Artificial intelligence (AI) currently acts solo or as a human companion in decisions made in several sensitive domains, such as healthcare, finance, and law. AI systems, even those carefully designed to be fair, have been heavily criticized for delivering misjudged and discriminatory outcomes against individuals and groups of people. The continuous unfair and unjust AI outcomes indicate that the significant impact of human and societal factors on AI biases is currently being overlooked. It is now urgent to view and understand AI as a sociotechnical system, inseparable from the conditions in which it is designed, developed, and deployed. This work addresses this critical issue by proposing a systematic methodology under which human cognitive biases intertwine within the overall AI lifecycle. By identifying how harmful human actions influence AI’s biases, we reveal human-to-AI biases' hidden pathways, leveraging the major human heuristics as identified in cognitive science. Central to this effort is a mapping approach that systematically connects human heuristics to specific AI biases, uncovering meaningful patterns of human–AI influence and interdependence that shape fairness outcomes. We envision that this work will inspire a genuine human-centric AI fairness approach by revealing the causes and effects of hidden biases.
Tantalaki et al. (Thu,) studied this question.