Argumentative Large Language Models for Explainable and Contestable Decision-Making

The diversity of knowledge encoded in large language models (LLMs) and their ability to apply this knowledge zero-shot in a range of settings makes them a promising candidate for use in decisionmaking. However, they are currently limited by their inability to reliably provide outputs which are explainable and contestable. In this paper, we attempt to reconcile these strengths and weaknesses by introducing a method for supplementing LLMs with argumentative reasoning. Concretely, we introduce argumentative LLMs, a method utilising LLMs to construct argumentation frameworks, which then serve as the basis for formal reasoning in decisionmaking. The interpretable nature of these argumentation frameworks and formal reasoning means that any decision made by the supplemented LLM may be naturally explained to, and contested by, humans.

Previous methods for improving the reasoning of LLMs do not necessitate a direct relationship between the reasoning steps and the final decision. Our argumentative approach, on the other hand, provides this as a feature of the system. This is because the system prediction is directly derived from the generated argumentation framework using a formally defined and deterministic procedure, thus providing faithful explainability. Further, argumentative LLMs also provide a guarantee of contestability, in that if a human intervenes in the reasoning process (such as by adding or removing an argument, or changing the strength of an argument), this will have a measurable effect on the output of the decision-making system.

Rather than prompting an LLM to produce ‘thoughts’, as in Wei et al. [2022] or Yao et al. [2023], that either enrich the context of the LLM, or provide disparate reasoning steps to compare, our approach can be seen as providing ‘thoughts’ for and against particular outputs, in the spirit of Miller [2023]. This makes it a natural fit for highly complex decision-making tasks, wherein an option, or set of options, must be chosen from a number of possible alternatives.