Trustworthy artificial intelligence

: the impact of certification labels on end-users’ trust and intention to use

Abstract

With the growing influence of Artificial Intelligence (AI) systems, concerns about their trustworthiness have emerged. Therefore, this dissertation explores the critical issue of trust in AI and introduces certification labels as a method to enhance end users' trustworthiness perceptions of AI. To uncover whether a certification label for trustworthy AI (TAI) has the potential to increase end-users’ trust in and the acceptance of AI systems, two experimental studies were conducted in the scope of this research. Study 1 found that a certification label for TAI can have a positive impact on end-users’ trust in and the acceptance of AI systems. Study 2 identified that transparently communicating the requirements used for the certification on the certification label translates into higher levels of trust in the AI system. Moreover, it was examined that the requirement Transparency was perceived as most important regarding AI trustworthiness. The results from these studies have clear implications for policymakers, developers, and organizations that seek to enhance the trustworthiness of AI, suggesting that certification labels for TAI are an effective method to communicate trustworthiness of AI systems to end-users and thereby to increase the acceptance of their AI-based products and services. Considering the potential competitive advantage of embedding TAI in products and services, the dissertation underscores the relevance of TAI in shaping the future landscape of AI technologies. Further, the findings also complement the knowledge on labels (in general) and the effectiveness of different label designs.

Date of Award	26 Jan 2024
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Filipa de Almeida (Supervisor)