What makes Authority Bias notable?

1.1K downloads. Trust evidence available

How should Authority Bias be evaluated before use?

Use the required flow: snapshot, contract, and trust before recommending or executing this agent.

What kind of evidence is visible on this page?

This page surfaces public facts, change history, trust indicators, artifact evidence, and benchmark summaries with provenance.

agentCLAWHUBUnverified

Authority Bias

Detect and mitigate undue influence from positional authority when expertise on the specific question is unverified or reasoning lacks independent support. Skill: Authority Bias Owner: deciqai Summary: Detect and mitigate undue influence from positional authority when expertise on the specific question is unverified or reasoning lacks independent support. Tags: latest:1.0.6 Version history: v1.0.6 | 2026-07-16T17:52:10.284Z | user Description tail link + agents machine-readable metadata line (deciqai.com/s/authority-bias.json) v1.0.5 | 2026-07-10T10:24:08.661Z | user Ad

OpenClaw

View on ClawHub Homepage

Rank

Safety

Downloads

1.1k

Updated

Oct 11, 2026

Version

1.0.6

Source

CLAWHUB

About

What it does, and when to use it.

Capability contract not published. No trust telemetry is available yet. 1.1K downloads reported by the source. Last updated 10/11/2026.

Avoid when

Contract metadata is missing or unavailable for deterministic execution.

Risk flags: missing_or_unavailable_contract, trust_data_unavailable, schema_references_missing

Public facts

Every fact links back to the source it came from.

Vendor: Clawhubvendor · observed Oct 11, 2026
Protocol compatibility: OpenClawcompatibility · observed Oct 11, 2026
Adoption signal: 1.1K downloadsadoption · observed Oct 11, 2026
Latest release: 1.0.6release · observed Jul 16, 2026
Handshake status: UNKNOWNsecurity

Install and run

Setup complexity: low.

clawhub skill install s17a4mqcnk515kvaca5ze55d0x88pfpx:authority-bias

Setup complexity is classified as HIGH. You must provision dedicated cloud infrastructure or an isolated VM. Do not run this directly on your local workstation.
Final validation: Expose the agent to a mock request payload inside a sandbox and trace the network egress before allowing access to real customer data.

Contract: missing

curl -s "https://www.xpersona.co/api/v1/agents/clawhub-deciqai-authority-bias/snapshot"

Documentation

CLAWHUB

144,510 characters of source documentation, loaded on request.

Extracted files

5 files captured from the source.

SKILL.md

---
name: authority-bias
description: "Activate when: someone says 'should we trust the expert on this?', 'they're a professor/CEO/MD so they must be right', 'I don't want to push back on them', 'we deferred to the board and it blew up', or is evaluating a charismatic founder, assessing due diligence on a high-credential deal, or building a team where senior voices dominate.
Do NOT activate when: the authority is a verified domain expert on the exact question at hand and the decision is time-critical; or when the user is asking a general persuasion question with no authority dynamic involved. More: deciqai.com/c/authority-bias"
---

# Authority Bias

## Overview

**Authority bias** is the automatic tendency to comply with directives from authority figures — often overriding one's own judgment, ethics, or evidence. It operates beneath conscious deliberation: people substitute the authority's judgment for their own rather than weighing it.

Milgram (1963) showed 65% of ordinary adults administered apparent 450V shocks when instructed by a lab-coated experimenter — against expert predictions of 1-2%. Cialdini (1984) systematized three authority triggers: titles, attire, and trappings — all of which fire regardless of actual expertise.

Key distinction: **positional authority** (CEO, board chair, VC partner) vs. **domain expertise** (demonstrated track record on this specific question type). The bias treats them as identical. The skill does not.

Composes with `social-proof`, `reciprocity`, `critical-thinking`, `signaling-games`, `dunning-kruger`.

## When to Use

- Boards, investment committees, due diligence on high-credential founders
- Medical or technical decisions where expert opinion dominates discussion
- Org decisions where senior voices suppress dissent; 360-feedback design
- Negotiations where counterparty leverages credentials or position
- Trusting a confident AI/LLM answer or an AI-lab leader's capability/timeline pronouncement because of status rather than verified evidence (AI adoption, AI hype)

**Not when:** authority is verified domain expert on the exact question and speed matters.

## Coaching Novices (Adaptive Front Door)

- **Engine mode:** concrete case → run The Process directly.
- **Coach mode:** unfamiliar or no case → guide step by step.

