The Problem of Other Minds

Each subject encounters their own experience from the first person and other minds through bodies, behaviour, language, and shared activity. The problem of other minds asks what justifies the inference that anyone else is conscious or minded. The problem is not normally lived as a sceptical crisis; social perception is immediate and practical. It becomes explicit when familiar evidence comes apart, as with non-human animals, severe motor impairment, unfamiliar cultures, or artificial systems.

No solution can require direct access to another's experience without making ordinary knowledge impossible by definition. The relevant standard is fallible public evidence: which patterns best support attribution, which theory connects them to mentality, and how uncertainty should change across cases.

The asymmetry

In one's own case, pain is known through undergoing it. In another's case, pain is known through expression, action, testimony, context, and physiology. This creates an epistemic asymmetry without necessarily creating an ontological gap.

The asymmetry is compatible with reliable knowledge. Scientists know unobservable entities through effects, and people know historical events through traces. The special difficulty is that the association between behaviour and experience appears calibrated from one first-person case: one's own.

Radical scepticism argues that all other bodies could be mindless. A practical theory instead asks whether the mindless hypothesis explains the evidence as well as ordinary attribution and whether the demand for certainty is reasonable.

The argument from analogy

Mill's classic argument has the form:

1. In my case, bodily states and behaviour $B$ are associated with experience $E$.
2. Others exhibit relevantly similar bodies and behaviour $B$.
3. Similar effects probably have similar causes.
4. Therefore others probably have experience $E$.

The argument is inductive and weakens as relevant similarity decreases. Shared biology, development, expressive behaviour, and causal mechanisms make the human case stronger than a comparison based on surface output alone.

The one-case objection says the induction begins from a sample of one. The reply is that evidence is not limited to one correlation. Mechanistic knowledge, evolutionary continuity, intervention, and convergent behaviour support the same model. One's own case supplies the bridge between mechanism and experience; science tests its scope.

Inference to the best explanation

Other-mind attribution can be an inference to the best explanation. Beliefs, desires, perceptions, pain, and intentions explain flexible behaviour, learning, communication, and context-sensitive correction better than a catalogue of movements.

The explanation gains strength when it:

• predicts novel behaviour rather than redescribing observed output;
• unifies action across settings;
• maps onto mechanisms and interventions;
• distinguishes error, pretence, and impairment;
• supports reciprocal interaction and communication.

A mindless duplicate is logically possible on some theories but explanatorily idle if it adds no prediction and merely denies the best-supported cause. Zombie conceivability may challenge metaphysical entailment without defeating empirical inference.

Criteria and expression

Wittgensteinian approaches reject the picture that mental states are private objects inferred from external signs. Pain behaviour is not merely evidence contingently linked to a hidden state; expression belongs to the public criteria through which the concept of pain is learned and applied.

This does not make pain identical with behaviour. Pretence, suppression, and paralysis remain possible. Criteria provide defeasible standards internal to the practice, while symptoms are merely inductive signs.

The approach dissolves some scepticism: certainty about an inner object was never the basis of ordinary attribution. Its limitation is explaining unfamiliar cases where established criteria conflict or where new architectures require extending the practice.

Theory theory

The theory theory says people attribute beliefs and desires through a tacit folk- psychological theory. Observed behaviour and context are used to infer hidden states that predict further action.

This explains systematic attribution and development through learning. It also permits error: the folk theory can misclassify minds or omit important states. Scientific psychology can refine the model.

Critics argue that ordinary social understanding is too rapid, embodied, and context- sensitive to consist mainly of theoretical inference. The theory may describe reflective interpretation better than face-to-face engagement.

Simulation theory

Simulation theory proposes that one understands others by using one's own cognitive system as a model: imaginatively adopt another's situation, generate a candidate response, and attribute the resulting state.

Simulation explains empathy and sensitivity to perspective. It can outperform abstract rules when agents share capacities. Its risks are projection and egocentric bias. Other people differ in values, concepts, bodies, and experience; running one's own system may systematically mislead.

Theory and simulation can cooperate. Background knowledge selects inputs and corrects projection, while simulation supplies detailed expectations.

Direct social perception

Phenomenological and enactive views argue that emotion and intention are often perceived directly in expressive action. One sees anger in a face or hesitation in movement rather than first perceiving neutral geometry and inferring a hidden state.

