The effects of human training data (HTD) explanation on purchase intention for artificial intelligence (AI) technologies

Abstract

Consumers exhibit resistance to Artificial Intelligence (AI) technologies despite their widespread deployment. One common thread linking various reasons for this is the perceptions of AI technologies’ lack of inherent human qualities. This paper explores the effect of human training data (HTD) explanation on purchase intention for AI technologies as a potential source-based explainable AI (XAI) solution to this problem. We define HTD explanation as a statement that describes the role that humans play within a given AI system’s training data in lay terms that is easy to understand for end-users. This includes descriptions of what human data is used and for what purpose (e.g., “Our ready-made meals are developed by an AI system trained on millions of people’s food and taste preferences”). We predict HTD explanation to have a positive effect on consumer purchase intention that is mediated by perceived transfer of human essence. We tested our hypotheses in four between-subjects experiments. Study 1 supports the proposed main effect. Studies 2 and 3 show evidence for the mediation mechanism. Study 4 provides marginal evidence that HTD explanation may reduce adoption intention among consumers high pre-existing trust in AI technologies. Our findings suggest that the psychological processes linking HTD explanation to consumer outcomes may be more complex than originally theorized. This study introduces HTD as a form of source-based XAI solution and offers preliminary insight into PET and deep humanization effects, where consumers may perceive AI technologies to possess human qualities by virtue of HTD use.