Self-Knowledge and Introspection

People normally know their own current beliefs, intentions, sensations, and experiences without observing behaviour or inferring from evidence. An avowal such as "I intend to leave" has a first-person authority unlike another person's attribution. Yet subjects confabulate motives, misclassify emotions, overlook perceptual limitations, and remain ignorant of most processing. Self-knowledge must therefore explain privileged access without assuming transparent or infallible minds.

Different mental states may be known differently. Sensations can be attended to; beliefs may be known by considering the world; intentions may be known through practical commitment; traits and unconscious motives may require interpretation. A single inner sense need not cover them all.

Authority without infallibility

First-person avowals are generally:

• immediate: not based on observed behaviour;
• authoritative: normally given greater weight than third-person guesses;
• asymmetric: others do not know the state in the same way;
• expressive: saying "I am in pain" can manifest the state as well as report it.

They are not incorrigible. A person can mistake anxiety for anger, deny a belief revealed across contexts, or sincerely confabulate why they chose. Authority is a default epistemic status grounded in a capacity, not immunity from correction.

The target is selective reliability. A theory should predict where first-person access is strong, where it is weak, and why sincere error does not erase the asymmetry entirely.

Inner-sense theories

An inner sense monitors first-order mental states through a higher-order mechanism, analogous to perception. A current pain or thought causes a representation of itself, providing self-knowledge.

The model explains error and absence. Monitoring can misrepresent, fail, or have limited resolution. It also connects self-knowledge with higher-order theories of consciousness.

The analogy risks regress if knowledge of the higher-order state requires another monitor. It can stop when the monitoring state constitutes the self-ascription rather than needing separate inspection. A further problem is immediacy: inner observation seems to position the subject as a spectator of their own mind rather than the thinker.

Transparency

To answer "Do I believe it will rain?" one often considers whether it will rain, not an inner object. This is the transparency method: determine one's attitude by attending outward to the reasons and subject matter.

The method fits belief's rational nature. A subject's answer commits them to a view of the world rather than reports a detected internal item. It also explains why self- knowledge can be authoritative without a private observation channel.

Transparency has limits. Evidence may be inconclusive while one already has a belief; people can discover attitudes from behaviour; sensations are not known by deciding whether the body is damaged. It is best viewed as one method for judgment-sensitive attitudes rather than a universal theory.

Rationalist accounts

Rationalist theories connect self-knowledge to agency. A rational subject can know what they believe because beliefs are commitments responsive to reasons; they can know what they intend because forming an intention settles a practical question.

Self-knowledge is partly constitutive of rational agency. The capacity to avow and revise attitudes is not an external detector but participation in the norms governing them.

This explains authority for active judgments and intentions. It is less suited to passive, recalcitrant, unconscious, or fragmented states. A subject may judge that they ought not fear while still fearing, or discover a standing belief inconsistent with their avowal.

Expressivism

On an expressivist view, first-person avowals express mental states as a groan expresses pain or an assertion expresses belief. They are not conclusions from evidence, which explains their immediacy and resistance to ordinary requests for justification.

Mature avowals also make truth-evaluable claims. "I am afraid" can be mistaken and can inform another person. Expressivism must combine manifestation with assertion rather than replace one by the other.

The view is strongest for current states enacted in speech and weaker for retrospective self-description or theoretical claims about one's personality.

Interpretive accounts

People sometimes know themselves as others do: by observing patterns, recalling action, and constructing an explanation. Unconscious motives, traits, biases, and long-term preferences often require this interpretive route.

Interpretation is holistic and can be socially scaffolded. Others may identify a pattern the subject misses. Therapy, conversation, and measurement can improve self-knowledge rather than merely override a private authority.

An exclusively interpretive theory loses the difference between current avowal and third-person attribution. A plural theory allows direct, transparent, expressive, and interpretive routes for different targets.

Confabulation

In classic confabulation cases, subjects offer fluent reasons for choices influenced by factors they do not report. Split-brain patients may explain actions initiated by information unavailable to the speaking hemisphere. Choice-blindness subjects defend an option they did not choose.

These findings show:

• sincerity does not guarantee causal accuracy;
• the mind generates coherent self-explanations from partial access;
• knowledge of a decision's content can coexist with ignorance of its causal basis;
• first-person reports are outputs of mechanisms open to empirical study.

They do not show that subjects are generally wrong about whether they feel pain or what they currently judge. Confabulation is target-specific: causal explanation of a state is harder than avowing its occurrence or content.

Change blindness and perceptual introspection

Change blindness and inattentional blindness reveal limits on visual access. Subjects may experience a scene as richly present while failing to detect large changes. This does not straightforwardly prove that the experience itself is sparse; it may show limited comparison, memory, or attention.

Introspective reports underdetermine whether phenomenal detail exists beyond access. The dispute about overflow in Kinds and Concepts of Consciousness illustrates the problem.

The correct lesson is methodological: performance failure cannot be mapped directly onto absence of experience, while confidence in richness cannot settle its structure.

Theory-ladenness

Concepts shape self-description. A person may classify bodily arousal as fear, anger, spiritual presence, or illness depending on training and context. Categories can reveal patterns and also create expectations.

This does not make all introspection fictional. Perceptual observation is also theory- laden without being arbitrary. Reports can be constrained by recurring experience, behaviour, physiology, and correction.

Cross-cultural variation is evidence about how phenomenal, conceptual, and social components interact. It cautions against treating one vocabulary as a transparent inventory of universal mental kinds.

First-person concepts and immunity to error

Some first-person judgments are immune to error through misidentification. If one judges "I am in pain" on the basis of feeling pain, one may be wrong about the diagnosis but not ordinarily wrong about whose pain is presented. No identification of a person as oneself occurs.

This is weaker than infallibility. The subject can misclassify the sensation or use the concept incorrectly. Immunity concerns the route by which the subject is selected, not the truth of every predicate.

The phenomenon supports a minimal first-person structure without proving a Cartesian ego. An embodied self-model can organise states around one perspective without an inner act of identification.

Externalism and self-knowledge

If content depends on environment and community, a subject can know what they think without knowing all factors that fix it. This issue is developed in Internalism and Externalism.

The distinction is between first-order authority and semantic theory. One can know "I think that water is wet" while not knowing water's chemistry or the causal history of the term. Externalism embeds self-knowledge in a world the subject need not fully understand.

Sceptical switching cases remain pressure: if the subject cannot know whether the environment fixes water or twin-water content, the exact external content may outrun reflective access.

Machine self-report

A machine can report internal variables, uncertainty, goals, or processing history. Whether this is introspection depends on the target and mechanism, not first-person grammar. A dedicated, validated monitoring channel is stronger evidence than a fluent description generated by the same process used for general text.

The existing LLM Introspection series examines systems whose reports may be trained reconstructions, functional self-models, or candidate avowals. Human confabulation supplies a caution, not an automatic argument that machine and human reports have equal standing.

A plural account

Self-knowledge is distinctive without being one faculty. Its authority comes from the subject's constitutive and reliable relations to states, while its limits reflect the different mechanisms and concepts through which those states become available.

Selected references

• Bar-On, Dorit. Speaking My Mind (2004).
• Moran, Richard. Authority and Estrangement (2001).
• Nisbett, Richard E. and Timothy D. Wilson. "Telling More Than We Can Know" (1977).
• Schwitzgebel, Eric. Perplexities of Consciousness (2011).
• Shoemaker, Sydney. "Self-Knowledge and Inner Sense" (1994).