Algorithmic Fairness Perceptions in the Global South: Evidence from Bangladesh on Ride-Sharing, Beauty Filters, and Large Language Models

Context matters as much as outcomes in fairness judgments: Bangladeshi respondents rated an identical 20% price surge as significantly less fair during a medical emergency than during a casual trip (2.00 vs. 2.17 on a 5-point scale), suggesting that people evaluate algorithmic fairness not just by what happens, but by why and when it happens.

Demand for transparency and user control is nearly universal: Between 85.7% and 90.3% of survey participants wanted protections like transparency about how algorithms work, consent before algorithms affect them, and real control over algorithmic decisions—with this demand consistent across gender, income levels, and prior knowledge of algorithmic bias.

Beauty filters cause measurable psychological harm linked to self-image: The study found that feeling personally affected by AI beauty filters predicted reduced self-confidence with a strong correlation (R² = .30, p < .001), suggesting the harms are real and measurable rather than merely theoretical concerns.

Global South perspectives on fairness differ from Western research assumptions: By directly surveying people in Bangladesh rather than assuming fairness standards from North American and European research, the study reveals that algorithmic fairness judgments are shaped by local context and values—highlighting a significant gap in how fairness research has been conducted globally.

People already hold critical views of algorithmic systems: Respondents were already skeptical of surge pricing in ride-sharing; context didn't create fairness concerns but rather sharpened existing judgments, suggesting communities may already have developed fairness intuitions that researchers should listen to rather than teach.