Skin-in-the-Game: Decision Making via Multi-Stakeholder Alignment in LLMs

TL;DR

This paper presents the Skin-in-the-Game (SKIG) framework that improves moral reasoning in large language models by simulating stakeholder accountability, empathy, and risk assessment, validated across multiple benchmarks.

Contribution

The paper introduces SKIG, a novel framework that enhances LLMs' moral reasoning by incorporating multi-stakeholder perspectives and accountability mechanisms.

Findings

SKIG improves performance on moral reasoning benchmarks.

Accountability simulation enhances ethical decision-making.

Extensive ablation shows key components' importance.

Abstract

Large Language Models (LLMs) have shown remarkable capabilities in tasks such as summarization, arithmetic reasoning, and question answering. However, they encounter significant challenges in the domain of moral reasoning and ethical decision-making, especially in complex scenarios with multiple stakeholders. This paper introduces the Skin-in-the-Game (SKIG) framework, aimed at enhancing moral reasoning in LLMs by exploring decisions' consequences from multiple stakeholder perspectives. Central to SKIG's mechanism is simulating accountability for actions, which, alongside empathy exercises and risk assessment, is pivotal to its effectiveness. We validate SKIG's performance across various moral reasoning benchmarks with proprietary and opensource LLMs, and investigate its crucial components through extensive ablation analyses.