Language Models in Dialogue: Conversational Maxims for Human-AI Interactions

TL;DR

This paper introduces a set of conversational maxims tailored for human-AI interactions, analyzing how language models adhere to these principles and proposing new maxims for transparency and benevolence to improve dialogue quality.

Contribution

It extends Grice's conversational maxims by adding benevolence and transparency, and evaluates language models' understanding of these principles in dialogue settings.

Findings

Language models show varying adherence to the maxims.

Models prioritize certain principles, affecting their understanding.

New maxims help address AI-specific conversational challenges.

Abstract

Modern language models, while sophisticated, exhibit some inherent shortcomings, particularly in conversational settings. We claim that many of the observed shortcomings can be attributed to violation of one or more conversational principles. By drawing upon extensive research from both the social science and AI communities, we propose a set of maxims -- quantity, quality, relevance, manner, benevolence, and transparency -- for describing effective human-AI conversation. We first justify the applicability of the first four maxims (from Grice) in the context of human-AI interactions. We then argue that two new maxims, benevolence (concerning the generation of, and engagement with, harmful content) and transparency (concerning recognition of one's knowledge boundaries, operational constraints, and intents), are necessary for addressing behavior unique to modern human-AI interactions. We evaluate the degree to which various language models are able to understand these maxims and find that models possess an internal prioritization of principles that can significantly impact their ability to interpret the maxims accurately.