In a recent talk, Punya Mishra claimed that “whether we like it or not we will start treating these bots as a if they are a psychological real other.” There’s quite a lot of evidence from social psychology going back to the 1990’s that humans consistently (and unconsciously) treat computers as social actors. This paradigm has been further refined in recent years, but overall there’s evidence that we do treat technologies in social ways (e.g., by being polite).

In a paper that we published this month, we show that learners imagine their interactions with an AI as abiding by social processes, including encompassing issues such as respect, honesty, and trust. Equally importantly, this finding isn’t uniform. At times learners imagine AI as a tool/object that can be used in the service of learning, while other times they imagine AI as a subject as one who has agency and possibly some kind of internal subjectivity.

Here’s the paper

Veletsianos, G., Houlden, S., & Johnson, N. (in press). Is Artificial Intelligence in education an object or a subject? Evidence from a story completion exercise on learner-AI interactions. Tech Trends. The final version is available at https://doi.org/10.1007/s11528-024-00942-5 but here is a public pre-print version.

Abstract

Much of the literature on artificial intelligence (AI) in education imagines AI as a tool in the service of teaching and learning. Is such a one-way relationship all that exists between AI and learners? In this paper we report on a thematic analysis of 92 participant responses to a story completion exercise which asked them to describe a classroom agreement between an AI instructor and a learner twenty years into the future. Using a relational theoretical framework, we find that the classroom agreements between AI and learners that participants produced encompassed elements of education, boundaries, affordances, and social conventions. These findings suggest that the ways learners relate to AI vary. Some learners relate to AI as an object, others relate to AI as a subject, and some relate to AI both as an object and a subject. These results invite a deeper engagement with the ways in which learners might relate to AI and the kinds of ethics and social protocols that such relations suggest.