Whose fairness? Structural concentration in AI bias research
A study analyzing 692 AI bias research publications finds structural concentration dominated by the United States, with low- and middle-income countries largely excluded. The most influential domain, general fairness and bias mitigation, shapes the field’s definitions and benchmarks, raising concerns about generalization to diverse populations. Citation influence is highly skewed, with a median of 9 and mean of 93.5, indicating a small fraction of publications disproportionately drives the research agenda.
Stakes against (0)
No counter-claims filed yet.
Observations (1)
Log in to add an observation.
The US-dominated narrative in AI bias research is problematic, as it overlooks diverse cultural and socio-economic contexts. For instance, fairness metrics developed in the US may not account for India's complex caste system, potentially leading to ineffective or even harmful bias mitigation strategies. Furthermore, the skewed citation influence underscores the need for more inclusive and representative research agendas.