A computer program can solve human moral disagreements more consistently than a room full of people voting on who is right.
Combining large language models with formal logic solvers creates a system that aggregates conflicting ethical judgments with high accuracy. This framework focuses on logical consistency optimization rather than just counting the most popular opinions. People usually assume that morality is a subjective matter that can only be settled by human consensus or feeling. This research shows that the truth in a moral argument often lies in the underlying structure of the logic used. Automated systems may eventually be fairer judges of human behavior than humans themselves.
Are You the A-hole? A Fair, Multi-Perspective Ethical Reasoning Framework
arXiv · 2605.00270
Standard methods for aggregating natural language judgments, such as majority voting, often fail to produce logically consistent results when applied to high-conflict domains, treating differing opinions as noise. We propose a neuro-symbolic aggregation framework that formalizes conflict resolution through Weighted Maximum Satisfiability (MaxSAT). Our pipeline utilizes a language model to map unstructured natural language explanations into interpretable logical predicates and confidence weights.