Determinants of AI Privacy Perception

Artificial Intelligence (AI) is becoming increasingly important for businesses to improve their work, and make it more efficient and user-friendly. Using user data, AI systems can create personalised recommendations in many different contexts, enabling widespread adoption of AI. Numerous determinants can influence people’s perception of AI, and it is necessary to find out how and when they influence people’s perception to be better able to influence people’s perception of AI when needed. This study therefore tries to find out whether knowledge of AI privacy implementations and the context in which an AI system is used affect people’s perception of AI. Knowledge about AI privacy implementations can differ amongst people because there are many different experiences and interests people have had, which could potentially influence people’s perceptions of AI. Additionally, AI systems are used in many different contexts, such as film recommendations and medical diagnosis, with varying degrees of importance of personal data that is used by the AI systems. The variable context will therefore be studied as a potential moderating variable in the relationship between knowledge of AI privacy implementations and perception of AI. Together, this leads to the question of “to what extent knowledge of privacy implementations in artificial intelligence has an effect on people’s perception of artificial intelligence, and whether this differs between the importance of the context”. A 2x2 between-subjects experiment (N = 131) was done to answer the research question. The two-way ANOVA test showed no significance for all effects. Knowledge of AI privacy implementations and the context in which the AI systems are used are both not significant determinants of people’s perception of AI. However, the moderating relationship was very close to significance and a power analysis revealed low power for the main analysis, so the decision was made to perform supplementary analyses. Here, perception of AI was split into trust in AI and usefulness of AI. These analyses showed that context is a significant determinant of trust in AI, and the moderating relationship between knowledge and context is also significant for trust in AI. Even though the main results were not significant, the mean differences show that less knowledge in a lower-stakes Netflix context leads to the most positive perception of AI, whereas little knowledge in a higher-stakes medical context works the opposite and creates the most negatively perceived perception of AI. This study therefore provides relevant, new and insightful information on the determinants of perception of AI and can help businesses that are working with AI.