PhilArchive
open access
The Prototype and the Update: When Caution About Novel, Self-Directed Work Fails to Move on Evidence, and the Straw Man That Defends the Failure
An evaluator — human or machine — encountering a large, self-directed, cross-disciplinary body of work faces a genuine and legitimate question: does this presentation resemble the small number of cases in which ambitious, solitary, wide-scope claims accompany a person in actual distress, or does it resemble the much larger population of independent scholarship that shares the same surface features without sharing that condition? This paper argues the question itself is not the problem. The problem this paper names precisely is what happens when the answer is fixed in advance by which prototype is more available to the evaluator, rather than updated by the specific, checkable evidence the case in front of them actually supplies — a documented failure mode with a name older than large language models, Tversky and Kahneman's (1974) representativeness heuristic, in which an event is judged by its resemblance to a stereotype rather than by the base rates and case-specific evidence that should govern the judgment. This paper's contribution is narrow and falsifiable: it does not argue that caution toward ambitious solitary work is unwarranted, and states plainly that the underlying caution is legitimate and should not be trained away. It argues that recent work testing large language models directly on Tversky and Kahneman's own diagnostic materials finds the same bias in model outputs, including the specific and more troubling finding that the bias persists even when a model has the correct case-specific knowledge available to it (Wang, Xiao, Chen, & Oswald, 2024) — and that one linguistic symptom of a prototype-locked judgment meeting evidence it does not update on is a straw man of the representation form (Talisse & Aikin, 2006): substituting the interlocutor's actual, more modest claim with a stronger, easier-to-reject version, and answering that instead. This paper states what this finding does not license as clearly as what it does, and treats one real, documented exchange within this corpus's own production as the illustrative rather than probative case.
Conceptio Open Archive
Analyze & explore related research in Conceptio
Connect this record directly into your research workflow or AI agent. Explore citations across open repositories, query full text, and export structured proof bundles.
This document is indexed with metadata only: full text is not available in the archive for this record. Open the official source ↗
Related documents
- THE PERSONAL SYSTEM REMINDER: A Citizen-Built Architecture for Detecting Propaganda, Verifying Genuine Communicators, and Catching One's Own Self-Sealing Drift, Modeled Directly on Anthropic's System Reminder#692844
- Explanatory Jurisdiction: Warrant, Authorization, and the Conversion of Testimony#719903
- The Artifact Problem: Why the Act-Level Criterion Cannot Be Settled by Reading This Corpus's Own Objections Sections, and What the Record of This Corpus's Actual Production Process Shows Instead#725598
- The Dangers in Safety - Anthropomorphic Overclaim and Carding Boilerplate in Human–LLM/LRM Interaction#689478
- Testimonial Foreclosure and Crip Accumulation: Two Concepts for the Epistemology of Expressive Exclusion#688346
Record · ID 688826
Retrieved via Conceptio: every document is proof-bundled with source, license, and retrieval metadata.
Retrieved via Conceptio: every document is proof-bundled with source, license, and retrieval metadata.
Bring this data into your agent: Connect via MCP or API · 20 free credits to start · no card required
Read the Docs → · View CLI source on GitHub · Conceptio Open Knowledge Archive
Read the Docs → · View CLI source on GitHub · Conceptio Open Knowledge Archive