 Web Ontology Language Primer

At its broadest, ontology asks:

What exists, what kinds of things exist, and what basic structures relate them?

An inventory is only the beginning. An ontology must also address whether objects, events, properties, relations, numbers, institutions, and mental states exist in the same way. It must say what makes an entity the same entity through change and whether some things depend on other things for their existence.

A rock and a corporation can both be treated as entities. The rock persists through a physical organization of matter. The corporation depends on legal rules, records, offices, agreements, and collective practices. Both can be real without having the same mode of existence or the same conditions of identity.

Ontology therefore concerns more than a hidden substance behind appearances. It studies the inventory, categories, identity conditions, dependence relations, and general structure of a world.

The Term and Its Scope

The word ontology combines Greek roots associated with being or that which is, and with study or account. The discipline is often introduced through the question “What is there?” Yet contemporary ontology also investigates the most general features of what there is and the ways different entities relate. The term itself became a disciplinary label in early modern philosophy rather than arriving unchanged as an ancient Greek field name. Stanford Encyclopedia of Philosophy: Metaphysics

This broad scope matters because ontology is sometimes reduced to one of two narrower projects:

• a search for the ultimate material from which everything is made;
• a search for the essential definition of each kind of thing.

Both can raise ontological questions, but neither exhausts the field. Ontology also asks whether events are entities, whether properties can exist independently of their instances, how social institutions depend on collective practices, and what makes a person or organization persist over time.

Four Dimensions of an Ontology

Inventory: what does the theory admit?

Every theory relies on some account of what it takes to be real or indispensable to explanation. A physical theory may quantify over fields and particles. A theory of mind may include experiences, beliefs, intentions, or functional states. A social theory may refer to institutions, norms, roles, and collective agents.

Using a noun does not automatically settle the ontology. A theory may speak conveniently about an “average consumer” without treating that consumer as an individual entity. The ontological question is whether the theory is committed to such a thing as part of its account of the world.

This is why ontology includes the study of ontological commitment: what must exist, or be treated as existing, if a set of claims is true?Stanford Encyclopedia of Philosophy: Logic and Ontology

Categories: what kinds of entities are there?

Ontologies distinguish among general kinds:

There is no universally accepted final table. The dispute is partly about whether these divisions are discovered in reality, imposed by language and cognition, or produced through an interaction between the world and a purpose of inquiry.

Identity: what makes something the same thing?

Identity through change is a central ontological problem.

• A person normally remains the same person after changing a name.
• A corporation may survive a complete change of staff.
• An essay may retain its identity across revisions because a version history connects them.
• A product may acquire new packaging without becoming a new model.
• A machine-learning system may be called the same product after an update even when the deployed model artifact has changed.

The answer depends on the kind of entity under discussion. Legal continuity can matter for a corporation, biological continuity for an organism, provenance for a document, and a stable identifier plus version policy for a software artifact. A matching name is not enough to establish identity, and a changed property does not by itself establish a new entity.

Dependence and structure: what relies on what?

Some entities appear capable of existing independently; others exist only through a supporting structure.

A purchase depends on participants, an object of exchange, a time, and a transaction. A currency depends on institutional rules and practices. A software process depends on code and an execution environment. A fictional character can exist as an object of a story and discourse without being an actual historical person.

These examples separate three claims that are often conflated:

an expression refers to something
≠ a model represents it as an entity
≠ a corresponding entity exists in the actual world

An ontology has to state which level it is describing.

Ontology, Essence, Epistemology, and Metaphysics

Ontology overlaps with several neighboring subjects but answers a distinct question.

Essence concerns what an entity must be in order to count as the kind of thing it is. Ontology also asks whether that kind exists, how its instances persist, and how it relates to other kinds.

Epistemology concerns knowledge, evidence, and justification. Whether intentions are mental states is an ontological issue. How an AI system can infer an intention from language and behavior is an epistemological and methodological issue. A complete inquiry needs both: an account of the target and an account of the evidence for finding it.

Metaphysics is usually broader. It includes ontology but also ranges over causation, time, modality, freedom, grounding, and the overall structure of reality. The borders vary across traditions, so ontology should not be treated as a universally fixed department inside metaphysics.

Computational Ontology: Making Commitments Explicit

In knowledge representation, ontology names an engineered artifact as well as a philosophical inquiry.

Thomas Gruber’s influential formulation describes an ontology as a specification of a conceptualization. The practical aim was knowledge sharing: agents need an explicit vocabulary defining the classes, relations, functions, and other objects used in a shared domain.A Translation Approach to Portable Ontology Specifications

A computational ontology can therefore be understood as:

an explicit, inspectable account of the entity types, properties, relations, and constraints that a system recognizes within a domain.

Consider a commerce domain:

Customer ──has──> Need
Brand ──offers──> Product
Merchant ──sells──> MarketOffering
MarketOffering ──realizes──> Product
Customer ──buys──> MarketOffering
Customer ──uses──> Product
Order ──contains──> MarketOffering

The arrows do not settle the model. Designers still have to decide:

• Is a product a design, a model, a service capability, or an individual item?
• Does price belong to the product, the market offering, or a time-bounded quote?
• Are two merchants selling one product or two distinct offerings?
• Does an offering cease to exist when it is withdrawn?
• Does a purchase establish that a customer’s need was met?

These are ontological choices with operational consequences. They determine how data can be joined, which inferences are valid, and what an automated system can act upon.

What an Engineering Ontology Contains

Classes and individuals

A class represents a kind such as Person, Organization, or Product. An individual represents a particular member of one or more classes. In OWL 2, classes can be understood as sets of individuals, while properties express relations among individuals or between individuals and data values.W3C: OWL 2 Primer

Class membership need not be exclusive. One individual might be both a customer and an employee. A useful ontology states when multiple classifications are compatible and when classes are disjoint.

Properties and relations

Properties express information such as a name, status, date, or quantity. Relations connect entities through claims such as worksFor, owns, published, or purchased.

