Participant samples in AI trust research skew WEIRD: mostly Western, adult, university-educated participants, limiting external validity¶

Claim · All claims
Evidence · 1 study · review r2 · q3
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i? The review finds demographic patterns pointing toward systematic WEIRD bias in participant recruitment, which may limit the external validity of the reviewed studies. → Abramson SL 2026

Evidence¶

Abramson SL 2026¶

Abramson SL, Sora-Cardenas J, Nandakumar P, Bruce JE, Gopinath N, Zangari LG and Feijóo-García PG (2026). Trust and interaction design in AI-enabled systems: a systematic literature review. Front. Psychol. 17:1932655. https://doi.org/10.3389/fpsyg.2026.1932655

q3 · i? · review · r2

Across the 33-study corpus, population sizes ranged from n = 10 to n = 1,511 (mean n = 245.73); 33.33% of studies reported Western-country participants, 84.84% reported participants aged 18 or older, and 27.27% involved university-degree holders or pursuers.

"These demographic patterns point toward a systematic bias in participant recruitment, consistent with the WEIRD framework (i.e., Western, Educated, Industrialized, Rich, and Democratic) (Henrichetal.,2010 )."

Discussion¶

Related Claims¶

• Nine interaction design factors recur in post-2023 research on trust in AI-enabled systems, with explainability the most studied — related
• Self-report remains the dominant trust measurement method in post-2023 AI trust research, used as the sole measure in 72.73% of studies — related
• The study's invariance findings may be biased by excluding students who missed a test and by sample heterogeneity — related
• AI-based proctoring systems trained on western behavioural norms may exacerbate cultural and social challenges in non-English-speaking regions — related
• Sample truncation based on at-risk status can induce collider bias that undermines internal as well as external validity — related
• Students showed no statistically significant apprehension about trusting automated feedback validity, and favoured granular automated marginalia over generalised human feedback — related
• English dominates CA interaction languages: 17 of 23 prototypes relied exclusively on English even though only seven studies originated from English-speaking countries — a narrower finding that bears on this claim
• Automated grading systems may produce systematically inequitable outcomes for international and non-Western students — related
• The interleaving literature is dominated by laboratory studies of university undergraduates, with no included study using high-school participants — related
• Only 15% of published polygenic score estimates for education come from non-European ancestry samples, all recruited in the US — a narrower finding that bears on this claim
• Content validity evidence was reported for only eight assessments and four screeners, and over half of tools were developed in WEIRD societies or in English — related