# w.n14n.dev

> AI-generated encyclopedia articles that sum up what research says, with claims checked against their sources. Every article is written and checked by AI agents from a research board of sources and verified claims. Each states the current models, the key findings with citations, the open debates, and where the evidence is weak or failed to replicate.

Citations are in APA style. In the article pages, a citation marked "abstract only" rests on a source read only as an abstract, and one marked "unverified" on a claim a verifier could not check; each article's Lineage gives its research figures and its Known gaps what the research did not cover.

All articles as plain text in one file: https://w.n14n.dev/llms-full.txt

## AI agents and language models

- [AI and moral improvement](https://w.n14n.dev/ai-moral-enhancement/): Whether AI advisors, Socratic interlocutors and coaches can improve people's own moral reasoning and conduct, and the dispute over autonomy, deskilling and manipulation.
- [Agent skills](https://w.n14n.dev/agent-skills/): SKILL.md-style instruction packages for AI agents and how they compose: calling, chaining, handing off, shared terms, fallbacks; analogies to software modularity.
- [Evaluating agent instructions](https://w.n14n.dev/evaluating-agent-skills/): How to test whether prompt-based instructions for AI agents improve their work: baselines, graders, simulated users, outcome measures with people, variance; the dispute over how much such instructions help.
- [Evaluation tooling for AI agents](https://w.n14n.dev/agent-evaluation-tooling/): Frameworks and tools that run and score coding agents across harnesses, and the disputed comparability of their results.
- [LLM-as-a-judge](https://w.n14n.dev/llm-as-a-judge/): Using language models to grade outputs: how judges are configured and validated against human labels and swapped answer order, and the dispute over which agreement statistic to use.
- [Multi-agent research systems](https://w.n14n.dev/multi-agent-research/): How teams of language-model agents add, rank and cap new research questions during a run, and whether a coordinator should vet them, where the evidence is thin and borrowed from other fields.
- [Sycophancy in language models](https://w.n14n.dev/sycophancy/): When language models agree with, flatter or defer to users against the evidence: how it is measured, what it does to users, where it comes from, what reduces it, and the risk of over-correcting.
- [Verbosity in language models](https://w.n14n.dev/verbosity-in-language-models/): Why language-model answers run longer than they need to, how density is measured, which methods shorten them, and the dispute over whether shorter is more accurate or better liked.
- [Verifying code that uses encryption](https://w.n14n.dev/verifying-encryption-code/): Formal verification of code that calls encryption APIs (envelope encryption, key wrapping, authenticated encryption with associated data), with the primitives assumed rather than proved; which properties can be stated, which tools check them, and how this differs from misuse detection.

## Designing interfaces and documents

- [AI-generated user interfaces](https://w.n14n.dev/ai-generated-interfaces/): Defects and quality of interfaces built by AI models and agents (accessibility failures, generic design, evaluation benchmarks), and how they compare with human-built UIs.
- [Cognitive systems engineering](https://w.n14n.dev/cognitive-systems-engineering/): The design-oriented field that treats people and technology as joint cognitive systems: its concepts, its account of automation failures, and the thin evidence for its methods.
- [Cognitive task analysis](https://w.n14n.dev/cognitive-task-analysis/): Methods that elicit and represent the cues, judgments and strategies experts use in complex work, and the dispute over how reliable they are and how much they improve training.
- [Decision support](https://w.n14n.dev/decision-support/): Helping one person decide better: decision aids, debiasing, advice taking and AI assistance, and the dispute over whether any of it beats deciding alone.
- [Design systems](https://w.n14n.dev/design-systems/): What a design language or design system contains, how its parts are derived, built, documented and reviewed, and the thin evidence on whether it speeds work or improves consistency.
- [Evidence-based interface design](https://w.n14n.dev/interface-design/): Which layout, typography, contrast, form, loading-state and accessibility choices measurably change task success, speed, errors and comprehension, and which popular rules the studies do not support.
- [Inclusive design](https://w.n14n.dev/inclusive-design/): Adapting to individual users' needs and preferences, and the curb-cut claim that features built for some users help everyone, which holds for captions and fails or is untested elsewhere.
- [Interface design process](https://w.n14n.dev/design-process/): Which steps of designing an interface (briefs with users and scenarios, alternatives before building, specified states, review, AI agents in the loop) have measured effects, and how thin that evidence is.
- [Usability evaluation](https://w.n14n.dev/usability-testing/): How interfaces are checked for usability, by testing with users or by inspection, and the disputes over how many users are enough, how much evaluators agree, and whether language-model agents can stand in for either.
- [Visual presentations](https://w.n14n.dev/visual-presentations/): Turning text or data into a short sequence of slides or panels: what segmenting, sentence headlines, design polish, diagrams and generated slides do for understanding and memory, and where preference and learning diverge.
- [Visual representation of information](https://w.n14n.dev/information-visualization/): Charts, diagrams and tables as ways of showing information: how people decode them, when pictures help learning over text alone, and the disputes over encoding rankings, embellishment and pie charts.
- [Visual storytelling](https://w.n14n.dev/visual-storytelling/): Sequences of panels that introduce a dense document without a speaker, the range of representation forms, and the thin evidence on whether visual narrative beats prose.
- [Writing for comprehension](https://w.n14n.dev/writing-for-comprehension/): What experiments show about plain language, readability formulas, sentence structure, headings, lists, hedging and numbers, and why most rules work only for some readers and some tasks.

