The computing capacity of Autonomous Vehicles (AVs) has allowed the public to rediscover the classic trolley dilemma in a modern context. This paper aims to present an in-depth explanation of driver’s moral reasoning for AV moral dilemma situations. Moral dilemma vignettes for AVs were designed based on real crash data and in- depth interviews with drivers. With the vignettes, a thought experiment with 33 participants was conducted; think aloud method and open-ended interviews were used to examine participants’ reasoning processes. This paper qualitatively interpreted the data by thematic analysis. The results suggest that 1) human drivers’ moral reasoning relies on diverse moral values and 2) reasoning can be explained based on safety, justice, and crash context. The results can be used as an analysis and communication tool for AV engineers and machine ethicists to determine how well current AV algorithms convey actual human moral reasoning.