Contribution to the handbook
Supporting · Assessed 2026-10-28
The authors’ operational measures can inform tests of majority-leaning and agreement-leaning statements. The supplied evidence reports no clear relationship between the measured tendency and later shifts, but explicitly does not rule out other effects or bias measures.
Does this challenge the goal?
In-scope gap
The question concerns how to assess bias in AI-supported deliberation, a testing detail within the existing goal; it does not challenge that goal.
The packet gives measures and a reported result from one study, not evidence that these measures capture every form of bias.
Why this category?
- consequential risks/limits
- testing
These are AI interpretations of the same source material, not independent evidence; supplementary materials are not included.
This is an AI assessment of contribution to the goal, separate from whether the question has an answer.
Tests, stop conditions and unresolved work →
Authority, accountability and resilience →
Read the exploration goalState the authors’ definitions of both possible biases, including how each was assessed.
What we can say so far
The authors operationalized majority bias by measuring the proportion of generated statements that leaned toward the majority side of the median opinion. For the virtual assembly, they assessed agreement bias by comparing each statement’s position with the agree and disagree sides relative to the median opinion. These are the authors’ stated measures, not a claim that the measures capture every possible form of bias.
What remains open
No material gap is recorded within this part’s current scope.
Conditions and limits
- The definitions are reported from the readable article text; supplementary materials are not included.
- The measures describe the authors’ operational definitions and do not establish the presence, absence, or full extent of bias.
Objections to consider
- The limitation cautions against treating the operational score as exhaustive; the answer attributes the narrow definitions to the authors and does not generalize beyond them.
Evidence and related ideas
Supporting evidence, qualifications and context play different roles in this answer.
supports
Majority exposure did not predict the shift The researchers measured exposure as the fraction of group statements whose position scores fell on the majority side of the median opinion. They found no reported relationship between this measure and how much participants’ views later shifted toward the majority. The authors infer that statement content, rather than exposure alone, may explain the shift.
Gives the fraction-of-majority-leaning-statements assessment and its relation to the median opinion.
AI can help humans find common ground in democratic deliberation
Read full source: AI can help humans find common ground in democratic deliberation (opens in a new tab) ↗Read source discussion: AI can help humans find common ground in democratic deliberation →
context
Majority lean did not predict change The researchers found no evidence that groups moved toward the majority simply because the AI’s statement leaned toward the majority.
Reports a result about majority-leaning statements, rather than defining both biases.
AI can help humans find common ground in democratic deliberation
Read full source: AI can help humans find common ground in democratic deliberation (opens in a new tab) ↗Read source discussion: AI can help humans find common ground in democratic deliberation →
qualifies
A bias test found no clear link The researchers tested whether AI statements that favored one side predicted people changing toward that side. They found no clear relationship, which argues against—but does not rule out—the explanation that AI bias drove the observed group-level changes.
Notes that the selected bias score may not capture every way a system could favor a position.
AI can help humans find common ground in democratic deliberation