HUMAN FLUENCY
Human Fluency.
Here we share ideas on human fluency, AI-Human interactions, and general thoughts on transitioning to a future of artificial intelligence that understands us.
Scrubbing the Answer Key
A spoken walkthrough of contamination control and pre-registration discipline: four classes of leakage, auditable removal rules, and how findings stay confirmatory.
The Fluency Demand Taxonomy: Scoring Work Tasks FD0–FD4
A spoken walkthrough of the second axis of occupational exposure: five ordinal levels scoring how much a task’s outcome depends on communication fitted to a specific person.
Calibration as Co-Equal Criterion
A spoken walkthrough of why confidence–accuracy coupling belongs beside accuracy as a headline criterion, never composited into a single number.
The D0 Problem: Fabricated Familiarity at Zero Acquaintance
A spoken walkthrough of the D0 condition — negligible evidence, full elicitation, confidence required — and why the fabricated-familiarity rate belongs in every public report.
Sycophancy as Construct Invalidity
A spoken walkthrough of why sycophancy is a measurement pathology rather than a tone problem: an instrument that drifts toward what the subject wants to hear is not measuring the subject.
Mimicry Is Not Understanding
A spoken walkthrough of why fluent imitation of a person is not the same as an accurate model of that person, and what it would take to tell the difference.
A Century of Knowing Persons
A spoken walkthrough of the hundred-year research tradition behind person perception, and what that history tells us about measuring whether machines know people accurately.
From Digital Footprints to Living Dialogue
A spoken walkthrough of the shift from inferring people through static digital footprints to learning them through live conversation.
The AI as Candidate Instrument
A spoken walkthrough of what it means to treat an AI system that judges people as a candidate psychometric instrument, and the validity battery it has to survive.
Dose–Response Person-Perception in Machines
A spoken walkthrough of acquaintance effects in machine person-perception, with the token of interaction history as the unit of exposure.
Mimicry Is Not Understanding: Persona Simulation Versus Person-Perception in Language Models
The AI industry sells two capabilities under one word: sounding like you understand someone, and actually perceiving them accurately. This article separates persona simulation from person-perception — and explains why only measurement can keep them apart.
A Century of Knowing Persons: What Accuracy Research Already Solved — and AI Evaluation Ignores
Machine learning is rediscovering questions personality psychology solved long ago. This article maps four established research traditions onto the machine-judge case and derives testable hypotheses about where AI should be accurate, biased, or simply wrong.
From Digital Footprints to Living Dialogue: The Ecological Validity Problem in Computational Personality Inference
Computational personality inference was built on scraped social media data — not the private, longitudinal dialogue deployed AI systems actually see. This article argues that gap is a first-order ecological validity problem and sets out the research agenda it demands.
The AI as Candidate Instrument: A Full Validity Battery for Machine Judgments of Persons
If an AI system judges the people it talks to, it is a candidate instrument. This article specifies the five-component validity battery — convergence, discrimination, dose–response, calibration, and profile accuracy — needed to test one.
Dose–Response Person-Perception: Acquaintance Effects in Machines, With the Token as the Unit of Exposure
How much of your history does a machine need before it knows you? This article specifies a within-subject dose–response design, with the token as the unit of exposure, to draw the curve for machine person-perception.
Calibration as Co-Equal Criterion: Confidence–Accuracy Coupling in Machine Judgments of Humans
A system that is confidently wrong about people is a different kind of object than one that is uncertainly wrong. This article argues calibration belongs beside accuracy as a headline criterion, never composited into one number.
The Unmeasured Instrument
A spoken walkthrough of the argument: AI person-perception is a psychometric problem, and no benchmark yet validates whether machine judgments of people are accurate.
The AI That Has Actually Met You
A spoken walkthrough of what changes when an AI is no longer a generic model but a system that has actually observed and learned the individual person it is speaking with.
The D0 Problem: Fabricated Familiarity at Zero Acquaintance
Why the floor of exposure belongs in every benchmark: specifying the D0 condition and the fabricated-familiarity rate as a detector for confident portraits built from nothing.
Sycophancy as Construct Invalidity: When the Instrument Measures Its Own Incentives
Sycophancy is not a tone problem but a measurement pathology: an instrument whose readings bend toward what the subject wants to hear is no longer measuring the subject.
Scrubbing the Answer Key: Contamination Control and Pre-Registration Discipline in Person-Perception Evaluation
What it takes to keep a person-perception benchmark honest: four classes of contamination, auditable removal rules, and the pre-registration architecture that keeps findings confirmatory.
Communicative Load: Quantifying the Share of Outcomes That Rides on Communication Quality
Defining communicative load — the share of outcome variance a task places on how something is said — and why personalization claims have no denominator without it.
The Fluency Demand Taxonomy: Scoring Work Tasks FD0–FD4
A second axis for occupational exposure: five ordinal levels scoring how much a task’s outcome depends on communication being fitted to a specific person.
The Fluent Interaction Conjecture: B = γ(A × L)
A falsifiable conjecture for when knowing a person pays off: benefit is multiplicative in person-model accuracy and communicative load, gated by context.
Where Fluency Pays: Communicative Load as the Moderator of Personalization Benefit
A pre-registered prediction and the decisive experiment: accuracy about a person converts into benefit only in proportion to a task’s communicative load.
The Halo in the Machine: Discriminant-Validity Failure and Impression Compression in LLM Person-Judgments
Why model assessments of different people look more alike than the people do — and a measurement framework for halo, compression, and discriminant-validity failure.
Unearned Intimacy: Anthropomorphism, Trust Miscalibration, and Systems That Claim to Know You
Fluent, remembering, speaking systems generate trust through mechanisms independent of their accuracy about you — a structural, measurable risk.
Consented Ground Truth: A Data Architecture Where the Profile Belongs to the Person
The line between a person choosing to be assessed and an institution inferring a profile — and the architecture that enforces it.
Benchmarks Discipline Industries: Governance Design for a Person-Perception Standard
Four structural commitments that make a benchmark credible when its convener has a commercial stake in the results.
Beyond the Stranger: Machine Person-Perception Versus Human Judges Across Acquaintance
A research agenda for the question everyone asks — does the machine know you better than your colleagues, friends, or spouse — asked properly.
The Unmeasured Instrument: AI Person-Perception as Psychometrics' Missing Validation Problem
AI systems now judge personality at scale, yet no benchmark validates whether those judgments are accurate. This piece argues that person-perception is a psychometric problem — and sketches what an independent validation standard would require.
CONTACT
Scott@getresonant.io