Date created: | Last Updated:

Category: Uncategorized

Description: Many controversies arise from disagreements about which decision-making strategies to use in which situations (e.g., following rules or reliance on cost-benefit reasoning). In this talk, we propose a new theory of moral decision-making based on strategy selection that explains how these differences arise. Using a new paradigm with realistic moral dilemmas (across four experiments, total N=2328), we show how moral strategy selection learning (i.e., metacognitive learning) from the consequences of previous moral decisions can influence people's reliance on different decision-making strategies. Using computational modeling, we showed that many participants learned about decision strategies in general (metacognitive learning) rather than specific actions. Their learning transferred to incentive-compatible donation decisions and moral convictions beyond the experiment. Further, we use the new scenarios to compare large language model (LLM) and human moral decision-making, finding that human decisions are consistent with the strategy-selection account but not LLM decisions. LLMs show stronger omission bias and an additional type of “yes-no” bias not shared by humans. We conclude that strategy selection is an important mechanism of human moral decision-making and discuss the implications of this account. Maier, M.*, Cheung, V.*, & Lieder, F. (2023). Metacognitive Learning from Consequences of Past Choices Shapes Moral Decision-Making. https://doi.org/10.31234/osf.io/gjf3h Cheung, V.*, Maier, M.*, & Lieder, F. (2024). Large Language Models Amplify Human Biases in Moral Decision-Making. https://doi.org/10.31234/osf.io/aj46b