Can we trust AI? An empirical investigation of trust requirements and guide to successful AI adoption

Abstract

Purpose Artificial intelligence (AI) fosters economic growth and opens up new directions for innovation. However, the diffusion of AI proceeds very slowly and falls behind, especially in comparison to other technologies. An important path leading to better adoption rates identified is trust-building. Particular requirements for trust and their relevance for AI adoption are currently insufficiently addressed.Design/methodology/approachTo close this gap, the authors follow a qualitative approach, drawing on the extended valence framework by assessing semi-structured interviews with experts from various companies.FindingsThe authors contribute to research by finding several subcategories for the three main trust dimensions ability, integrity and benevolence, thereby revealing fundamental differences for building trust in AI compared to more traditional technologies. In particular, the authors find access to knowledge, transparency, explainability, certification, as well as self-imposed standards and guidelines to be important factors that increase overall trust in AI.Originality/valueThe results show how the valence framework needs to be elaborated to become applicable to the AI context and provide further structural orientation to better understand AI adoption intentions. This may help decision-makers to identify further requirements or strategies to increase overall trust in their AI products, creating competitive and operational advantage.