Whether a value should become an entity depends on the required reasoning. A city can be stored as a string if nothing else refers to it. It should usually become an identified entity if the system must connect it to regions, offices, time zones, routes, or multiple records.

Axioms and constraints

An ontology can state more than permitted vocabulary. It can encode claims that support validation and inference:

• every order has a buyer;
• a cancelled order cannot simultaneously be awaiting payment;
• every researcher is an employee;
• people and organizations are disjoint kinds;
• a relation is symmetric, transitive, functional, or inverse to another relation.

Formal languages differ in what they express and in the assumptions they make. An ontology should therefore be judged partly by the reasoning task it must support, not by the size of its vocabulary.

Identity, provenance, and time

Operational systems encounter aliases, duplicate records, namesakes, revisions, and claims that change over time. An ontology alone does not solve entity resolution. It has to work with identifiers, provenance, temporal data, and confidence to answer:

• Do two records refer to the same entity?
• When was a property true?
• Which source supports the assertion?
• Is the assertion observed, inferred, or supplied by a user?

Without these distinctions, a clean graph can still represent an incorrect world.

An Ontology Is More Than a Taxonomy or Schema

A taxonomy arranges categories, often in a hierarchy. It tells a system that a researcher is a kind of employee or that a phone is a kind of product.

A database schema specifies storage: tables, columns, keys, and types.

An ontology addresses semantics: what the records represent, which relations are meaningful, what identity means, and which constraints or inferences follow.

The three can overlap, but they are not interchangeable:

taxonomy: how kinds are classified
schema: how data is stored
ontology: what the modeled things mean and how they may relate

A relational database can implement an ontology. A graph database can also lack one. The storage technology does not settle whether the domain has been conceptually defined.

Does a Large Language Model Have an Ontology?

Language models learn many implicit categories and relations from text. They can produce statements such as “a company releases a product” and often resolve an ambiguous word from context. This behavior shows useful conceptual organization.

It does not establish that the model contains one explicit, stable, internally consistent ontology. Information about a company or product is distributed across parameters and contextual activations. Different prompts can elicit incompatible classifications or identity assumptions.

Typical failures include:

• treating a brand and its owning company as one entity;
• treating a product and a purchasable offering as one object;
• confusing a name with a verified identity;
• confusing a discourse entity with an actually existing entity;
• applying different identity criteria across contexts.

For reliable action, an AI system often needs several layers:

language model
+ domain ontology
+ current entity and state data
+ provenance and time
+ permissions and action rules

The language model interprets variable expression. The ontology constrains the objects and relations the system recognizes. Data supplies current instances and states. Authorization determines which actions are permitted. Understanding “update the essay” is insufficient until the system identifies the essay, language edition, revision, destination, and authority to publish.

Are Ontologies Discovered or Designed?

Every practical ontology is selective. The same cup of coffee can be modeled as:

• a physical object with a temperature and container;
• a sellable item with a price and inventory status;
• a nutritional object with ingredients and calories;
• an experience with flavor and context;
• an environmental object with a supply chain and emissions.

Different purposes can justify different boundaries. Purpose, however, does not make every model equally adequate. Merging two different people, treating a temporary quote as a permanent product price, or discarding provenance can produce false claims and harmful actions.

A useful ontology can be examined through seven questions:

1. What problem is the ontology intended to support?
2. Which entities does it recognize, and which does it omit?
3. Are identity conditions clear for each important kind?
4. Are objects, events, states, properties, and relations kept distinct?
5. Do the axioms support the required inferences without producing contradictions?
6. Can the model represent time, provenance, and uncertainty where they matter?
7. Can old data still be interpreted when the ontology changes?

An ontology is a map of what a system takes to exist and how those things fit together. It is answerable to reality, but it should not be confused with reality itself.

Entities Are Local; Ontology Is the Framework

An entity is something a system can identify, describe, relate, track, or act upon. An ontology states:

• what can count as an entity;
• which kinds of entity exist in the domain;
• what makes entities identical or distinct;
• which properties they can bear;
• which relations and events they can enter;
• which conditions constrain their existence and change.

The relationship can be summarized in one sentence:

An entity is one recognized item; an ontology is the system’s account of what kinds of items may be recognized and how they are organized.

A person can understand a room from a description, anticipate how a device will respond, imagine an alternative future, or infer a conclusion without directly manipulating the world. These abilities require some usable representation of a situation.

A mental model is a selective representation of objects, relations, mechanisms, or possible states that a thinker can inspect or manipulate in order to understand, infer, predict, or act.

The broad process is:

complex situation
→ selective representation
→ organized relations and mechanisms
→ inference or simulation
→ judgment and action
→ correction from evidence

A model is useful because it leaves things out. That selectivity is also the source of its limits.

The Scientific Term and the Popular Toolkit

In cognitive science, mental model names a family of hypotheses about internal representation and reasoning. Philip Johnson-Laird’s theory proposes that people reason by constructing representations of possible situations rather than only by applying formal syntactic rules. Cambridge University Press: Mental Models and Reasoning

In business and self-education, mental models often refers to a toolkit: opportunity cost, Bayesian updating, feedback loops, margin of safety, inversion, or second-order effects. That use is broader. It combines internal representations, scientific models, decision rules, frameworks, and heuristics.

The two usages overlap but should not be treated as identical. A theory about how reasoning works is different from a curated list of techniques for improving judgment.

A Model Is Not a Copy of Reality

Models perform at least four operations:

1. Selection: identify objects relevant to the task;
2. Compression: omit detail;
3. Organization: establish categories, relations, order, or mechanism;
4. Simulation: vary conditions and examine what follows.

A transit map distorts geographical distance but preserves connections useful for travel. A topographic map preserves different relations and serves a different task. Neither is simply the territory at smaller scale.

