Atoosa Kasirzadeh and Iason Gabriel embark on an ambitious analysis of how large-scale conversational agents, such as AI language models, can be better designed to align with human values. The premise of the article is grounded in the philosophy of language and pragmatics, employing Gricean maxims and Speech Act Theory to establish the importance of context and cooperation in achieving effective and ethical linguistic communication. The authors underscore the necessity of considering pragmatic norms and concerns in the design of conversational agents and illustrate their proposition through three discursive domains: science, civic life, and creative exchange.
The authors present a novel approach, suggesting the operationalization of Gricean maxims of quantity, quality, relation, and manner, to aid in cooperative communication between humans and AI. They also emphasize the diversity of utterances, asserting that there is no single universal condition of validity that applies to all. Instead, the validity of utterances often depends on different sorts of truth conditions which require different methodologies for substantiation, based on context-specific criteria of validity. They further stress the centrality of contextual information in the design of ideal conversational agents and highlight the need for research to theorise and measure the difference between the literal and contextual meaning of utterances.
The authors also delve into the implications of their analysis for future research into the design of conversational agents. They discuss the potential for anthropomorphisation of conversational agents and the constraints that might be imposed on them. They note that while anthropomorphism can sometimes be consistent with the creation of value-aligned agents, there are situations where it might be undesirable or inappropriate. They also advocate for the exploration of the potential for conversational agents to facilitate more robust and respectful conversations through context construction and elucidation. Lastly, they suggest that their analysis could be used to evaluate the quality of interactions between conversational agents and users, providing a framework for refining both human and automatic evaluation of conversational agent performance.
The research article resonates with broader philosophical themes, particularly those concerning the interplay between technology and society. It touches upon the ethical dimensions of AI, hinting at the moral responsibility of designing AI systems that align with human values and norms. The exploration of Gricean maxims and Speech Act Theory in the context of AI conversational agents provides a unique blend of AI ethics, philosophy of language, and pragmatics, reflecting the interdisciplinary nature of contemporary AI research. In doing so, the article stimulates dialogue about the role of AI in shaping our social and communicative practices, challenging conventional boundaries between humans and machines, and highlighting the potential of AI as a tool for fostering effective and ethically sound communication.
In terms of future avenues of research, the authors’ analysis opens up a myriad of possibilities. First, while the paper focuses primarily on the English language, a fruitful direction of research could involve the exploration of norms and pragmatics in other languages, thereby ensuring the cultural inclusivity and sensitivity of AI systems. Second, the proposed alignment of AI conversational agents with Gricean maxims and discursive ideals could be further operationalized and tested empirically to assess its effectiveness in real-world scenarios. Third, the article alludes to the potential of AI in fostering more robust and respectful conversations, which suggests an opportunity to investigate how AI can play an active role in shaping discourse norms and facilitating constructive dialogues. Lastly, the authors’ work can be further enriched by drawing from other sociological and philosophical traditions, such as Luhmann’s system theory or Latour’s actor-network theory, to offer a more comprehensive and nuanced understanding of the complex interplay between AI, language, and society.
Abstract
Large-scale language technologies are increasingly used in various forms of communication with humans across different contexts. One particular use case for these technologies is conversational agents, which output natural language text in response to prompts and queries. This mode of engagement raises a number of social and ethical questions. For example, what does it mean to align conversational agents with human norms or values? Which norms or values should they be aligned with? And how can this be accomplished? In this paper, we propose a number of steps that help answer these questions. We start by developing a philosophical analysis of the building blocks of linguistic communication between conversational agents and human interlocutors. We then use this analysis to identify and formulate ideal norms of conversation that can govern successful linguistic communication between humans and conversational agents. Furthermore, we explore how these norms can be used to align conversational agents with human values across a range of different discursive domains. We conclude by discussing the practical implications of our proposal for the design of conversational agents that are aligned with these norms and values.
In Conversation with Artificial Intelligence: Aligning language Models with Human Values