Notes
Trust Me, I'm AI: Examining the Influence of XAI and/or AI-Seal on User Trust and Understanding
231In recent years, artificial intelligence (AI) has become increasingly relevant and is being applied in numerous areas. Due to the rapid development, the performance of the algorithms is already so advanced, even experts are not always able to comprehend the actions (Kaplan, 2017). However, since the majority of AI-users do not have detailed knowledge about AI, their perception of AI-systems is primarily that of a black-box, resulting in mistrust and reduced acceptance of AI in society (von Eschenbach, 2021). To address this issue, the concept of explainable artificial intelligence (XAI) has been proposed as a method to provide insights into the often hidden logic of algorithms, using explanations to increase understanding and trust in AI-systems (Das & Rad, 2020). Additionally, the AI-Seal has been introduced as a certification measure to ensure safe and reliable AI applications. However, both methods, apart from promising advantages, still have downsides, such as overly complex explanations causing mistrust (Papenmeier et al., 2019) or the potential for blind trust in the AI-Seal (Kloker et al., 2022). Therefore, the question arises as to whether a combination of both XAI and AI-Seal can potentially maximize benefits while minimizing the disadvantages, considering the different ways of processing XAI and an AI-Seal. Additionally, previous AI-research has largely overlooked human-related factors which could play a crucial role in increasing understanding and trust in AI. Thus, this online-experiment (N = 436) used a 2 (XAI vs. no XAI) x 2 (AI-Seal vs. no AI-Seal) between-subject design in which participants interacted with an AI-system, to analyze the effectiveness of XAI, AI-Seal, or a combination of both in increasing understanding and trust in AI, while also considering the influence of human-related factors such as a persons’ need for cognition and the propensity to trust in technology.
232Artificial intelligence / explainable-AI / trust / understanding / AI-seal