"Direct" does not mean infallible or unprocessed. Social perception is learned, context-sensitive, and embodied. It means the other person's minded expression is the intentional object of perception, not a theoretical entity always reached through conscious inference.

The view fits ordinary immediacy but is strongest for expressive occurrent states. Long- term beliefs, concealed motives, and unfamiliar agents still require interpretation.

Testimony

Others tell us what they think and feel. Testimony is not merely behaviour among other behaviours; linguistic practices make speakers accountable to claims and permit challenge, correction, and elaboration.

First-person reports have defeasible authority, as Self-Knowledge and Introspection explains. Their weight depends on sincerity, conceptual competence, memory, incentives, and the reliability of the reporting mechanism.

Rejecting testimony whenever direct verification is impossible would destroy social knowledge generally. Accepting every first-person sentence would ignore confabulation, role-play, and generated text. The evidential question is mechanistic and contextual.

Animals

The full capacity-by-capacity evidence audit is in Animal Minds; the present section states the role of that evidence in the general other-minds problem.

Animal minds are supported by evolutionary continuity, homologous mechanisms, flexible learning, pain-related trade-offs, planning, communication, and metacognitive behaviour. Different capacities require different evidence; failure on a language-heavy task is not evidence of absent perception or sentience.

Anthropomorphism attributes human states from superficial resemblance. Anthropodenial rejects relevant continuity because the subject is non-human. Both replace discriminating evidence with a prior bias.

The analogical argument degrades by dimension, not along one species ladder. A creature may provide strong evidence of pain and weak evidence of propositional self-conception.

Artificial systems

The general assessment is developed in Machine Minds and the distinct question of Artificial Consciousness.

Artificial systems can separate evidence that human cases bundle: fluent report without shared biology, planning without bodily vulnerability, and self-description without an independently validated introspective channel.

Theories therefore assign different weights:

• functionalism emphasises organisation and counterfactual capacities;
• biological views emphasise evolved or neural mechanisms;
• higher-order and workspace theories seek specified architectures;
• phenomenal-first views question whether public function reaches experience;
• interpretivism asks whether mental attribution yields robust prediction.

No behaviour is theory-free proof, but unfamiliar substrate does not make behaviour irrelevant. Architecture, learning history, world interaction, persistent goals, metacognitive control, and intervention can triangulate. The LLM Introspection series examines this problem under unusually rich testimony and unusually weak structural similarity.

Scepticism and symmetry

If another human's perfect behaviour is insufficient because experience is private, the same scepticism applies to everyone but oneself. Selectively applying it to machines or animals requires an independent reason - biology, mechanism, history, or theory - not privacy alone.

The symmetry also runs the other way. Human certainty is supported by dense convergent evidence unavailable in many artificial cases. Equal logical possibility does not mean equal posterior probability.

The rational attitude can be graded. One may be highly confident in human consciousness, moderately confident across many animals, and uncertain about unfamiliar systems while still taking low-probability moral risk seriously.

Theory-ladenness

Evidence for other minds depends on what a theory says consciousness, belief, or agency requires. This does not make attribution arbitrary. Theories themselves are constrained by human, animal, clinical, and mechanistic evidence.

The relation is reciprocal:

graph LR
  T[Theory of mind] --> W[Weight assigned to evidence]
  E[Behaviour, reports, mechanisms] --> T
  W --> J[Attribution judgment]
  J -. new tests .-> E

Progress comes from cases where theories diverge and interventions can test their proposed requirements, not from pretending one operational measure is neutral.

Assessment

Knowledge of other minds is neither direct acquaintance nor an unsupported leap. It is a mature form of abductive, criterial, testimonial, and interactive knowledge whose strength varies with convergence. The philosophical problem becomes practically urgent where those sources disagree.

Selected references

• Gallagher, Shaun. "Direct Perception in the Intersubjective Context" (2008).
• Goldman, Alvin I. Simulating Minds (2006).
• Hyslop, Alec. Other Minds (1995).
• Mill, J. S. An Examination of Sir William Hamilton's Philosophy (1865).
• Wittgenstein, Ludwig. Philosophical Investigations (1953).