---
id: "concept-weird-bias-in-ai"
type: "concept"
source_timestamps: ["§ Agentic AI's Diversity Challenge"]
tags: ["ai-bias", "cultural-diversity", "training-data"]
related: ["claim-weird-bias", "action-enrich-training-data", "entity-world-values-survey"]
definition: "The tendency of major foundation models to reflect the values and psychological profiles of Western, Educated, Industrialized, Rich, and Democratic populations."
sources: ["agentic"]
sourceVaultSlug: "hbr-seg-agentic"
originDay: 6
articleStem: "hbr-new-28-agent-teams-different-models"
sourceUrl: "https://hbr.org/2026/06/the-strongest-teams-of-ai-agents-will-be-built-using-different-models"
sourceTitle: "The Strongest Teams of AI Agents Will Be Built Using Different Models"
---
# WEIRD Bias in Foundation Models

WEIRD bias is the phenomenon where major large-language models (like ChatGPT) exhibit psychological profiles and values that heavily skew toward **WEIRD** populations — **W**estern, **E**ducated, **I**ndustrialized, **R**ich, and **D**emocratic societies. Research by **Atari et al.** demonstrates that these models fail to accurately capture or reflect the diversity, values, and problem-solving approaches of non-WEIRD populations.

The practical consequence: relying on a *single* major LLM inherently limits the cultural and cognitive diversity of the resulting agentic system, **regardless of how it is prompted** — which is another reason [[concept-cosmetic-ai-diversity]] fails and why the article recommends enriching training data (see [[claim-weird-bias]] and [[action-enrich-training-data]], using datasets like the [[entity-world-values-survey]]).

**Enrichment validation — STRONG:** This is the most rigorously grounded claim in the source. Atari et al. (2023), *"The Cultural Psychology of GPT,"* shows that GPT-3.5 and GPT-4 systematically align with WEIRD psychological profiles across multiple cultural-psychology benchmarks, clustering around Western patterns and failing to represent non-WEIRD norms. It builds implicitly on the broader WEIRD literature (Henrich et al.).