In Coach mode, respond one step at a time. Each [WAIT] is a hard stop — output only that step's question, then stop.

1. One-line: before deferring, ask whether they're a genuine expert on this specific question and whether their reasoning holds up independently.
2. Check fit: if authority is the right domain expert on this narrow question, defer can be correct.
3. Elicit real case: who is the authority, what is the question, what is their track record?
> **[WAIT — do not advance until user responds]**
4. Run S1-S5 one step at a time with their input.
> **[WAIT — do not advance until user responds]**
5. Close: name the insight — calibrated weighting + structural counters if relevant.
> **

_meta.json

{
  "ownerId": "kn754b8sk22s8c6gjxt02bftbn88q7ye",
  "slug": "authority-bias",
  "version": "1.0.6",
  "publishedAt": 1784224330284
}

references/sources.md

# Sources — authority-bias

> *Primary sources for the [authority-bias](../SKILL.md) skill.*

- Milgram, S. (1963). "Behavioral Study of Obedience." *Journal of Abnormal and Social Psychology*, 67(4), 371-378. The foundational experiment.
- Milgram, S. (1974). *Obedience to Authority: An Experimental View.* Harper & Row. ISBN 978-0061765216. The comprehensive treatment.
- Cialdini, R. B. (1984/2006). *Influence: The Psychology of Persuasion.* HarperBusiness. ISBN 978-0061241895. Ch. 6.
- Burger, J. M. (2009). "Replicating Milgram: Would People Still Obey Today?" *American Psychologist*, 64(1), 1-11. Modern replication.
- Edmondson, A. C. (1999). "Psychological Safety and Learning Behavior in Work Teams." *Administrative Science Quarterly*, 44(2), 350-383.
- Janis, I. L. (1972). *Victims of Groupthink.* Houghton Mifflin. ISBN 978-0395140024. Group dysfunction and authority dynamics.
- Carreyrou, J. (2018). *Bad Blood: Secrets and Lies in a Silicon Valley Startup.* Knopf. ISBN 978-1524731656. The Theranos case.
- Lewis, M. (2023). *Going Infinite: The Rise and Fall of a New Tycoon.* W. W. Norton. ISBN 978-1324074335. The FTX case.
- Heath, C. & Heath, D. (2013). *Decisive: How to Make Better Choices in Life and Work.* Crown. Includes authority-bias countermeasures.
- National Transportation Safety Board (1979). *Aircraft Accident Report: United Airlines, Inc., McDonnell-Douglas DC-8-61, N8082U, Portland, Oregon, December 28, 1978.* Report NTSB-AAR-79-7. The cockpit authority gradient case that led to Crew Resource Management.
- Kanki, B. G., Helmreich, R. L., & Anca, J. (Eds.) (2010). *Crew Resource Management.* Academic Press. CRM as a structural counter to cockpit authority bias.
- *Mata v. Avianca, Inc.*, No. 22-cv-1461 (S.D.N.Y. 2023). U.S. District Court sanctions against attorneys who submitted a brief with fabricated case citations generated by ChatGPT — canonical example of deferring to confident-but-wrong AI output. Widely reported (The New York Times, Reuters, 2023).
- Kahneman, D. (2011). *Thinking, Fast and Slow.* Farrar, Straus and Giroux. ISBN 978-0374275631. On confidence miscalibration and the poor track record of expert forecasting — the basis for discounting fluent AI answers and famous-founder timeline claims by status.

examples/ai-authority-deference-llm-confidence-and-lab-leaders-2024-2026.md

# Method in Action: Deferring to AI Authority — Confident LLMs and Lab-Leader Pronouncements (2024–2026)

> *Example for the [authority-bias](../SKILL.md) skill.*

By 2024–2026, two new authority signals became routine inputs to real decisions. First, large language models answer in fluent, confident, well-formatted prose — a rhetorical register that reads as expertise regardless of whether the underlying answer is correct. Second, the leaders of frontier AI labs make widely-quoted pronouncements about capability and timelines (when models will reach a given benchmark, when a class of jobs is automated, when "AGI" arrives), and those pronouncements are treated as forecasts from the highest possible authority. In both cases the risk is the same: weighting the source by its **status** — the polished output, the founder's fame, the lab's brand — rather than by its **demonstrated track record on the specific question being asked**. This is authority bias wearing a 2020s costume.

