Building Interpretable Models for Moral Decision-Making
Artificial Intelligence
2026-02-05 v2 Computers and Society
Machine Learning
Abstract
We build a custom transformer model to study how neural networks make moral decisions on trolley-style dilemmas. The model processes structured scenarios using embeddings that encode who is affected, how many people, and which outcome they belong to. Our 2-layer architecture achieves 77% accuracy on Moral Machine data while remaining small enough for detailed analysis. We use different interpretability techniques to uncover how moral reasoning distributes across the network, demonstrating that biases localize to distinct computational stages among other findings.
Cite
@article{arxiv.2602.03351,
title = {Building Interpretable Models for Moral Decision-Making},
author = {Mayank Goel and Aritra Das and Paras Chopra},
journal= {arXiv preprint arXiv:2602.03351},
year = {2026}
}
Comments
8 pages, 4 figures, accepted to AAAI'26 Machine Ethics Workshop