A Hypothesis-Driven Framework for the Analysis of Self-Rationalising Models

TL;DR

This paper introduces a hypothesis-driven statistical framework using Bayesian networks to analyze and compare the decision processes of self-rationalising LLMs, aiming to assess explanation faithfulness and interpretability.

Contribution

It proposes a novel framework that models LLM decision processes with Bayesian networks and compares generated explanations to evaluate their alignment.

Findings

Models did not strongly resemble GPT-3.5 decision processes.

Framework enables systematic comparison of LLM explanations with structured models.

Potential to improve LLM interpretability in future work.

Abstract

The self-rationalising capabilities of LLMs are appealing because the generated explanations can give insights into the plausibility of the predictions. However, how faithful the explanations are to the predictions is questionable, raising the need to explore the patterns behind them further. To this end, we propose a hypothesis-driven statistical framework. We use a Bayesian network to implement a hypothesis about how a task (in our example, natural language inference) is solved, and its internal states are translated into natural language with templates. Those explanations are then compared to LLM-generated free-text explanations using automatic and human evaluations. This allows us to judge how similar the LLM's and the Bayesian network's decision processes are. We demonstrate the usage of our framework with an example hypothesis and two realisations in Bayesian networks. The resulting models do not exhibit a strong similarity to GPT-3.5. We discuss the implications of this as well as the framework's potential to approximate LLM decisions better in future work.