A mixed methods study on university students' attitudes towards artificial intelligence ethics

This research aims to reveal the attitudes of university students from different academic backgrounds, representing today's users and potential stakeholders of the artificial intelligence ecosystem, towards artificial intelligence ethics, and to address how artificial intelligence systems which have vital effects shape our views on its ethical dimensions, by turning towards thinking about the possibilities of changing the "incompatible alignments of technical systems with social systems" in favor of human beings through moral philosophy. It is also aimed to increase the representation of micro-analysis level studies in the literature. In contrast to contemporary discourses on machine agency "that shift the sphere of agency from humans to machine generations", an understanding that acknowledges the interdependence of humans and artifacts is adopted, drawing strength from the idea of epiphylogenesis, which describes "the co-evolution of humans and technology". This two-stage research, which followed a sequential explanatory design from mixed methodologies, was based on the determination that documents containing ethical principles or guidelines for artificial intelligence converge around the principles of "fairness", "transparency", "non-maleficence", "privacy" and "responsibility". In the quantitative phase, the "Attitude Towards the Ethics of Artificial Intelligence" was used as a data collection tool. The instrument was translated into Turkish and 395 participant responses were obtained through an online survey. Following the validity and reliability analyses, statistically significant differences based on gender for the "Fairness", "Non-maleficence" and "Privacy" factors and academic discipline for the "Transparency" factor were discovered through difference analyses. Through qualitative research, semi-structured interviews were conducted as a data acquisition technique with a phenomenological approach, focusing on the subjective perspectives of the 16 participants on their interactions with these systems, their experiences and their relationality with the principles discussed, in order to examine the structure of the levels of difference identified in the primary phase of the relevant groups and to reach the underlying reasons for their understanding. With the main themes of "Fairness perception", "Fairness effect", "Malicious intent perception", "Malicious intent motivation", "Malicious intent outcomes", "Privacy perception", "Privacy tolerance", "Transparency perception" and "Transparency tolerance", which were determined as a result of the thematic analysis of the interview texts, an explanation that describes the essence of the higher sensitivity levels of women participants and the participants from the disciplines Humanities and Social Sciences in the relevant dimensions was tried to be put forward. Regarding the dimension of "Responsibility", for which no significant difference was found between the groups through quantitative analysis, the qualitative process revealed that the majority of the participants expect regulatory and legal frameworks that assign a multi-stakeholder responsibility to ensure that these principles are respected and that a human being is always in the 'loop'. While explaining the differences in attitudes between the groups, the research has allowed exploring the reflection of the principles that are envisioned to shape AI guidelines on Turkish students, deepening understanding by capturing country-specific nuances, and revealing legitimate concerns in all their diversity. The findings are expected to provide insights for inclusive AI development and related governance and oversight efforts that will provide common responses to central ethical issues, taking into account local values and preferences. It is hoped that the study will be useful for the development of AI ethics curricula and training materials, and for enhancing the competitive advantage of domestic businesses, especially those aiming to invest in AI. This research also points to many potential research directions.