Hey Siri, should I keep this title? How algorithms support (my) decision making under uncertainty

Algorithms are increasingly present in everyday life. These tools have moved beyond their traditional online habitats to analysing data about how people choose to commute or browse a dinner menu. This thesis investigated when and why people use algorithms as decision aids to guide their choices. I examined this central question in three experimental designs that spanned from applied investigations with radiologists to studies in the laboratory and then returning to examining algorithm-based scenarios in the real-world. Across 14 experiments, our results highlight the central role of the individual’s knowledge about the algorithm. In various formats, we provided individuals with algorithm performance information that allowed them to compare their own abilities (or the abilities of others) to an algorithm. We show that once equipped with this information, individuals adopt a strategy of selective reliance on the decision aid. That is, individuals ignore the algorithm when their abilities surpass the algorithm and appropriately defer to it when faced with choices under uncertainty. Our systematic investigations show that further opportunities to learn about the algorithm not only encourage reliance on its recommendations but also engagement in experimentation and verification of one’s knowledge about its capabilities. Together, our findings emphasise the decision-maker’s capacity to learn about the algorithm to provide insights for how we can improve the use of decision aids