How does AI affect decision makers in the public sector?
Madalina Busuioc, co-director of the AIDG Lab, investigates the cognitive biases that arise among decision makers when using artificial intelligence tools in decision-making.
Currently, artificial intelligence algorithms are used as decision aids by human decision makers in a variety of bureaucratic frameworks. In three experimental studies, the research explicitly examines how decision makers rely on algorithmic recommendations and what impact this has on their decision-making. Two possible types of bias formed the focus of the study: automation bias - unwarranted overreliance on algorithmic advice despite counter-evidence and selective adherence - biased adherence to algorithmic advice when predictions confirm what decision-makers already think or believe.
The study finds evidence of selective adherence to algorithms: namely, increased deference to algorithmic recommendations when predictions match prevailing stereotypes, increasing bias in decision-making at the expense of already marginalized communities and individuals. The research helps unravel an important cognitive mechanism by which decision bias may worsen in interactions with automation.