Claudia Sau King Chow
This paper uses a new measurement of AI’s intelligence related to task performance to examine how expectations about the operating efficiency of an AI technology influence the intention to adopt it. We suggest four levels of user acceptance/rejection of AI services, including the level of compensatory acceptance. Our conceptual model is specifically designed for the AI context, with two key variables: cybersecurity and anthropomorphism, and three mediating constructs: i) perceived level of AI’s intelligence, ii) perceived performance expectancy, and iii) perceived effort expectancy. The hypotheses were tested by surveying 494 potential virtual banking users in Hong Kong and analyzing the data with Structural Equation Modelling (SEM). We find that consumer acceptance of AI services is positively related to perceived performance expectancy and effort expectancy and to the perceived level of AI’s intelligence. These findings support an extended behavioral intention: the compensatory level of AI acceptance. Our empirically tested and generalizable results have implications for academics and practitioners.