Effective altruism (EA) emphasizes maximizing the social benefit of donated resources, yet it remains unclear whether people can learn efficiency information from repeated outcome feedback and whether such learned efficiency representations subsequently guide charitable choice. To investigate this issue, we employed reinforcement learning (RL) computational modeling and electroencephalography (EEG), using a self-prosocial contrast framework (reward task vs. charity task) to examine whether donors can learn efficiency and amount information across multiple trials and use these learned representations to guide subsequent decisions. Results from Experiment 1 showed that in the charity task, efficiency information was assigned a higher value weight and updated with a lower learning rate, with donation preferences primarily driven by efficiency differences. In contrast, amount information dominated choices in the reward task. Experiment 2 replicated this "efficiency-first" behavioral pattern and key computational findings under a revised feedback framework. EEG results further revealed a significant negative-going sensor-level EEG cluster over right temporoparietal scalp electrodes for amount information, with weaker neural responses in the charity task than in the reward task. These findings demonstrate that individuals can learn and utilize efficiency information to guide charitable decisions, supporting the practical feasibility of EA's rational altruism framework.
He et al. (Mon,) studied this question.