The philosophy of science describes an important use of models as surrogative reasoning: investigators learn about a target system by constructing and manipulating a model of it. Stanford Encyclopedia of Philosophy: Models in Science

Model quality therefore depends on a purpose. The relevant questions are what the model preserves, what it suppresses, and whether those choices support the intended inference.

Neighboring Concepts

Concept

A concept supports recognition and classification. A model organizes several concepts and relations. User, need, product, and outcome are concepts; a representation of how a product changes a user’s situation begins to form a model.

Theory

A theory is normally a systematic set of claims and explanations with evidential commitments. A model can instantiate a theory, simplify it, or represent a particular case. One theory can support multiple models.

Framework

A framework identifies dimensions or questions through which to inspect a subject. A model also represents how elements relate or change. A list of people, process, and technology is a framework until their interactions are specified.

Method

A model represents a structure or mechanism. A method specifies a procedure. A causal graph is a model; a randomized experiment is a method for identifying causal effects.

Heuristic

A heuristic reduces search or computation through a rule of thumb. It may arise from a model but need not represent the mechanism that makes the rule successful.

Decision model

A decision model is one functional class of model:

mental or analytical model:
How is this situation structured, and what might follow?

decision model:
Given evidence, ends, and constraints, which action should be selected?

Six Functions of Models

This classification groups models by the work they perform. A single model can serve more than one function.

1. Representation and structure

These models answer: What exists in the problem, where is the boundary, and how are the parts related?

Maps, hierarchies, networks, process diagrams, system boundaries, and representations of a business model belong here. They determine what can be noticed and discussed before any causal claim or decision is made.

2. Explanation and causation

These models answer: Why did this happen, and through what mechanism?

Causal chains, incentive structures, supply and demand, bottlenecks, path dependence, and feedback loops organize explanatory relations. Their characteristic failure is to turn a plausible story or correlation into an asserted mechanism without a test.

3. Inference and belief revision

These models answer: What should be believed from these premises or this evidence?

Deduction, induction, abduction, Bayesian updating, base rates, counterfactual reasoning, and falsification supply different structures for inference.

Bayesian updating primarily changes belief:

prior assessment
+ evidence
→ posterior assessment

It becomes a decision model only after consequences, values, constraints, and an action rule are added.

4. Prediction and simulation

These models answer: What could happen if conditions changed?

Scenario models, sensitivity analysis, system dynamics, second-order effects, trend models, and Monte Carlo simulation belong here. A useful predictive model exposes ranges, assumptions, uncertainty, and the conditions under which its forecast should no longer be trusted.

5. Evaluation and decision

These models answer: How should feasible actions be compared?

Opportunity cost, expected utility, multi-criteria analysis, minimax rules, decision trees, margin of safety, reversibility, and exploration versus exploitation connect beliefs about the world to choice.

They necessarily introduce value. What counts as benefit, which loss is intolerable, and whose outcomes matter cannot be derived from probability alone.

6. Action and feedback

These models answer: How will a decision be executed, observed, and corrected?

OODA, PDCA, hypothesis–experiment–feedback cycles, control loops, iterative trials, and after-action review distinguish plans from execution and execution from verified effect.

Without feedback, a model remains an imagined relation to the world. When consequences update the next representation and action, the system can learn.

How the Functions Connect

representation
What are we dealing with?

explanation
Why does it behave this way?

inference
What should the evidence change?

prediction
What might happen under other conditions?

decision
Which action should receive priority?

feedback
Did action change reality as expected?

This is a functional map, not a mandatory sequence. Failed action can force a new representation. Conflicting values can cause the option set to be redesigned. A forecast error can expose a weak causal mechanism.

One Problem, Several Models

Consider whether to discontinue a product.

1. A structural model defines the product, users, market, costs, and dependencies.
2. A causal model explains why adoption or retention stalled.
3. Base rates and new evidence revise confidence in competing explanations.
4. Scenarios estimate the consequences of continuing, narrowing, selling, or stopping.
5. Opportunity cost, downside, and reversibility compare actions.
6. A staged experiment and feedback loop test the commitment.

No celebrated model substitutes for the whole chain. Selecting a model is itself a diagnosis: is the current uncertainty about facts, mechanism, prediction, values, or execution?

Why Popular Mental Models Look Like Decision Models

Business and self-improvement writing is organized around practical questions: What should I do? What should I do first? When should I stop? How can I reduce error? That selection pressure favors models close to action.

Yet serving a decision is not the same as being a decision model:

• first-principles analysis reconstructs assumptions and problem boundaries;
• systems thinking identifies interaction and feedback;
• causal models estimate what intervention might change;
• Bayesian updating revises belief;
• opportunity cost compares actions;
• OODA connects observation, orientation, decision, and action.

Decision models are the action-selection subset of a wider ecology of representations, explanations, inferences, predictions, and feedback systems.

How to Evaluate a Model

Identify its target

What situation, system, or class of problems does it represent? Where is the boundary?

State its function

Is it describing, explaining, predicting, evaluating, deciding, or controlling? A classification model does not establish causation merely because its categories are useful.

Expose assumptions and omissions

Which relations are fixed? What has been left outside? Under which conditions should the model fail?

Demand checkable implications

A model that can accommodate every possible outcome cannot learn much from evidence.

Compare with simpler alternatives

Complexity should earn its cost by changing a judgment or improving prediction. More variables and terminology do not by themselves bring a model closer to reality.

Update from contact with the world

When prediction or action fails, does the model change, or does it acquire an endless list of exceptions?

Philosophical Limits

Human beings do not encounter a complete, uninterpreted reality from nowhere. Perception, language, concepts, and measurement already select and organize. This does not imply that all models are equally good.

Prediction failure, blocked action, counterexamples, measurement, and other people’s experience constrain representation. Multiple models can be useful without becoming immune to evidence.

Models are also not automatically value-neutral. What enters the model, which outcomes are measured, and whose risk is represented can determine which actions appear reasonable.

