Artificial intelligence ethics by design. Evaluating public perception on the importance of ethical design principles of artificial intelligence

Abstract

Despite the immense societal importance of ethically designing artificial intelligence, little research on the public perceptions of ethical artificial intelligence principles exists. This becomes even more striking when considering that ethical artificial intelligence development has the aim to be human-centric and of benefit for the whole society. In this study, we investigate how ethical principles (explainability, fairness, security, accountability, accuracy, privacy, and machine autonomy) are weighted in comparison to each other. This is especially important, since simultaneously considering ethical principles is not only costly, but sometimes even impossible, as developers must make specific trade-off decisions. In this paper, we give first answers on the relative importance of ethical principles given a specific use case—the use of artificial intelligence in tax fraud detection. The results of a large conjoint survey ([Formula: see text]) suggest that, by and large, German respondents evaluate the ethical principles as equally important. However, subsequent cluster analysis shows that different preference models for ethically designed systems exist among the German population. These clusters substantially differ not only in the preferred ethical principles but also in the importance levels of the principles themselves. We further describe how these groups are constituted in terms of sociodemographics as well as opinions on artificial intelligence. Societal implications, as well as design challenges, are discussed.

Consumers are willing to pay a price for explainable, but not for green AI. Evidence from a choice-based conjoint analysis

“…In areas where the stakes of AI decisions or recommendations are higher, people may well care more about transparency of AI systems than with personal assistants. For instance, one can expect this to be the case in the areas of health (Juravle et al, 2020), but also in other settings where it is not the individual but the state who uses AI such as in tax fraud detection (Kieslich et al, 2021) or in the criminal justice system (Waggoner et al, 2019). Similarly, transparency may be valued more where individuals are not the users of a service but the object, as in recruiting decisions (Langer et al, 2021).…”

Ever heard of ethical AI? Investigating the salience of ethical AI issues among the German population

“…Researching the perceived importance of ethical requirements proposed by the EU, Choung, David, and Ross (2022) showed for a US sample that, again, accountability was perceived as most important. As in the study by Kieslich et al (2022), they also report that all ethical requirements are evaluated equally important. Ikkatai, Hartwig, Takanashi, and Yokoyama (2022) surveyed Japanese citizens to evaluate public attitudes towards AI ethics.…”

“…This indicates that fairness and discrimination issues may not be salient or even conceivable for a fair share of the German population. Another study by Kieslich et al (2022) researched preferences for the ethical design of an AI system in the use case of tax-fraud detection. They report that accountability was perceived as the most important ethical criterion among the respondents, while machine autonomy was perceived as the least important.…”

“…RQ1: What topics are present in the German population with regard to AI? Further, prior studies suggest that several individual characteristics may influence public opinion on AI such as interest in or knowledge of AI as well as sociodemographic aspects like gender, age, educational level or socioeconomic status (Carradore, 2021;Fietta et al, 2021;Ikkatai et al, 2022;Kieslich et al, 2022;Zhang & Dafoe, 2019). Given the one-sided media reporting about AI and the focus on specific stakeholders by economic and political actors, we suppose that AI can be considered as a topic of the elite.…”

“…Consequently, there must be public awareness of social problems for them to lead to political actions. While we have some knowledge of how citizens react to the ethical design of AI (Kieslich, Keller, & Starke, 2022;König, Wurster, & Siewert, 2021;Shin, 2021aShin, , 2021bShin, , 2021cShin, Zhong, & Biocca, 2020), we only have little to no knowledge about the salience of ethical AI among the public in the first place. Hence, we ask, what issues are top of citizens' minds and, in particular, how prominent ethical issues feature among these.…”