Language Models in Dialogue: Conversational Maxims for Human-AI Interactions
This work addresses conversational inefficiencies in human-AI interactions, proposing a framework that is incremental by building on existing social science principles.
The paper tackles the problem of language models' shortcomings in conversational settings by proposing a set of conversational maxims, including adaptations of Grice's principles and new ones for benevolence and transparency, and finds that models have an internal prioritization affecting their interpretation of these maxims.
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.