Mental models are therefore not a collection of clever labels to memorize. They are revisable structures for reducing complexity while preserving relations that matter to a task.

Models let finite minds approach reality through selective representation. Decisions ask those minds to act, and accept responsibility, before the representation can ever become complete.

Further Reading

• What Decision-Making Requires: Judgment, Choice, and Commitment
• Mental Representation: How the Mind Represents a World

Sources

• Cambridge University Press: Mental Models and Reasoning
• The Mental Models Global Laboratory: Publications
• Stanford Encyclopedia of Philosophy: Mental Representation
• Stanford Encyclopedia of Philosophy: Models in Science
• Stanford Encyclopedia of Philosophy: Thought Experiments

A useful working definition is:

A mental representation is a content-bearing state through which a cognitive system presents an object, property, relation, event, or possible situation.

This definition belongs to a representational approach to mind. It does not assume that every cognitive process requires a detailed internal model, or that a representation must resemble what it represents. It marks a problem: how can a mental state carry content and make that content available to perception, memory, inference, imagination, or action?

What Makes a State Representational?

Mere correlation is not enough. Smoke correlates with fire, and a tree ring correlates with a period of growth, but neither must be a mental representation. Within a cognitive system, a representational state typically has several connected features:

• it is about a target or state of affairs;
• it has content that can be used by other processes;
• it plays a role in recognition, prediction, inference, memory, or control;
• in many cases, it can represent its target incorrectly.

A belief that the door is locked represents the door as being in a particular condition. A visual experience represents edges, colors, surfaces, and locations. A remembered route preserves relations among places. A motor plan presents a possible sequence of bodily movement.

The philosophical term intentionality names this directedness or aboutness of mental states. It should not be confused with intention, the practical commitment to perform an action. Stanford Encyclopedia of Philosophy: Intentionality

Content, Vehicle, Target, and Use

Four questions keep the subject clear.

The vehicle and the content are different. Marks on a paper map are the vehicle; the terrain and spatial relations presented by the map are its content. A neural pattern may implement a representation without literally containing the represented object.

The same content can also be carried in different formats. The prospect of rain may occur as an inner sentence, an image of dark clouds, a probability estimate, or a practical readiness to take an umbrella. The formats need not be interchangeable: a diagram makes some spatial relations easy to inspect, while a sentence makes some logical relations easier to state.

Representation Is Not an Inner Picture

The picture metaphor is attractive because visual imagery is familiar. It becomes misleading when applied to every kind of thought.

Representations may include:

• perceptual organizations of the current environment;
• images and spatial models;
• concepts and categories;
• proposition-like contents that can be true or false;
• expectations about causes and outcomes;
• affective appraisals of danger or value;
• motor plans and action-oriented states;
• models of other people’s beliefs and goals.

Even visual representation is not a complete internal photograph. Perception selects and organizes information according to attention, learned categories, bodily capacities, and current goals. A carpenter, a mover, and a tired traveler may represent the same chair differently because different properties matter to what each person is doing.

Representation is therefore selective. It reduces information in order to make relevant relations available. That compression is a source of both intelligence and error.

Absence, Possibility, and Fiction

One reason to posit mental representation is that thought is not confined to what is currently present.

People can represent:

• a childhood home that no longer exists;
• tomorrow’s weather;
• an unbuilt bridge;
• a fictional detective;
• a counterfactual choice;
• an impossible geometric or logical condition.

The represented object does not have to exist. What exists now is the cognitive state and its role in thought. This separation between a present state and its content makes memory, planning, fiction, hypothesis, and counterfactual reasoning possible.

It also creates a difficult question: if no unicorn exists, what makes one thought a thought about unicorns rather than about some other nonexistent thing? A theory of representation must explain the determination of content, not merely point to an external object.

Misrepresentation Is a Test Case

A perceptual illusion, false belief, distorted memory, or failed prediction represents the world as being a way it is not.

Misrepresentation matters because it prevents a circular definition. If a state represented a snake only when a snake caused it, a person could never mistake a rope for a snake. Yet such mistakes are ordinary.

A good account must explain both success and specific error:

• why a state counts as representing a snake;
• why it can occur when no snake is present;
• why the system treats the state as relevant to fear, attention, and action;
• how later evidence can correct it.

The capacity to detach content from the immediate environment supports imagination and planning, but it also allows hallucination, prejudice, and false belief. These are consequences of the same general capacity, not separate additions to it.

Formats of Mental Representation

Cognitive scientists have proposed many representational formats. Some are sentence-like and compositional: they contain parts that combine according to rules. Others preserve analog or spatial relations. Still others are distributed across patterns of activity rather than stored as discrete symbols.

The relevant contrast is often task-dependent.

No single format has to explain every cognitive achievement. Human cognition may recruit several formats and translate among them imperfectly.

Concepts and Compositional Thought

Concepts help a thinker classify things, recognize recurring structures, and form new judgments. The concept bird can occur in thoughts about sparrows, migration, extinction, or flight even when the particular examples differ.

One influential view treats concepts as mental representations that function as constituents of thought. Other theories treat concepts as abilities, such as the ability to discriminate and infer, or as abstract objects that are not identical with events in any individual mind. Stanford Encyclopedia of Philosophy: Concepts

The broader category of mental representation should therefore not be identified with concepts alone. A spatial image, a sensory expectation, and a motor plan may be representational without functioning as concepts. Nor does the usefulness of a representational model settle the metaphysical question of what concepts ultimately are.

Language and Nonlinguistic Representation

Natural language gives thought durable labels, explicit syntax, and public expression. It lets people preserve distinctions, combine claims, compare reasons, and submit private judgments to social criticism.

But mental representation is broader than natural language. Infants and nonhuman animals can track objects, locations, quantities, and agents without possessing an adult language. Adults rely on facial recognition, musical expectation, spatial imagery, and practical skill that may resist complete verbal description.

