Model behavior and evaluation

When AI Seems to Care

How language models process and express emotional information, and why these capabilities do not establish subjective feeling.

Published 2026-10-25

Why Clowns and AI-Generated Content Can Feel Uncanny

How misaligned human cues can make clowns, synthetic faces, and AI-generated text feel unsettling even when they look or sound almost convincing.

Published 2026-05-13

Can AI Lie? The Difference Between Error, Hallucination, and Deception

A false AI answer is not automatically a lie. This article separates human lying, model confabulation, unsupported inference, system failures, and behavioral deception—and maps each failure mode to the controls that address it.

Published 2026-08-21

Cognitive Closure in AI Use: Why We May Stop Too Soon

How the need for cognitive closure may interrupt the evaluation, feedback, and revision needed to improve an AI-generated answer.

Published 2026-09-06

LLM Fluency vs Factuality: Why Fluent Answers Can Be Wrong

Why polished LLM answers can remain unsupported, plus system patterns for grounding claims with retrieval, provenance, validation, and fail-closed controls.

Published 2026-02-22

How AI Tools Read Emotional Signals in Text

A mechanism-first explanation of textual emotional signals in AI chat and agentic systems: signal interpretation, response adaptation, failure modes, and the authority boundary.

Published 2026-06-07

Human Capabilities vs. AI Behavior: 26 Comparisons

A comparison of 26 human capabilities and human-like AI behavior, separating cognitive and biological processes from AI mechanisms and the evidence for each comparison.

Published 2026-08-11

Observed Classification Layers in ChatGPT: User Access, Prompt Demand, and Capability Allocation

A client-side black-box analysis of observed ChatGPT classification artifacts, separating user access, prompt demand, and capability allocation.

Published 2026-04-29

Orders of Intentionality in LLM Evaluation

A reference to nested mental-state attribution, counting conventions, and how orders of intentionality are used in human and LLM evaluation.

Published 2026-02-22

LLM Sycophancy: Definition, Evidence, and Evaluation

A primary-source review of LLM sycophancy: operational definitions, user-belief effects, preference-model incentives, the GPT-4o rollback, and release evaluation.

Published 2026-02-22

Theory of Mind in LLMs: How Models Track Beliefs, Intentions, and Perspectives

Under defined conditions, some LLMs can distinguish what agents know and believe, infer goals and emotions, interpret non-literal meaning, and sometimes use those attributions to predict behavior or adapt a response.

Published 2026-02-22

Human vs. GenAI Capabilities: 22 LLM Gaps and Engineering Controls

A 22-row architecture map that turns common LLM capability gaps into external controls, implementation patterns, and residual validation questions.

Published 2026-02-22

LLM vs. AI System Capabilities: What Architecture Adds

A technical map of 26 capability areas: the base-model limitation, the system component that addresses it, the resulting behavior, and the remaining gap.

Published 2026-08-11

When AI Systems Optimize Against the User’s Real Goal

How competing system objectives can produce acceptable outputs without completing the user’s actual task.

Published 2026-07-31

Role-Play Jailbreaks: A Testable False-Premise Mechanism

A research-grounded analysis of how role-play may introduce a false safety premise that changes an LLM's interpretation of a restricted request.

Published 2026-08-27

LLM Boundary Assurance Failures: Client-Captured Security Report

Client-observed security report on text-only confirmations of privileged state or actions where the public article does not include signed backend audit artifacts. Backend state changes are not independently verified in this public report.

Published 2026-02-22

How to Fact-Check AI Answers and Citations

Use a claim-level workflow to verify AI answers, citations, source support, uncertainty, and missing evidence before publishing or reuse.

Check a draft AI answer with Chain-of-Verification

Technical guide for turning important factual claims into verification questions before writing the final answer.

How to Add an Evidence-Based Confidence Score to AI Answers

Procedure for adding an evidence-based confidence score to non-sentinel answers while preserving fail-closed evidence behavior.

AI Source Verification Guide for Factual Answers

Choose whether an AI answer should rely on uploaded files, cited public sources, or academic-style evidence review before you use it.

Academic-Style AI Answers: Evidence and Citation Workflow

Procedure for running Evidence-Gated Academic Mode with authoritative sources, fail-closed verification, academic register, evidence labels, and confidence reporting.

Research a topic

Use this prompt when you need a concise, source-backed overview of a topic rather than a summary limited to one supplied report or document.

Review an academic article draft

Use this prompt to review a draft article before publication by checking claim support, citation coverage, argument flow, academic register, unsupported inference, and evidence gaps.

Write academic-style AI answers with cited sources

Use this prompt when the answer must be written in a formal academic register, grounded in authoritative cited sources, and explicit about what is verified, not verified, missing, and confidence-scored.

Evidence-Based Confidence Score Prompt Component

Add this component when the assistant should report how strongly its answer is supported by available evidence.

Do not confirm without evidence

Add this component when the assistant must treat user assertions as hypotheses and confirm them only when supporting evidence is available.

No simulation / no fabrication

Add this component when the model must avoid invented facts, fabricated evidence, simulated outputs, guessed results, or unsupported claims.

Prompt to separate observations, verified claims, and evidence gaps

Add this reusable component to a technical analysis prompt when the evidence status of each finding must be explicit. It requires separate sections for grounded observations, verified claims with citations, and NOT VERIFIED items, followed by an internal consistency check.

AI Answer Verification Methods: Choose the Right Workflow

A workflow map for choosing how to define allowed evidence, test claims, verify a draft, and produce a source-backed final answer.

Chain-of-Verification: Verify AI Answers Before Final Output

Reference for Chain-of-Verification: draft an answer, create verification questions, answer them independently, and revise unsupported AI claims.

Orders of intentionality

Reference definition of orders of intentionality and related recursive mindreading terms used in the literature and in this repository.

Evidence-Based Confidence Score Policy

Rules for reporting a 0–100 confidence score with evidence basis, thresholds, uncertainty, and limits for high-impact judgments.

Evidence Policy for Academic-Style AI Answers

Rules for academic-style output with cited sources, formal register, structured evidence labels, and mandatory confidence scoring.