Testing a dynamic model of trust in AI: How trust develops and affects critical thinking in the American workforce

While trust in AI has been viewed as positive and extensively examined in existing literature, it is crucial to foster critical thinking about this innovation. Against this backdrop, this study addresses two interests. First, we propose and test a dynamic, model of trust in AI. Second, integrating critical thinking, we examine how calculative, cognitive, and affective forms of trust affect critical thinking about AI. To this end, we analyze data from a three-wave panel survey of employees from various industries in the U.S. Results from a cross-lagged panel model support a developmental model of trust. Calculative trust leads to cognitive trust, which in turn leads to affective trust. Reciprocally, affective trust leads to more cognitive trust, and cognitive trust leads to more calculative trust subsequently. Results also reveal that calculative trust predicts critical thinking about AI, but cognitive and affective trust do not. We discuss theoretical and practical implications of results.