AI Alignment research seeks to align human and AI goals to ensure independent actions by a machine are always ethical. This paper argues empathy is necessary for this task, despite being often neglected in favor of more deductive approaches. We offer an inside-out approach that grounds morality within the context of the brain as a basis for algorithmically understanding ethics and empathy. These arguments are justified via a survey of relevant literature. The paper concludes with a suggested experimental approach to future research and some initial experimental observations.

翻译：摘要：人工智能对齐研究旨在协调人类与人工智能的目标，以确保机器的自主行为始终符合伦理规范。本文认为，尽管共情常被更偏向演绎的方法所忽视，但它对这一任务至关重要。我们提出一种由内而外的路径，将道德植根于大脑的语境中，以此作为算法理解伦理与共情的基础。通过系统梳理相关文献，这些论点得到了论证。本文以建议未来研究的实验方法及初步实验观察作为结论。