The language-of-thought hypothesis proposes that some cognition uses an internal representational system with syntactic and compositional structure. Even if such a system exists, it would not follow that all thought is silent English, Chinese, or another public language. Stanford Encyclopedia of Philosophy: The Language of Thought Hypothesis

Language and representation therefore interact in both directions. Prelinguistic representations support language learning; language then reorganizes attention, memory, abstraction, and deliberate reasoning.

Representation for Action

A mind does not represent the world solely to produce detached descriptions. Representation also guides intervention.

Action-oriented states may encode what can be reached, avoided, grasped, followed, or changed. A driver needs to register that the distance to another car is closing dangerously. Exact measurements of every visible surface would add information without improving control.

A practical loop looks like this:

perceive a situation
→ represent task-relevant relations
→ anticipate possible outcomes
→ prepare and perform an action
→ use feedback to revise the representation

This loop makes representation dynamic. An action changes the available information, and the new information changes the next action. Some cognitive work may be distributed across brain, body, tools, and surroundings rather than completed inside the head before movement begins.

Where Is Mental Content?

Neuroscience can identify activity associated with recognition, memory, spatial coding, and decision. Computational models can explain how information is encoded and transformed. Functional accounts can show how a state contributes to successful behavior.

None of those descriptions alone answers why a state has one content rather than another. Location, causal role, learning history, biological function, and relations to the environment may all matter.

Internalist theories give primary weight to conditions within the individual. Externalist theories argue that at least some content depends partly on relations to objects, social practices, or historical environments outside the thinker. On an externalist account, two internally similar states need not have identical content if the subjects stand in different relations to the world.

The question “Where is the representation?” is therefore partly spatial, but it is also explanatory. It asks what realizes the state, what fixes its content, and what system makes that content usable.

Do Minds Need Internal Representations?

Classical cognitive science often explains cognition through representational structures and procedures that operate on them. This approach is powerful for memory, planning, inference, and thought about absent possibilities. Stanford Encyclopedia of Philosophy: Cognitive Science

Embodied, ecological, and enactive approaches challenge the idea that intelligent activity always requires a rich internal reconstruction of the world. Skilled movement can exploit stable environmental cues and continuous sensory feedback. Written notes, diagrams, and instruments can carry information that a person does not need to reproduce internally. Stanford Encyclopedia of Philosophy: Embodied Cognition

The dispute becomes more productive when it is divided by task:

• remembering an absent event and planning several years ahead appear to require some way of carrying unavailable content;
• rapid sensorimotor coordination may use sparse, action-specific states;
• external artifacts can become parts of a wider cognitive system;
• calling every causally useful state a representation risks making the concept empty.

The question is not only whether representations exist, but which tasks require which kinds of representation at which level of detail.

Mental Representation and AI Representation

Machine learning systems transform inputs into vectors and intermediate activation patterns. Researchers call these internal states representations because they preserve relations that support classification, prediction, generation, or control.

The comparison with human minds is informative but limited. Both biological and artificial systems can compress information, abstract across cases, preserve some relations while losing others, and produce systematic errors. Yet functional similarity does not establish identical mental status.

Three questions should remain separate:

1. Structural: Does the system form internal states with reusable organization?
2. Functional: Do those states reliably support inference, prediction, or action?
3. Phenomenal and psychological: Are those states part of a subject’s conscious or lived mental content?

Model analysis can produce evidence about the first two. It does not by itself show that a model has a point of view, conscious experience, bodily needs, or human-like understanding. The term representation can be legitimate in both fields while carrying different explanatory commitments.

A Working Conclusion

Mental representation is best treated as a structured research problem rather than an inner object that everyone already understands.

It connects four questions:

What is represented?
What carries the representation?
In what format is the content organized?
How does the state contribute to cognition and action?

Representations allow a mind to respond to more than the immediate stimulus. They support memory, imagination, planning, inference, and coordination. Their selectivity also makes distortion and error possible.

Whether every form of cognition needs internal representation, whether content is fixed entirely inside the individual, and whether artificial representations can become genuinely mental remain open disputes. Any clear account should state which kind of representation, which task, and which explanatory level it is addressing.

Sources

• APA Dictionary of Psychology: Mental Representation
• Stanford Encyclopedia of Philosophy: Mental Representation
• Stanford Encyclopedia of Philosophy: Intentionality
• Stanford Encyclopedia of Philosophy: Concepts
• Stanford Encyclopedia of Philosophy: The Language of Thought Hypothesis
• Stanford Encyclopedia of Philosophy: Cognitive Science
• Stanford Encyclopedia of Philosophy: Embodied Cognition

Series: Thinking, Intention, and Action (1/4). This article explains how human thought forms judgment and action. Next: How AI Systems Reason and Act

What Is Thinking?

The word thinking can name an activity, a capacity, a style, or a sequence of mental contents. These uses overlap, but they should not be collapsed.

A useful working definition is:

Thinking is the activity through which a person organizes, transforms, relates, simulates, and evaluates information in order to form understanding, judgment, possibility, plans, or directions for action.

Thinking draws on perception, memory, emotion, language, concepts, bodily states, and prior knowledge. Its products include beliefs, questions, explanations, images, decisions, and intentions.

A simplified cycle looks like this:

encounter a situation
→ represent relevant features
→ compare, combine, infer, or simulate
→ form a judgment or possible action
→ receive feedback
→ revise the representation

Actual thought is rarely a neat sequence. Perception, memory, feeling, expectation, and action continually constrain one another.

Thinking, Thought, Reasoning, Mind, and Cognition

Several English terms mark different parts of the subject.

Thinking is therefore neither the whole mind nor a synonym for reasoning. Reasoning is one kind of thinking. Cognition is broader than deliberate thought because perception, memory encoding, language processing, and learned recognition can occur without a person explicitly working through a problem.

The borders are not sharp. When a person enters a room and recognizes a door, cognition is already at work. When the person asks whether a large table will fit through that door and mentally rotates it, the activity more clearly counts as thinking.

