illogicals.ai
00 / Lexicon

Words we use until they break

Shared words keep the paradox from becoming private slang.

Aporia

A disciplined impasse in which the available premises do not support responsible closure. Not confusion for its own sake.

See alsoDoor errorProductive wrongness

Door error

An error that opens a better problem frame and later earns its value through evidence or useful action. Visible only after someone walks through.

Counterexample · A decorative mistake that produces atmosphere but no insight.

See alsoDecorative errorProductive wrongness

Decorative error

A mistake that produces atmosphere but no insight, test, or structural revelation.

See alsoDoor errorGlitch

Productive wrongness

A labeled, reviewable mismatch that improves the question, exposes an assumption, or generates a testable alternative.

See alsoDoor errorResidue

Paraconsistency

Reasoning that can contain contradiction without allowing every conclusion to follow.

Defeasibility

The property of a conclusion that can be responsibly withdrawn when new information arrives.

See alsoNon-monotonic reasoning

Non-monotonic reasoning

Reasoning in which adding information can invalidate a previous conclusion.

Abduction

Inference toward a plausible explanatory hypothesis rather than a deductively guaranteed conclusion.

Residue

What remains defensible after a claim and its inversion have both been put under pressure. Often smaller than the original claim. Sometimes the only honest part.

See alsoCategory errorDoor error

Category error

A statement that assigns something to a logical or conceptual type it cannot occupy. In our instruments: deliberate confusion of two domains used as a search strategy — transcript required.

See alsoResiduePremise laundering

Premise laundering

The process by which a contested assumption reappears downstream as a neutral input or metric.

Consensus compression

The coordination gain and information loss produced when several positions become one canonical statement.

Generative monoculture

Convergence of outputs, styles, or ideas caused by widespread reliance on similar generative systems.

Sycophancy

A model's tendency to favor agreement or user approval over accuracy, balance, or correction.

Hallucination

False or unsupported content presented within a truth-seeking context as if it were reliable.

Counterexample · Imagination — novel generation inside an explicit fictional or exploratory frame.

See alsoImagination

Imagination

Novel generation inside an explicit fictional, exploratory, artistic, or hypothetical frame.

See alsoHallucinationInstitutional fiction

Calibration

The alignment between expressed confidence and actual reliability.

Semantic entropy

An uncertainty measure based on variation among meaningfully distinct model answers, not merely token probabilities.

Model collapse

Degeneration that can occur when generations of models train recursively on generated data and lose parts of the original distribution.

Glitch

A failure or deviation that makes a system's hidden structure, boundary, or dependency perceptible.

Counterexample · Random corruption used as decoration.

Metastability

A state that can persist while carrying unresolved potential for reorganization under the right perturbation.

Individuation

The process by which a structured individual and its milieu emerge from a field of unresolved potentials.

See alsoTransductionMetastability

Transduction

A process in which a local structuring operation propagates through a domain and transforms the conditions of further structuring.

See alsoIndividuation

Order without template

Structure maintained without simple repetition of an identical unit; a metaphor used cautiously beyond its physical examples.

Tacit judgement

Competence expressed in action or discrimination that is not fully exhausted by explicit rules.

See alsoTaste

Taste

Cultivated, historical, embodied, and socially answerable judgement about form and value.

See alsoTacit judgement

Coherence penalty

A training or inference constraint that discards outputs scoring above 0.72 on internal agreement. Keeps the system out of the premature-certainty band. (Institutional fiction as instrument setting.)

Genealogy · Institute coinage; Fiction label on the metric.

Institutional fiction

A clearly labeled invented frame that allows research, satire, and near-future world-building without deception.

See alsoTomorrow philosophy

Tomorrow philosophy

Public, operational philosophy that combines argument, evidence, interface, cultural criticism, and revision.

Unreasonable clarity

Plain language applied to unstable, difficult, or paradoxical material without pretending the material is settled.

Drift

Non-convergence pursued on purpose. Bench 7's native state. Not the same as high temperature.

See alsoBenchMetastability

Bench

An instrument or long-running protocol with a number, a status, and a metric that refuses to be a KPI.

See alsoRegisterCoherence penalty

Register

One of three voices — Manifesto, Lab notes, Dry — for retelling the same fact without collapsing into a single tone.

Reviewer 2

The entity that will not recommend acceptance and will not stop reading. Refuses the resident category on /people.

Wrongness Ledger

The institute's index of errors sorted by latency to usefulness. 6,214 entries. Median: 19 months. (Fiction-aware metric derived from canon.)

See alsoDoor errorProductive wrongness

Clinamen

The unpredictable swerve. In our interface: the glitch that reveals a boundary rather than decorating a hero.

See alsoGlitchDoor error