Social decision-making under uncertainty

Social decisions are fraught with uncertainty. In my thesis, I explore how uncertainty shapes social decision-making across three empirical projects (9 studies, total N = 3,346). I start by comparing social and individual decisions under risk (i.e., choices over outcomes with known probabilities) across a series of online studies. I find that choices for others are markedly more risk-averse than choices for self because they are more attuned to disadvantageous prospects. Supporting the hypothesis that enhanced uncertainty in social decisions might produce such a precautionary pattern, I observe greater indecision and less confidence in social relative to individual choices, while motivational differences cannot explain their divergence. As a next step, I set out to experimentally manipulate uncertainty about others’ experiences of outcomes in a laboratory experiment. My observations suggest that enhanced uncertainty about others’ experience plays a supporting role in a particular manifestation of precautionary preferences for others, namely a stronger aversion to harm others than oneself. Finally, I consider how uncertainty shapes interactions between humans and artificially intelligent machines as a broader category of social decisions. I find that people’s need to explain others’ minds and actions extends to interactions with machines. This demand to reduce uncertainty about machines’ inner workings is more pronounced for settings involving high stakes and/or scarcity – a pattern paralleling demand for explanation in realms evading machines. Two central insights emerge: first, enhanced uncertainty about others’ experience of outcomes in social choices elicits precautionary preferences whereby decision-makers minimise potential harm to others. Second, humans’ notorious need to reduce uncertainty about others’ internal states extends to interactions with artificial social agents, where it translates into a robust demand for interpretability of algorithmic decisions.