Procedural Dilemma Generation for Evaluating Moral Reasoning in Humans and Language Models

TL;DR

This paper introduces a framework for generating diverse moral dilemmas to evaluate moral reasoning in humans and AI, revealing patterns in judgments and highlighting areas for improvement in scenario design.

Contribution

It presents a novel procedural generation method for moral dilemmas and a large benchmark dataset for assessing moral reasoning in language models.

Findings

Both humans and models rated harm as less permissible when it was a necessary means.

Harmful outcomes that are evitable received higher permissibility ratings.

No significant difference based on whether harm resulted from action or omission.

Abstract

As AI systems like language models are increasingly integrated into decision-making processes affecting people's lives, it's critical to ensure that these systems have sound moral reasoning. To test whether they do, we need to develop systematic evaluations. We provide a framework that uses a language model to translate causal graphs that capture key aspects of moral dilemmas into prompt templates. With this framework, we procedurally generated a large and diverse set of moral dilemmas -- the OffTheRails benchmark -- consisting of 50 scenarios and 400 unique test items. We collected moral permissibility and intention judgments from human participants for a subset of our items and compared these judgments to those from two language models (GPT-4 and Claude-2) across eight conditions. We find that moral dilemmas in which the harm is a necessary means (as compared to a side effect) resulted in lower permissibility and higher intention ratings for both participants and language models. The same pattern was observed for evitable versus inevitable harmful outcomes. However, there was no clear effect of whether the harm resulted from an agent's action versus from having omitted to act. We discuss limitations of our prompt generation pipeline and opportunities for improving scenarios to increase the strength of experimental effects.