What Does Thinking Operate On?

Thinking is often described as operating on representations. A representation carries content: it is about an object, relation, event, possibility, or state of affairs.

The relevant format may include:

• words and propositions;
• concepts and categories;
• visual or spatial images;
• sounds and rhythms;
• motor patterns;
• bodily and emotional signals;
• models of situations;
• expectations about other people;
• simulations of possible futures.

Thinking that it may rain tomorrow might involve an inner sentence, an image of dark clouds, a numerical forecast, or simply a readiness to carry an umbrella.

Representational theories treat thinking, reasoning, and imagining as sequences of intentional mental states. They help explain how thought can be about things that are absent or merely possible. But the nature and necessity of internal representations remain disputed, especially by embodied and enactive approaches to mind. Stanford Encyclopedia of Philosophy: Mental Representation

Thinking Is an Activity of Transformation

Representing something is not yet enough. Thinking changes the organization of what is represented.

Common operations include:

• categorizing;
• comparing;
• decomposing and combining;
• ordering;
• abstracting;
• associating;
• negating;
• inferring;
• imagining;
• estimating;
• evaluating;
• replacing one model with another.

Consider the concept bird. A thinker does more than retain images of several birds. The thinker can extract common structure, recognize typical and atypical members, revise an expectation about flight, and explain why penguins remain birds.

The American Psychological Association defines thinking broadly enough to include the manipulation or experience of ideas, images, and other elements of thought, and includes processes such as imagining, remembering, problem solving, free association, and concept formation. APA Dictionary of Psychology: Thinking

Thinking Is Constrained

Human thought is not unrestricted computation. It occurs within limits set by:

• attention;
• working memory;
• available time;
• prior knowledge;
• learned categories;
• language;
• emotion;
• bodily condition;
• social expectations;
• current goals;
• the information environment.

A tired, frightened, angry, or socially threatened person may notice different evidence and retrieve different memories from the same situation.

Limits do not merely cause errors. They also make thought possible. Concepts compress experience. Heuristics reduce search. Habits prevent every action from becoming a fresh planning problem. The question is not whether thinking uses shortcuts, but when a shortcut remains appropriate and when it hides a relevant difference.

Thinking and Consciousness Are Not the Same

Consciousness concerns what is present in subjective experience. Thinking concerns how information is organized and transformed.

People can consciously rehearse an argument or visualize a route. Yet many processes that shape the resulting thought are not directly accessible: pattern recognition, memory retrieval, linguistic parsing, affective appraisal, and automatic association.

Often a person becomes conscious of an answer without becoming conscious of the process that produced it.

awareness of a thought
≠
awareness of how the thought was generated

Metacognition improves the monitoring and regulation of thought. It does not make the mind fully transparent to itself.

Thinking and Language Are Interdependent but Distinct

Language gives human thinking extraordinary resources. It allows people to name distinctions, stabilize concepts, combine propositions, preserve intermediate steps, communicate reasons, and construct possibilities far removed from immediate experience.

Language also externalizes thought. A vague impression can be written as a claim, inspected, criticized, and revised.

But not all thought takes the form of sentences. Visual imagery, spatial transformation, musical expectation, motor planning, facial recognition, and some forms of problem solving can proceed without ordinary verbal expression.

A cautious conclusion is:

Language organizes, extends, and externalizes human thinking, but does not exhaust it.

The language-of-thought hypothesis proposes that some thinking occurs in an internal representational system with combinatorial structure. It offers an explanation of the productivity and systematicity of thought, but remains a theoretical position rather than a neutral definition of all thinking. Stanford Encyclopedia of Philosophy: The Language of Thought Hypothesis

Concepts Make Reusable Thought Possible

Concepts allow different objects, events, and situations to be treated as instances of a common kind. They support categorization, inference, memory, learning, and decision-making. This makes them plausible building blocks of many thoughts, although philosophers disagree over whether concepts are mental representations, abilities, or abstract objects. Stanford Encyclopedia of Philosophy: Concepts

Concepts perform at least two functions.

They compress experience

Concepts such as product, need, cause, and intention gather many cases into reusable structures.

They establish distinctions

The concept of causation allows a thinker to distinguish temporal sequence, association, common causes, reverse causation, and feedback.

Concepts can also distort. A broad category may erase relevant differences. A rigid category may divide a continuous process at the wrong place. A familiar label may create the illusion that the underlying phenomenon has already been explained.

Thinking Is Not Logic

Thinking is a psychological activity. Logic supplies standards for relations among claims.

human thinking ≠ formal logic

People think through images, analogies, emotions, habits, narratives, and intuitions as well as explicit arguments. Some of these processes generate insight before a valid argument is available. Others generate systematic error.

Logic can ask:

• What are the premises?
• Does a term retain the same meaning?
• Does the conclusion follow?
• Are the claims mutually consistent?
• Has an alternative been excluded without reason?

Logic cannot by itself establish that the premises are true or that a goal is worth pursuing.

Consider:

Every high-value product has many users.
This product has many users.
Therefore this product has high value.

The inference affirms the consequent. Even after the form is repaired, empirical questions remain: what counts as value, how was use measured, and what alternative causes could explain the audience size?

Logic is therefore one normative instrument for improving thought, not a complete model of the human mind.

Conceptual Thinking and Logical Thinking

Conceptual thinking and logical thinking are closely related, but they do different work.

Conceptual analysis asks:

• What does the central term mean here?
• Is one word carrying several concepts?
• Which conditions define or support the category?
• Which neighboring concepts must be separated?
• Does the present case belong under the concept?

Logical analysis asks:

• How do premises support a conclusion?
• Is the inference valid or probabilistically strong?
• Are any claims inconsistent?
• Does the conclusion exceed the evidence?
• Does a counterexample defeat the generalization?

Take the claim:

The user bought the product, so the product satisfied the user’s need.

