Artificial intelligence and the new forms of interaction: Who has the control when interacting with a chatbot?

Advances in artificial intelligence provide new tools of digital assistance that retailers can use to support consumers while shopping. The aim of this research is to examine how consumers react as a function of assistants' appearance (human- vs. not human-like) and activation (automatic vs. human-initiated). We advance a model of sequential mediation whose empirical validation on 400 participants in two studies shows that nonanthropomorphic digital assistants lead to higher psychological reactance. In turn, reactance affects perceived choice difficulty, which positively reflects on choice certainty, perceived performance and-ultimately-satisfaction. Thus, although reactance might appear as a negative outcome, it eventually leads to higher satisfaction. Furthermore, initiation (system vs. user initiation) does not activate the chain of effects, but significantly interacts with anthropomorphism so that individuals exhibit lower reactance when confronted with human-like digital assistants activated by the consumer. Overall, reactance is highest for non-human like digital assistants that are computer-initiated.