i

Description

Colonie de manchots Adélie serrés les uns contre les autres sur une plage de l'Antarctique

Source

Giuseppe Zibordi (NOAA Corps Collection) · original page

Licence

PD Public domain

Original photo, unretouched. The background is a version of it stylised by the site.

In-group bias

As soon as a line is drawn between "us" and "them," even at random, we start to prefer our own.

In brief

• What: we treat members of our own group better, believe them more and judge them more favorably.
• Why: belonging feeds identity and self-esteem; valuing our group means valuing ourselves.
• Detect: "they" understand nothing, "we" always do our best; trade-offs regularly favor the same team.
• Avoid: shared goals, criteria written down before comparing, mixed teams, regular exchanges between groups.

On this page

1. Definition
2. The mechanism
3. Where we encounter it
   1. In project management
   2. In management
4. Famous cases
   1. Tajfel's minimal groups 1970-1971
   2. Robbers Cave: the Eagles and the Rattlers summer 1954
5. Warning signs
6. Avoidance reflexes
7. With AI: detect, remedy, mitigate
   1. Detect
   2. Fix
   3. Mitigate
   4. Prompts to build into your routines
   5. How to build them in
   6. Precautions
8. Limits
9. Sources and verification
10. See also
    1. Combines with, or reinforces
    2. Related biases
    3. Counterweights and related reflexes

Definition#

Fictitious illustration

Fictitious illustration

In-group bias (also called in-group favoritism) is the tendency to evaluate more favorably, help more and trust more the members of our own group (the in-group) than those of other groups (the out-group). The group can be a profession (developers, operations staff), a team, a site, a vendor or "headquarters."

The founding experiment is that of Henri Tajfel and his colleagues in Bristol, in the early 1970s: it shows that merely being assigned to a group is enough to trigger favoritism, with no conflict of interest, no interaction and no shared history.

The mechanism#

• Categorization. The brain sorts people into categories; this accentuates similarities within a category and differences between categories.
• Social identity. In Tajfel and Turner's social identity theory (1979), part of our self-esteem comes from belonging; for it to be positive, my group has to be better than the others.
• Asymmetric attribution. The group's successes are attributed to its qualities, its failures to circumstances; for the other group, it is the reverse (see the fundamental attribution error).
• More love of "us" than hatred of "them." Research syntheses (Brewer, 1999) show that favoritism operates first through preference for one's own; hostility toward others appears mainly when resources are scarce or identity is threatened.

Where we encounter it#

In project management#

• Development and operations: each side sees the other as the source of the problems ("they deploy any old way" / "they code without thinking about production") and attributes its own incidents to circumstances.
• Teams and sites: a request from the Paris site is handled faster than one from the Lyon site; a defect is blamed on "the other site's team."
• Contractors and in-house staff: consultants are suspected of being slow, or conversely in-house staff of being "too close to the business"; competence is judged by the label.
• Design or code review: we are more lenient with our own team's code than with another team's.

In management#

• Hiring and promotion: we prefer candidates from the same school, the same former employer, the same network (see similarity bias).
• Resource allocation: budget, the best profiles or prestigious projects go more often to the team the manager knows, or comes from.
• Mergers and reorganizations: "us" (the old entity) and "them" (the new one) reproduce the divisions long after the announcement.

Famous cases#

Warning signs#

• "Them" and "us" keep coming up in conversations; the other team is referred to by a label rather than by people.
• The same facts are read differently depending on whether they concern one's own group or another.
• Trade-offs (budget, deadlines, assignments) regularly favor the same team.
• A newcomer is judged by their group of origin before being judged on what they do.
• Post-incident reviews look for "who" rather than "how" as soon as the culprit is from the other team.
• Hiring almost always picks profiles of the same origin.