Conceptual analysis separates purchase, use, outcome, and need satisfaction. Logical analysis then shows that the occurrence of a purchase does not entail a successful outcome.

conceptual clarification
→ stabilizes the units of reasoning

logical evaluation
→ tests the relations among those units

Clear concepts do not guarantee a sound argument. Sound-looking inference cannot repair a shift in meaning.

Forms of Thinking Are Overlapping Tools

Thinking can be classified in many ways, but the categories are not separate mental compartments.

Concrete and abstract

Concrete thought remains close to perceptible objects and particular situations. Abstract thought handles kinds, relations, rules, and structures that cannot be directly perceived.

Intuitive and analytic

Intuitive thought is often fast, automatic, and holistic. Analytic thought is usually slower, attention-demanding, and easier to express in steps. Expertise can make a sophisticated judgment feel immediate, so speed alone does not reveal whether a judgment is shallow or well trained.

Deductive, inductive, and abductive

• Deduction derives what follows from stated premises.
• Induction extends from observed cases to a broader pattern.
• Abduction proposes an explanation for an observed result.

Causal, counterfactual, and systemic

• Causal thinking asks which factors change an outcome through which pathways.
• Counterfactual thinking asks what might happen under an alternative condition.
• Systems thinking examines interaction, feedback, delay, and aggregate behavior.

Critical, creative, and normative

• Critical thinking tests evidence, assumptions, sources, and rival explanations.
• Creative thinking reorganizes existing material into new possibilities.
• Normative thinking asks what should be done and what reasons favor an action.

A difficult decision may require all of these at once.

Human Thinking Is Embodied, Affective, and Social

The image of a solitary mind manipulating neutral propositions captures only part of human thinking.

Embodied

Fatigue, pain, hunger, stress, skill, posture, and possibilities for action shape what can be noticed and considered. Thought is implemented by a living organism, not by a detached viewpoint.

Affective

Emotion changes attention, memory retrieval, risk assessment, and readiness to act. Emotion can mislead, but it also registers importance, threat, loss, and value. Removing emotion would not leave a complete decision-maker behind.

Goal-directed

People do not process every available fact. Needs and goals determine which features become relevant.

Social and distributed

Language, education, institutions, tools, and other people shape the structures through which individuals think. Complex thought may be distributed across:

a brain
+ a body
+ language
+ notes and diagrams
+ software and data
+ other people
+ inherited social knowledge

External tools do more than store completed thoughts. Writing, drawing, calculating, and conversation can change the thought that becomes possible.

Reflexive

Humans can make their own beliefs and methods into objects of inquiry: Why do I believe this? Did the meaning of a term shift? What evidence would change my mind?

Self-protective

Reasoning may serve accuracy, but it can also protect identity, desire, group membership, and emotional stability. A polished argument can still be motivated by selective attention or defensive interpretation.

Is Thinking for Knowledge or for Action?

It serves both.

People build models of the world in order to understand what is the case, but also to anticipate events, avoid danger, coordinate with others, choose actions, and correct unsuccessful behavior.

perceive a situation
→ form an interpretation
→ anticipate outcomes
→ choose an action
→ observe feedback
→ revise the interpretation

Usefulness and truth should not be identified. A false belief may be temporarily useful. An accurate belief may offer no immediate advantage.

Thinking therefore faces at least two standards:

• epistemic: Is the judgment adequately supported and closer to the truth?
• practical: Does it guide effective action toward an end worth pursuing?

Science, logic, ethics, and philosophy examine different parts of these standards.

What Makes Thinking More Mature?

Mature thinking is not simply more complicated thinking. It can be tested through a sequence of questions:

1. What exactly is the object of thought?
2. Which concepts organize it?
3. Are the conceptual boundaries clear?
4. Which claims are observations, and which are interpretations?
5. Which premises remain unstated?
6. Is the inference deductive, inductive, abductive, causal, or analogical?
7. Which rival explanations remain possible?
8. How are emotion, interest, and identity shaping attention?
9. What evidence would require revision?
10. Is the conclusion descriptive, causal, evaluative, or normative?
11. What action follows, if any?
12. Did the result of action revise the original model?

Mature thinking makes distinctions where they matter, tests what can be tested, preserves uncertainty where evidence is limited, and allows consequences to correct the model that produced the action.

Conclusion

Human thinking is more than formal inference and more than a stream of inner speech.

It is a dynamic activity in which perception, memory, concepts, language, imagery, emotion, reasoning, goals, bodily conditions, social resources, and action participate together.

represent a world
→ transform and relate representations
→ form judgments, possibilities, plans, and actions
→ use consequences to revise the model

Cognition is broader than thinking. Logic provides standards for some transitions in thought. Language organizes and externalizes thought. Concepts make classification and reusable inference possible. Consciousness presents some mental contents in experience without revealing every process that produced them.

Conceptual thinking is one method for improving this larger activity: clarify the units of thought, separate neighboring meanings, and then test the inferences among them. It is valuable precisely because human thinking also includes nonconceptual perception, imagery, emotion, creativity, social interaction, and the organization of action.

A Word Is Not a Concept

A word is a unit of language. A concept is something used in categorization, thought, inference, memory, learning, and decision-making.

The English word water is not identical to the concept of water. The concept includes ways of identifying water, expectations about its behavior, relations to other substances, and inferences that can be drawn about it. The same concept can be expressed in different languages, while a single word can express different concepts in different contexts.

A dictionary is therefore a starting point rather than a complete theory. It records established senses and patterns of use. Conceptual analysis must also ask what is being represented, how the representation is structured, and what work it performs in reasoning.

Philosophy offers no single accepted account of what concepts themselves are. Major approaches treat them as mental representations, cognitive abilities, or abstract objects. Stanford Encyclopedia of Philosophy: Concepts

Parts of Speech Are Grammatical Categories

Nouns, verbs, and adjectives describe how expressions function grammatically.

