2020/08/20 by Nicholas Lourie, Lourie, Nicholas, Ronan Le Bras +3 · 7 citations
Neuroscience · Psychology · Social Sciences · #Computation and Language (cs.CL) #Ethics and Social Impacts of AI #FOS: Computer and information sciences #Misinformation and Its Impacts #Optimism, Hope, and Well-being #Psychology of Moral and Emotional Judgment
paper · pdf · doi:10.48550/arxiv.2008.09094
openalex publication_date 2020/08/20 · openalex created_date 2022/09/08 · openalex updated_date 2026/07/28
As AI systems become an increasing part of people's everyday lives, it\nbecomes ever more important that they understand people's ethical norms.\nMotivated by descriptive ethics, a field of study that focuses on people's\ndescriptive judgments rather than theoretical prescriptions on morality, we\ninvestigate a novel, data-driven approach to machine ethics.\n We introduce Scruples, the first large-scale dataset with 625,000 ethical\njudgments over 32,000 real-life anecdotes. Each anecdote recounts a complex\nethical situation, often posing moral dilemmas, paired with a distribution of\njudgments contributed by the community members. Our dataset presents a major\nchallenge to state-of-the-art neural language models, leaving significant room\nfor improvement. However, when presented with simplified moral situations, the\nresults are considerably more promising, suggesting that neural models can\neffectively learn simpler ethical building blocks.\n A key take-away of our empirical analysis is that norms are not always\nclean-cut; many situations are naturally divisive. We present a new method to\nestimate the best possible performance on such tasks with inherently diverse\nlabel distributions, and explore likelihood functions that separate intrinsic\nfrom model uncertainty.\n