Stereotype Negativity and the Evaluative Information Ecology

Why are most stereotypes negative — across groups, cultures, and eras — even though most social behavior people actually encounter is positive? The cognitive-ecological research program of Christian Unkelbach, Hans Alves, and colleagues answers without invoking motivational derogation, essentialist group differences, or a general negativity bias: stereotype negativity falls out of the interaction between a basic cognitive principle (groups are characterized by what distinguishes them) and two structural properties of the evaluative information environment.¹²

The two ecological asymmetries

The evaluative information ecology model (EvIE; Unkelbach, Koch & Alves 2019) posits two properties of ordinary social information:³

• Positivity prevalence — positive attributes and behaviors are more frequent. Most people are trustworthy, helpful, and non-violent most of the time; “good” is the base rate.
• Negativity diversity — negative attributes and behaviors are more diverse. There are more ways to be bad than to be good: the vocabulary for disliked persons is more differentiated, and each specific negative trait is rarer than each specific positive trait.

The mechanism: distinctiveness selects for negativity

The cognitive half of the theory is a learning principle (Kruschke’s attentional learning; Fiedler’s sampling approach): perceivers characterize a novel group by its distinct attributes — those that differentiate it from already-familiar groups — and discount attributes shared with prior groups. This is adaptive for learning (it reduces interference and accelerates acquisition of new categories), but it interacts fatally with the ecology.⁴

Because negative attributes are individually rare, they are disproportionately likely to be unshared — to distinguish one group from another. Positive attributes, being common, are disproportionately shared and thus discounted. The arithmetic is stark: if 80% of people are trustworthy (odds 4:1), then among traits shared by two groups the odds rise to 16:1, while among distinct traits they fall to 1:1. A perceiver who samples distinct attributes gets a negatively biased sample of the group — even though the group’s actual composition is mostly positive. Stereotypes can thus be diagnostic and comparatively accurate (they really do capture what best differentiates groups) while being descriptively inaccurate (they apply to only a minority of the group’s members). Simulations formalize this: rare attributes can be highly diagnostic of group membership while characterizing few members.⁵⁶

Corollary: out-groups, minorities, and novel groups carry an automatic evaluative disadvantage. Since perceivers typically encounter majority/in-group members first, later-encountered groups are the ones characterized by distinct — and therefore mostly negative — attributes. No motivation to derogate is required; the bias is a byproduct of learning order.⁷

The evidence base

• Alves, Koch & Unkelbach (2018, Psychological Science) — the canonical demonstration. Participants successively formed impressions of two fictional groups with identical positive/negative attribute counts; they preferred the first-encountered group over the novel one as long as the groups’ unique attributes were negative. When unique attributes were positive, the preference reversed — the signature of the mechanism rather than a fixed out-group bias.⁸
• Woitzel & Alves (2024, Psychological Science) — five preregistered experiments (N = 2,615) extending the effect: the novel-group disadvantage generalizes to evaluations, memory for member valence, and stereotype formation; holds with three groups; and holds regardless of whether groups are majorities/minorities or in-groups/out-groups. Learning order, not group status, drives the bias. People perceive novel groups more negatively than they actually are.⁹
• Unkelbach, Weitzel & Alves (2026) — the direct stereotype-negativity paper that prompted this page (abstract via Cowen): the theory demonstrated across two studies and two existing datasets, plus the diagnosticity simulations. It also derives when positive stereotypes should occur — when positive attributes are made diverse and rare, the ecology reverses, and novel groups come out positive.¹⁰

Honest caveat: the ecological premises under test

The model’s two premises are not equally secure. The first simultaneous empirical test — Weitzel & Unkelbach’s 2026 experience-sampling study (European Journal of Social Psychology; 453 participants, 9,413 momentary reports over 5–6 days) — found clear support for positivity prevalence (positive affect at 80% of sampling signals) but only mixed support for negativity diversity: negative descriptors were relatively more diverse, yet the result depended on the analytic method used to measure diversity. The framework’s engine room (distinctiveness-based learning) is experimentally robust; one of its two ecological fuel lines is still being measured.¹¹

The program also has boundary conditions worth keeping in view: it explains content negativity (which attributes become stereotypical), not every feature of prejudice — discrimination behavior, institutional exclusion, and motivated derogation are real phenomena the model deliberately does not need, not phenomena it denies. The intervention implication is correspondingly narrow but useful: changing the information environment (what is frequent, what is distinctive) can shift stereotype content without changing minds directly.

Connections

• dog-consonant-bias — a neighboring non-motivational account of group-flavored bias emerging from channel structure rather than derogation motive.
• trust-and-social-capital — the base-rate asymmetry (most people trustworthy most of the time) is the same ecological fact that makes trust a usable default and betrayal memorable.
• assortative-mating — exposure-vs-preference decomposition in social sorting shares the EvIE move of explaining a social pattern by the structure of what perceivers encounter rather than by their dispositions.
• norms — rare-and-diagnostic signals shaping group-level expectations, at the collective rather than perceptual level.

Sources

• Tyler Cowen 2026 — Why most stereotypes are negative
• Hans Alves, Alex Koch, and Christian Unkelbach 2018 — A Cognitive-Ecological Explanation of Intergroup Biases
• Johanna Woitzel and Hans Alves 2024 — The Formation of Negative Attitudes Toward Novel Groups
• Anne I. Weitzel and Christian Unkelbach 2026 — An Experience-Sampling Study on the Frequency and Diversity of Positive and Negative Affective States

Footnotes

1. Tyler Cowen 2026 — Why most stereotypes are negative ↩

2. Hans Alves, Alex Koch, and Christian Unkelbach 2018 — A Cognitive-Ecological Explanation of Intergroup Biases ↩

3. Anne I. Weitzel and Christian Unkelbach 2026 — An Experience-Sampling Study on the Frequency and Diversity of Positive and Negative Affective States ↩

4. Johanna Woitzel and Hans Alves 2024 — The Formation of Negative Attitudes Toward Novel Groups ↩

5. Tyler Cowen 2026 — Why most stereotypes are negative ↩

6. Johanna Woitzel and Hans Alves 2024 — The Formation of Negative Attitudes Toward Novel Groups ↩

7. Hans Alves, Alex Koch, and Christian Unkelbach 2018 — A Cognitive-Ecological Explanation of Intergroup Biases ↩

8. Hans Alves, Alex Koch, and Christian Unkelbach 2018 — A Cognitive-Ecological Explanation of Intergroup Biases ↩

9. Johanna Woitzel and Hans Alves 2024 — The Formation of Negative Attitudes Toward Novel Groups ↩

10. Tyler Cowen 2026 — Why most stereotypes are negative ↩

11. Anne I. Weitzel and Christian Unkelbach 2026 — An Experience-Sampling Study on the Frequency and Diversity of Positive and Negative Affective States ↩