Chatbots and human-human relationships: the need for research on potential downstream harms from generative AI

Abstract

The rise of generative AI into society via chatbots potentially heralds both positive and negative consequences for human-human relationships. This Voices article highlights four main potential downstream negative consequences that merit further empirical investigation. First, we consider whether conversational systems infused with a large language model (LLM) might often lead to reduced human-human socialization, such as by providing relational benefits to users without requiring users to make reciprocal efforts necessary to maintain the relationship. Second, we discuss whether interacting with chatbots might lead to a reduced estimation of the value of other humans, such as by making other humans seem less competent by comparison. Third, we discuss whether tendencies to treat chatbots in coldly instrumental (or even abusive) ways might spillover into worse treatment of other humans. Finally, we raise the ‘use it or lose it’ possibility that relying on chatbots to help facilitate social interactions could degrade the human ability to engage in harmonious interactions with other humans.