---
id: "claim-ai-errors-ripple-differently"
type: "claim"
source_timestamps: ["§ Where Trust Breaks Down"]
source_url: "https://hbr.org/2026/02/how-to-foster-psychological-safety-when-ai-erodes-trust-on-your-team"
source_title: "How to Foster Psychological Safety When AI Erodes Trust on Your Team"
tags: ["sense-making", "error-recovery", "team-learning"]
related: ["concept-attribution-uncertainty", "prereq-collective-sense-making"]
speakers: ["Jayshree Seth", "Amy C. Edmondson"]
confidence: "high"
testable: true
sources: ["adoption"]
sourceVaultSlug: "hbr-seg-adoption"
originDay: 9
articleStem: "hbr-cl-79-psychological-safety-ai-trust"
sourceUrl: "https://hbr.org/2026/02/how-to-foster-psychological-safety-when-ai-erodes-trust-on-your-team"
sourceTitle: "How to Foster Psychological Safety When AI Erodes Trust on Your Team"
---
# AI Errors Disrupt Collective Sense-Making

**Claim (confidence: high · testable).** AI inaccuracies ripple through teams in *fundamentally different ways* than human mistakes.

The argument: when a human errs, teams engage in [[prereq-collective-sense-making]] — asking about context, reasoning, and data — which updates shared mental models and *strengthens* team bonds. Generative-AI errors **short-circuit this process**. Because of AI's black-box nature (see [[concept-attribution-uncertainty]]), teams cannot interrogate its assumptions or methodology. That absence of transparency prevents the normal checks, discussions, and mutual-accountability exploration teams rely on to recover from and prevent future errors. The vivid version is [[quote-black-box-sense-making]].

**Why it matters:** if AI errors cannot be metabolized socially, each one leaves residue — feeding [[concept-trust-ambiguity]] and [[concept-workslop-d79]] rather than being resolved.

**How to test it:** compare error-recovery behavior and mental-model updating after matched human vs. AI errors on the same task.

**Enrichment (indirectly supported):** Direct comparative studies (human-error vs. AI-error sense-making) are limited, but surrounding evidence supports the direction — Nature documents decreased psychological safety and increased stress from AI adoption (which would hinder open sense-making), the Partnership on AI notes opaque AI damages safety, and XAI literature confirms that opacity impairs users' ability to understand and contest outputs.