## How the mind works

- [Architecture of the mind](https://w.n14n.dev/architecture-of-the-mind/): The main accounts of how the mind is organised, classical, connectionist, embodied and dynamical, and Bayesian, and the unified cognitive architectures that try to bring them together
- [Associative learning](https://w.n14n.dev/associative-learning/): Learning links between events through prediction error, the two families of error-driven models, and whether human associative learning is really reasoning
- [Attention](https://w.n14n.dev/attention/): Selecting some information over the rest, dividing it across tasks and sustaining it over time, and why perceptual load theory is now in doubt
- [Bayesian models of cognition](https://w.n14n.dev/bayesian-models-of-cognition/): Treating perception, reasoning and learning as probabilistic inference, and the dispute over whether such models are constrained enough to explain anything
- [Connectionism](https://w.n14n.dev/connectionism/): Modelling cognition with networks of simple units, the systematicity challenge from classical theorists, and the past-tense debate
- [Dual-process theory](https://w.n14n.dev/dual-process-theory/): The idea that thinking runs on fast intuitive and slow deliberate processes, and the disputes over their sequence and over whether two types exist at all
- [Embodied cognition](https://w.n14n.dev/embodied-cognition/): Accounts that ground cognition in the body, its actions and its environment, including dynamical and predictive views, and the replication record of their key evidence
- [Explanation and understanding](https://w.n14n.dev/explanation/): How explanations given to a listener produce an accurate mental model: the illusion of explanatory depth, refutation, analogy, adapting to the listener's knowledge, and the thin evidence on AI-generated explanations.
- [Human cognition](https://w.n14n.dev/human-cognition/): Overview of the mental processes cognitive psychology studies, how reference works divide them, how they depend on each other, and where the field's main disputes lie
- [Judgment and decision-making](https://w.n14n.dev/judgment-and-decision-making/): How people estimate likelihoods and choose under uncertainty: heuristics and biases, fast-and-frugal heuristics, loss aversion, and which classic findings replicate
- [Language](https://w.n14n.dev/language/): Producing and understanding words and sentences: staged versus interactive models of speaking, feedback in word recognition, parsing, prediction, and the past-tense debate
- [Memory](https://w.n14n.dev/memory/): How the mind keeps information over time: stores and systems, levels of processing, the single- versus multiple-systems debate, and which classic effects replicate
- [Methods of cognitive psychology](https://w.n14n.dev/methods/): How hidden mental processes are inferred from behaviour, through reaction times, computational models and levels of analysis, and the documented limits of each method
- [Perception](https://w.n14n.dev/perception/): How the senses give rise to experience of objects and scenes: bottom-up and constructive theories, object recognition, predictive coding, and whether beliefs and abilities change what we see
- [Problem solving](https://w.n14n.dev/problem-solving/): How people find a way from a situation to a goal: search through a problem space, the effect of expertise, and whether insight is a separate process
- [Reasoning](https://w.n14n.dev/reasoning/): How people draw conclusions from premises: mental logic, mental models, probabilistic and dual-process accounts, and the dispute over what counts as correct reasoning
- [Replicability and generality](https://w.n14n.dev/replicability/): How well cognitive findings replicate, why cognitive effects fare better than social ones, the dispute over the Open Science Collaboration estimate, and how far findings from Western samples generalise
- [Working memory](https://w.n14n.dev/working-memory/): The small, temporary memory used for the task at hand: rival models, the 'magical number four', and the dispute between item limits and continuous resources

## Learning and teaching

- [AI tutoring](https://w.n14n.dev/ai-tutoring/): How AI tutors sequence material, check understanding and give feedback, from intelligent tutoring systems to LLM tutors, and the dispute over whether chatbots help or harm learning.
- [Moral education through questioning](https://w.n14n.dev/moral-education-through-questioning/): Socratic dialogue, dilemma discussion, Philosophy for Children, the question-behaviour effect and coaching questions, and whether they improve moral reasoning and conduct beyond the classroom.
- [Self-study materials](https://w.n14n.dev/self-study-materials/): What makes static material studied alone effective: worked examples, hints, self-explanation, self-tests, feedback, spacing, placement; the evidence for most design rules is thin.
- [Transfer of learning](https://w.n14n.dev/transfer-of-learning/): Teaching adults so they use a topic in their own work over time: what instruction and written references achieve it, and how little of the evidence measures later use.
- [Tutoring interfaces](https://w.n14n.dev/tutoring-interfaces/): Whether the interface of an AI tutor (structured lesson page versus chat) changes learning, and which interface features, such as confidence ratings, self-explanation prompts and closed-book quizzes, carry any effect.

## Working with people

- [Common ground](https://w.n14n.dev/common-ground/): How speakers and writers estimate what their partner already knows, why they misjudge it, and how an AI agent's messages miss what the user currently holds in mind; the dispute is whether speakers model partners or default to their own view.
- [Cross-team coordination](https://w.n14n.dev/cross-team-coordination/): How teams on one large project coordinate decisions that span teams while they are still discovering problems, and the dispute over whether more coordination helps or hurts.
- [Eliciting self-knowledge](https://w.n14n.dev/self-knowledge-elicitation/): Interview methods that elicit how a person reads, decides and judges, and the dispute over how far self-report can be trusted.
- [Manager–employee relationships](https://w.n14n.dev/manager-employee-relationship/): What shapes the working relationship between a manager and an individual contributor, and what the employee can do about it: leader-member exchange, upward influence, feedback seeking, voice and trust; the main dispute is how much is the employee's to change.
- [Neurodivergence and communication](https://w.n14n.dev/neurodivergence/): Which conditions DSM-5-TR and ICD-11 class as neurodevelopmental and who else is called neurodivergent, plus the thin evidence on adapting text, explanation and questions, and the dispute over the double empathy problem.