Understanding large language models demands distinguishing human projection from machine cognition

Reviews, theory, and governance2026Nature / Scientific ReportsApproved editorial review

Authors: Lingyu Li, Yan Teng, Yingchun Wang, Xia Hu

Keywords: Theory, Methodology, Governance

Source: Open primary source (opens in a new tab)

4
Authors
4
Findings
5
Limitations
3
Evidence

Editorial summary

English

Conceptual Perspective organizing literature from physics, neuroscience-inspired interpretability, cognitive psychology, and sociology across mechanistic, behavioral, and interactive scales. For each metaphor it distinguishes technical focus, minimal assumption, and unverified extension, then proposes machine experientialism as an interpretive frame.

There is no experimental sample or systematic review. The article conceptually synthesizes 80 references and presents one comparative table of four metaphor families. Metaphors illuminate different scales and are not complete explanations. Behavioral psychology enables controlled probes but does not license inference of a stable human mind. Sociology describes interactive roles but not intrinsic interests or commitments. Machine experientialism shifts the question toward representations constructed from text.

It is a conceptual proposal, not a tested theory. Literature selection does not follow a systematic protocol. Experientialism may reintroduce the cognitive vocabulary it seeks to delimit. It provides no new falsifiable predictions or metrics. Boundaries among scales are heuristic. It does not demonstrate LLM cognition, experience, or subjectivity. It does not empirically resolve the understanding-versus-pattern debate. It does not validate LLMs as substitutes for human participants.

Español

Perspective conceptual que organiza literatura de física, interpretabilidad inspirada en neurociencia, psicología cognitiva y sociología en escalas mecanística, conductual e interactiva. Para cada metáfora distingue foco técnico, supuesto mínimo y extensión no verificada, y propone el experientialismo de máquina como marco interpretativo.

No hay muestra experimental ni revisión sistemática. El artículo sintetiza conceptualmente 80 referencias y presenta una tabla comparativa de cuatro familias de metáforas. Las metáforas iluminan escalas distintas y no son explicaciones completas. La psicología conductual permite probes controlados pero no autoriza inferir mente humana estable. La sociología describe roles interactivos pero no intereses o compromisos intrínsecos. Machine experientialism desplaza la pregunta hacia representaciones construidas desde texto.

Es una propuesta conceptual, no una teoría probada. La selección bibliográfica no sigue protocolo sistemático. Experientialism puede reintroducir vocabulario cognitivo que pretende delimitar. No ofrece predicciones falsables ni métricas nuevas. Las fronteras entre escalas son heurísticas. No demuestra cognición, experiencia o subjetividad de un LLM. No resuelve empíricamente el debate comprensión-patrón. No valida LLM como sustitutos de participantes humanos.

Research question

What do disciplinary metaphors explain and obscure about LLMs, and can machine experientialism avoid anthropomorphic projection?

Method

Conceptual Perspective organizing literature from physics, neuroscience-inspired interpretability, cognitive psychology, and sociology across mechanistic, behavioral, and interactive scales. For each metaphor it distinguishes technical focus, minimal assumption, and unverified extension, then proposes machine experientialism as an interpretive frame.

Sample: There is no experimental sample or systematic review. The article conceptually synthesizes 80 references and presents one comparative table of four metaphor families.

Findings

  • Metaphors illuminate different scales and are not complete explanations.
  • Behavioral psychology enables controlled probes but does not license inference of a stable human mind.
  • Sociology describes interactive roles but not intrinsic interests or commitments.
  • Machine experientialism shifts the question toward representations constructed from text.

Limitations

  • It is a conceptual proposal, not a tested theory.
  • Literature selection does not follow a systematic protocol.
  • Experientialism may reintroduce the cognitive vocabulary it seeks to delimit.
  • It provides no new falsifiable predictions or metrics.
  • Boundaries among scales are heuristic.

What the study does not establish

  • It does not demonstrate LLM cognition, experience, or subjectivity.
  • It does not empirically resolve the understanding-versus-pattern debate.
  • It does not validate LLMs as substitutes for human participants.

Traceability

Scope: Full text

Version: journal; 8-page full text reviewed 2026-07-18

Consulted source: https://doi.org/10.1038/s44271-026-00508-6

Review: Codex full-text and visual 8-page methodological, statistical and claim-boundary review, 2026-07-18

Approval: Codex fidelity pass, 2026-07-18

English translation: approved, 2026-07-18

Models evaluated

  • LLMs discussed conceptually

Instruments and metrics

  • Three-scale metaphor map
  • Minimal-assumption boundary table

Data used

  • 80 cited works

Evidence and location

  • Research question, method, results, and discussion: Full text, pp. 1-8, visually reviewed on 18/07/2026
  • Figures, tables, results, and limitations: Primary PDF sha256 4598cd9d7d10b60bef4bee1d32099b401e04b78aa358312d7b1cbef5123f870b; methods, results, limitations, and appendices
  • Editorial decision and claim boundary: Critical record article-414, complete cross-check of 8 pages