Explaining Large Language Models Decisions Using Shapley Values

TL;DR

This paper introduces a Shapley value-based method to interpret large language model decisions, revealing how prompt components influence outputs and exposing token noise effects that challenge the use of LLMs as human behavior proxies.

Contribution

It presents a novel, model-agnostic approach using Shapley values to analyze prompt influence and identify biases in LLM decision-making processes.

Findings

Shapley values help quantify prompt component contributions.

Token noise effects can disproportionately influence LLM decisions.

Caution is needed when using LLMs as human behavior substitutes.

Abstract

The emergence of large language models (LLMs) has opened up exciting possibilities for simulating human behavior and cognitive processes, with potential applications in various domains, including marketing research and consumer behavior analysis. However, the validity of utilizing LLMs as stand-ins for human subjects remains uncertain due to glaring divergences that suggest fundamentally different underlying processes at play and the sensitivity of LLM responses to prompt variations. This paper presents a novel approach based on Shapley values from cooperative game theory to interpret LLM behavior and quantify the relative contribution of each prompt component to the model's output. Through two applications - a discrete choice experiment and an investigation of cognitive biases - we demonstrate how the Shapley value method can uncover what we term "token noise" effects, a phenomenon where LLM decisions are disproportionately influenced by tokens providing minimal informative content. This phenomenon raises concerns about the robustness and generalizability of insights obtained from LLMs in the context of human behavior simulation. Our model-agnostic approach extends its utility to proprietary LLMs, providing a valuable tool for practitioners and researchers to strategically optimize prompts and mitigate apparent cognitive biases. Our findings underscore the need for a more nuanced understanding of the factors driving LLM responses before relying on them as substitutes for human subjects in survey settings. We emphasize the importance of researchers reporting results conditioned on specific prompt templates and exercising caution when drawing parallels between human behavior and LLMs.