Avoidance reflexes#

1. Define a shared goal that requires contributions from both groups, with shared indicators, not two competing dashboards.
2. Write the criteria before comparing, and apply them in the same way to both groups: budgets, deadlines, hiring, ratings.
3. Avoid rewarding competition between teams (bonuses, rankings) when they depend on each other.
4. Mix the teams. Cross-team pairs, rotations, cross-functional projects, cross reviews: contact works mainly when it is regular, between equals, around a common goal.
5. Talk about people, not camps. "Marie, from the network team" rather than "the network people."
6. Reverse the perspective. Before judging the other group, ask yourself: "If my team had done the same thing, what would I have said?"
7. Anonymize when possible: applications, code reviews, comparisons of offers.
8. Bring the divisions out in the post-mortem: look at what both teams could have done to avoid the incident.

With AI: detect, remedy, mitigate#

Prompts to build into your routines#

1. Rewrite an incident without group labels. To apply to a report before sharing it.

Here is an incident report: [text].
Rewrite it, replacing every reference to a group (“team X,” “the operations staff,” “the Y site”) with the exact role or function involved. Describe the facts in order, without attributing fault to any group.
Then list the decisions made and the missing information on each side, with the same level of detail for everyone.

2. Play the other group. To understand a situation from the opposing team's point of view.

Here is a disagreement between two teams: [description, from both points of view if possible].
Tell it from the other team's point of view: its constraints, its goals, what it expects from us, and what seems legitimate to it in its position.
End with three questions we should ask it to check this reading.

3. A shared goal for two teams. To break out of a logic of competition.

Two teams, [team A and its goals] and [team B and its goals], depend on each other for [shared deliverable or service].
Propose three measurable shared goals that cannot be achieved by one team alone, and for each: the shared indicators, the decisions to make together, and the risks of drifting into competition.

How to build them in#

• After an incident: prompt 1 produces the neutral version that serves as the basis for the post-mortem.
• Before a difficult meeting: prompt 2 prepares you to listen; you then check with the people concerned.
• When setting goals: prompt 3 feeds the discussion, not the decision.

Precautions#

• AI does not know the other team's real point of view: prompt 2 gives a hypothesis to verify, not a fact.
• Do not compare individuals or teams by name with personal data in an unauthorized tool; use aggregated data.
• Do not delegate to AI an arbitration between groups or the selection of people.

Limits#

• Favoritism is more common than hostility: the meta-analysis by Mullen, Brown and Smith (1992) confirms the effect, but its magnitude varies with the status of the groups, the salience of membership and the type of judgment requested.
• Belonging is not a flaw: it motivates, bonds and protects. The bias begins when it distorts a judgment or an allocation that should be neutral.
• Experiments with limited stakes (points, small sums) do not on their own say what happens in a merger or a restructuring; the orders of magnitude are not transferable.
• Do not confuse it with legal discrimination, which requires a protected characteristic and measurable effects; this bias may contribute to it, but does not explain it on its own.

Sources and verification#

• Tajfel, Billig, Bundy & Flament (1971), "Social categorization and intergroup behaviour", European Journal of Social Psychology, 1(2), 149-178 — the minimal groups study.
• Tajfel (1970), "Experiments in Intergroup Discrimination", Scientific American, 223(5), 96-102.
• Tajfel & Turner (1979), "An integrative theory of intergroup conflict", in Austin & Worchel (eds.), The Social Psychology of Intergroup Relations, Brooks/Cole.
• Brewer (1999), "The Psychology of Prejudice: Ingroup Love and Outgroup Hate?", Journal of Social Issues, 55(3), 429-444
• Mullen, Brown & Smith (1992), "Ingroup bias as a function of salience, relevance, and status: An integration", European Journal of Social Psychology, 22(2), 103-122.
• Sherif, Harvey, White, Hood & Sherif (1961), Intergroup Conflict and Cooperation: The Robbers Cave Experiment, University of Oklahoma Book Exchange.
• Perry (2018), The Lost Boys: Inside Muzafer Sherif's Robbers Cave Experiment, Scribe.

See also#