Shared Information Bias¶
A group-discussion sampling bias that favors already shared facts over uniquely held task-relevant information.
Core Idea¶
Shared information bias occurs when a group discussing partly distributed evidence gives disproportionate attention to facts multiple members already knew, while uniquely held relevant facts are less likely to surface. A common fact has more potential speakers who may recall and mention it; preference-consistent advocacy can add a separate influence. The named effect concerns what enters discussion, not a guarantee that the group makes a bad choice.[^ref-0faec8d42e06]
A hidden profile is a useful test: the unique facts favor an option obscured by initially common information. But the hidden-profile arrangement is not the bias itself. Visible information sheets and public expert-role assignments have changed the discussion pattern in particular experiments, so the effect is conditional rather than inevitable.[ref-aebffb6da0e6][ref-b824f223935a]
Scope of Application¶
Stasser and Titus studied political-caucus simulation groups whose candidate attributes were shared or distributed among members. In their design, 83% of 18 groups in the shared condition chose the collectively best-supported candidate, versus 18% of 38 in the unshared condition. Those figures describe a specific experiment, not real political committees.[^ref-0faec8d42e06]
Bowman and Wittenbaum studied three-person groups comparing hypothetical cholesterol-reducing drugs. Memory-only groups showed a shared-information discussion advantage; groups with accessible sheets reversed it. Their experiment also found a conditional time-pressure effect on choices. A meta-analysis found an overall shared-information advantage but also task and duration moderators, so “shorter time always means more bias” is not warranted.[ref-aebffb6da0e6][ref-ed93667b15c0]
Clarity¶
The contrast depends on who held an item before discussion and on whether it was then raised or considered. Unequal possession alone is Information Asymmetry; the directional selection of commonly held items is Shared Information Bias. Groupthink concerns concurrence pressure, not necessarily this information-sampling pattern. A poor outcome alone does not establish which process caused it.[^ref-0faec8d42e06]
“Best candidate” in the political-caucus study means the one supported by the group's pooled experimental evidence, not objectively the best person. The drug-choice study is simulated, not evidence about treating real patients. A shared fact can be useful to repeat; bias is assessed relative to available task-relevant evidence rather than by assuming every unique claim is true.[ref-0faec8d42e06][ref-aebffb6da0e6]
Manages Complexity¶
The abstraction separates three stages: initial distribution of facts, discussion sampling of those facts, and decision outcome. It thereby locates one route by which a group may fail to pool its knowledge without treating every group error as the same phenomenon. It also identifies why an intervention might matter: accessible sheets or visibly assigned expertise can make unique information easier to bring into the conversation.[ref-0faec8d42e06][ref-aebffb6da0e6][^ref-b824f223935a]
Abstract Reasoning¶
Ask what task-relevant items the members collectively possessed, which were shared before the meeting, and which reached discussion. If common items are systematically overrepresented while distinctive evidence is omitted, the named bias is present. Next ask whether the omitted evidence was decisive; that determines possible outcome cost, not the identity of the sampling effect.[^ref-0faec8d42e06]
Knowledge Transfer¶
The political-caucus and fictional-drug examples have different decision contents but the same structure: distributed premeeting knowledge, group discussion, and a testable common-versus-unique sampling direction. The proposed DAG parent is live prime Bias, the general directional-deviation relation; group communication and initial ownership of facts keep this entry domain-specific. Groupthink, Common Knowledge and Information Asymmetry are neighbors rather than asserted parents.[ref-0faec8d42e06][ref-aebffb6da0e6]
[^ref-0faec8d42e06]: Garold Stasser and William Titus, “Pooling of Unshared Information in Group Decision Making: Biased Information Sampling During Discussion”, Journal of Personality and Social Psychology 48 (1985), 1467–1478, especially pp. 1–4, 7–8 Table 3 and p. 11 of the PDF. [^ref-aebffb6da0e6]: Jonathan M. Bowman and Gwen M. Wittenbaum, “Time Pressure Affects Process and Performance in Hidden-Profile Groups”, Small Group Research 43 (2012), original publisher abstract only. [^ref-b824f223935a]: Garold Stasser, Sandra I. Vaughan and Dennis D. Stewart, “Pooling Unshared Information: The Benefits of Knowing How Access to Information Is Distributed among Group Members”, Organizational Behavior and Human Decision Processes 82 (2000), original publisher abstract only. [^ref-ed93667b15c0]: Torsten Reimer, Andrea Reimer and Uwe Czienskowski, “Decision-Making Groups Attenuate the Discussion Bias in Favor of Shared Information: A Meta-Analysis”, Communication Monographs 77 (2010), original publisher abstract only.
Relationships to Other Abstractions¶
Current abstraction Shared Information Bias Domain-specific
Parents (1) — more general patterns this builds on
-
Shared Information Bias is a kind of Bias Prime
Shared information bias is a directional distortion in group discussion's sampling of available evidence.
Hierarchy path (1) — routes to 1 parentless root
- Shared Information Bias → Bias
Neighborhood in Abstraction Space¶
Shared Information Bias sits in a sparse region of the domain-specific corpus (69th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Group Dynamics & Collective Behavior (19 abstractions)
Nearest neighbors
- Showdown Cooperative Learning — 0.86
- Focus group — 0.86
- Cultural Consensus Theory — 0.84
- Rhetorical Listening — 0.83
- Event Sampling Methodology — 0.83
Computed from structural-signature embeddings · 2026-10-08