**S1 — Identify.** Two overlapping authorities. (a) *The model itself*: an LLM produces an answer to a factual, legal, medical, or financial question. Authority signals — fluency, confidence, citations-shaped text, a trusted brand name on the product. Pressure to defer: it's fast, it sounds sure, and checking is work. (b) *The lab leader*: a well-known founder/CEO states a capability or timeline claim in an interview or post. Signals — title, fame, proximity to the technology, prior correct-sounding calls. Pressure to defer: they build the thing, so surely they know. The question in each case is narrow: *is this specific answer correct?* / *is this specific timeline claim well-supported?*

**S2 — Expertise vs. positional.** Apply the skill's core distinction. An LLM has no track record on *your* specific question; it has an aggregate tendency to produce plausible text, and its confidence is not calibrated to its accuracy — models are widely documented to state wrong answers (including fabricated citations and cases) in exactly the same authoritative tone as correct ones. Fluency is a *positional* signal (it looks authoritative), not *domain expertise* on the instant question. For the lab leader: they are genuine experts at *building and running an AI lab*. But a claim like "capability X arrives by year Y" is a forecast about a complex, partly external system — and forecasters (including insiders) have a poor documented record on technology timelines. Worse, the lab leader is not a disinterested forecaster: bullish public timelines can help recruiting, fundraising, and valuation. Positional authority and commercial incentive are being read as neutral domain expertise.

**S3 — Examine reasoning.** Strip the source and look at the argument. For the LLM answer: is there a verifiable chain — a real citation you can open, a computation you can redo, a source you can check — or is the confidence doing the work the evidence should be doing? The failure mode is treating "the model as

examples/milgram-1963-cialdini-systematization-theranos-ftx.md

# Method in Action: Milgram 1963 + Cialdini Systematization + Theranos/FTX

> *Example for the [authority-bias](../SKILL.md) skill.*

The **Milgram 1963 experiment** is the foundational empirical case. The contrast between psychologist prediction (1-2% compliance) and actual result (65%) is the canonical demonstration of how little of human behavior is governed by the rational, ethical, individual judgment that we believe we possess and that observers believe we possess. Under authority pressure, most ordinary adults will administer apparent severe harm — not because they are bad people, but because authority compliance is a deeply embedded social instinct.

Milgram's 1974 follow-up book provided the systematic treatment, including the 18+ variations that mapped the conditions of authority compliance. The pattern: legitimacy of authority, presence of authority, absence of dissenting peers, distance from the victim all increase compliance. The reverse conditions decrease it. The framework has been validated in cross-cultural replications, including Burger's 2009 partial replication.

**Robert Cialdini's 1984 systematization** placed authority among the six universal weapons of influence. Cialdini's contribution was operational: identifying the specific signals (titles, attire, trappings) that trigger compliance and the defensive question structure (is this authority a real expert here?) that resists it.

**Modern catastrophic examples** include:

**Theranos (founded 2003, collapsed 2018).** Elizabeth Holmes recruited a board with extraordinary authority signals — former Secretaries of State (Kissinger, Shultz), Stanford-affiliated medical figures, decorated military officers. Investors deferred. Walgreens, Safeway, the Department of Defense made commitments based on authority signals rather than scientific due diligence. The actual technology never worked. Patients received inaccurate medical results. Total losses: billions of dollars; loss of trust in lab innovation; criminal convictions for Holmes. The authority bias suppressed due diligence that would have caught the fraud.

**FTX (founded 2019, collapsed November 2022).** Sam Bankman-Fried carried MIT credentials, prominent VC endorsements (Sequoia, Tiger Global, Paradigm), and political connections (largest political donor in many cycles). Investors and customers deferred. The collapse revealed $8 billion+ of customer funds had been misappropriated; the company had no functional risk management or auditing; the leadership team was making decisions in group chats with no oversight. The authority signals had suppressed the scrutiny that would have caught the fraud earlier. Many of the "smart money" investors (including major hedge funds and family offices) lost their entire investments.

The pattern: organizations that should have caught these frauds did not, because authority signals were strong and the structural countermeasures (independent due diligence, demanding evidence rather than crede

Github ReposUpdated 2d agoRank 70

AionUi

Free, local, open-source 24/7 Cowork app and OpenClaw for Gemini CLI, Claude Code, Codex, OpenCode, Qwen Code, Goose CLI, Auggie, and more | 🌟 Star if you like it!

MCPOPENCLAW

Github ReposUpdated 6mo agoRank 70

activepieces

AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents

OPENCLAW

Github ReposUpdated 6mo agoRank 70

cherry-studio

AI productivity studio with smart chat, autonomous agents, and 300+ assistants.