Unverified50% confidenceFactExact time
In multi-agent debate, when retrieved web sources contain both potentially harmful and safe factual content, social biases such as conformity pressure and authority obedience may override evidence-based safety boundaries.
1
Sources
50%
Confidence
Long-term
Relevance
11/4/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe safety risks of multi-agent debate stem not only from the capability deficiencies of individual models, but also from the social structure of the debate process itself.78% similarUnverifiedThe study recommends that safety constraints should be context-sensitive, distinguishing between malicious personal attacks and legitimate character challenges in a debating context.73% similarUnverifiedOnce a powerful information role is poisoned or encounters the Vortox, it transforms from a trustworthy anchor into a source of misdirection, which is a core risk in social deduction games72% similarUnverifiedTikTok's search blocking reflects the dilemma social platforms face in balancing information accessibility and potential risks when handling sensitive health topics.71% similarUnverifiedEveryday harms—such as subtle social bias, addictive design, and systemic effects on vulnerable groups—are hard to quantify and hard to turn into PR material, and are therefore excluded from safety assessments.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1133957API
curl https://kongchang.com/api/v1/knowledge/claims/1133957MCP
get_claim(id=1133957)