• Nouns commonly name or refer to people, objects, events, states, and abstractions.
• Verbs commonly express actions, occurrences, changes, and continuing processes.
• Adjectives commonly attribute properties, conditions, degrees, or evaluations.

These patterns do not create a one-to-one map between grammar and reality.

Death is a noun that can refer to an event. Run can be a verb naming an activity or a noun naming an instance, route, or sequence. Red can be an adjective attributing a color or a noun referring to the color itself. Choice is a noun derived from action language.

Grammar tells us how an expression is used in a sentence. It does not by itself settle what kind of entity, occurrence, or structure the expression is about.

What Kinds of Things Can Concepts Represent?

Objects and Entities

Objects are typically treated as things that exist and retain some identity across change: a person, a tree, a phone, or a building.

Not every entity is a natural physical object. A corporation is a social and legal entity constituted by rules, roles, records, assets, and relationships.

Events

Events happen or take place: an arrival, an election, an explosion, a birth, or a death.

Philosophers often contrast objects, which exist, with events, which occur. The contrast is useful but disputed. Objects and events also differ in how they occupy space and time and in how they are identified. Stanford Encyclopedia of Philosophy: Events

Actions

Actions are things agents do: promise, refuse, write, choose, repair, or work.

An action is not merely a bodily movement. Describing something as an action often introduces questions about intention, control, knowledge, ability, and responsibility. A movement that happens to a person and an action performed by that person may look similar while differing in agency.

Processes

Processes unfold over time: learning, aging, erosion, cognition, and economic development.

A process may contain many events without having one natural endpoint. Learning is an extended process; passing a particular examination is an event within a longer history.

States

States obtain or persist for a period: knowing an answer, being unemployed, remaining stable, or feeling anxious.

A state description does not by itself explain its cause, permanence, or value. Calling someone anxious identifies a condition; it does not establish why the condition arose or whether it defines the person.

Properties

Properties are attributed to objects, actions, and states: red, rigid, fragile, free, fair, or positive.

Some properties have agreed measurement procedures. Others depend on interpretive or normative standards. Length and fairness can both be predicated of something, but evidence for each is assessed differently.

Relations

Relations connect two or more terms: resemblance, causation, ownership, kinship, dependence, rights, and obligations.

Many concepts that sound like self-contained qualities are partly relational. Significance, for example, often involves a relation among an event, an interpreter, a context, and a set of concerns.

Quantities, Scales, and Models

Time, probability, price, risk, and efficiency do more than label a thing. They provide ways to order, compare, measure, or infer.

Their everyday uses may differ from their roles in a formal discipline. Saying that an outcome is “very likely” is not yet the same as assigning a probability under a specified model.

Institutions, Rules, and Norms

Money, offices, laws, promises, responsibilities, and justice depend on shared practices, constitutive rules, or normative judgment.

A banknote is a physical object, but its purchasing power cannot be explained by its physical composition alone. Institutional concepts connect material objects with recognized rules and social powers.

Abstract Objects

Numbers, sets, propositions, possibilities, and perhaps concepts themselves are common candidates for abstract objects. They do not occupy ordinary physical space, yet they appear indispensable to mathematics and reasoning.

Whether abstract objects exist independently and how human beings could know them remain major philosophical disputes. Stanford Encyclopedia of Philosophy: Abstract Objects

Events, Processes, Achievements, and States

Language also distinguishes different temporal structures.

These distinctions matter because grammar can hide them. “She is writing the report” describes an incomplete process. “She has written the report” presents the completion. “She knows the answer” describes a state rather than an achievement now in progress.

The metaphysics of events and the linguistic study of aspect use related distinctions, although there is no universally accepted inventory of categories.

Concepts Also Perform Roles in Reasoning

What a concept represents is only one question. A concept may also perform different tasks in an argument.

The same term can move among these roles.

Calling a mood positive may describe affect. Calling an employee’s attitude positive may evaluate conduct. Saying that someone should be positive adds a norm. A conversation can move across all three without marking the transition, turning an observation into a moral demand.

Similar errors occur with facts. A fact is a state of affairs that obtains. A causal explanation is a claim about why it obtains. A recommendation requires further premises about goals and values. Facts constrain recommendations, but they do not generate a complete practical conclusion on their own.

Why Ordinary Sentences Hide Multiple Concepts

Natural language compresses several layers of thought.

Consider the sentence:

The plan failed, so the strategy was wrong.

It contains at least four elements:

• plan: an arrangement of intended actions;
• failed: an event description combined with a standard of success;
• strategy: a system of goals, diagnosis, choices, resources, and trade-offs;
• so: a claimed inferential connection.

One failed plan may result from execution, timing, or chance. It does not by itself establish that the strategic diagnosis was false. The concepts must be separated before the inference can be tested.

Conceptual analysis therefore does more than define isolated nouns. It reveals compressed relations, shifting senses, and missing premises.

A Three-Pass Method for Conceptual Analysis

1. Identify the Linguistic Form

Is the expression functioning as a noun, verb, adjective, or complete proposition? Has an action or property been nominalized and treated as a thing?

2. Identify the Represented Category

Does it concern an object, event, action, process, state, property, relation, quantity, institution, or norm? Are different speakers referring to the same category?

3. Identify the Inferential Role

Is the expression describing, classifying, explaining, predicting, evaluating, prescribing, or guiding action? Has the argument moved from one role to another without an explicit premise?

After these passes, further research can address etymology, historical change, disciplinary definitions, evidence, philosophical disputes, and practical limits.

Conclusion

Concepts do not merely stand for “things,” and nouns do not map neatly onto objects. Thought concerns what exists, what happens, what agents do, what persists, what properties and relations obtain, and what rules or values should guide action.

Part of speech identifies a grammatical role. Ontological category identifies the kind of reality under discussion. Inferential role identifies what an expression is doing in an argument.

Keeping those three levels separate is a practical first step from word definition to conceptual analysis and from conceptual analysis to sound reasoning.