Increasing Algorithm Appreciation in AI-based Decision Support Systems through Encouraging Theory of Machine

Abstract

While the use of artificial intelligence (AI) for decision-making is widespread, the technology cannot be fully realized when the end-user mistrusts it. To increase algorithm appreciation, literature supports the idea of clarifying how AI works. Rather than presenting AI as a black box, the framework Theory of Machine proposes an approach to explain artificial intelligence to end-users by contrasting it to human thinking. The present research examines the effects of Theory of Machine priming on algorithm appreciation. One hundred twenty-eight participants were randomly assigned to a priming condition where artificial intelligence was introduced as in Theory of Machine or as a black box. Namely, comparing human to algorithmic judgment or by only giving technical descriptions of AI’s reasoning, respectively. Afterwards, participants performed an age guessing game where an algorithm aided as a decision support system. The extent to which participants aligned their answers to the algorithm’s advice was used as a measurement for algorithm appreciation (weight of advice). Additionally, task difficulty was manipulated to explore possible moderation effects. Based on previous literature it was hypothesized that a Theory of Machine framing will increase algorithm appreciation compared to the Black Box framing. The hypothesis was not confirmed. The results showed no significant difference between the means of the weight on advice score on the framing conditions or task difficulty. We conclude that a Theory of Machine framing does not influence algorithm appreciation. Several explanations for this effect, limitations of the study, and suggestions for further research on how to increase trust in AI are considered.