2025 ASEE Annual Conference & Exposition

The increasing integration of artificial intelligence (AI) into real-world applications requires researchers and developers to critically evaluate the ethical implications of their work. However, AI ethics education is limited in technical courses. This risks developing technology that may unintentionally harm society. Here, we present results from a pilot curriculum that integrates the various ethical topics related to AI into a graduate-level machine learning course. Activities include a combination of case studies, project-based learning, and critical classroom discussions on the ethical implications of current events in AI.

Two research questions guided the study: (1) How do engineering students perceive ethical issues in AI design and implementation before taking the class? (2) How do these perceptions develop or change by the end of the AI course? This study employed a pre-post course survey method. Data were collected from 56 students enrolled in the Spring 2024 graduate machine learning course at the University of Colorado Boulder. The survey included Likert scale questions addressing perception of ethics in diverse areas such as fairness, accountability/ decision-making, transparency and privacy. It also included open-ended essay questions about ethical dilemmas in high-risk AI applications, such as medical diagnosis and employee selection processes.

A combination of qualitative and quantitative analysis of the responses was used to identify the impact of our curriculum. We conducted factor analyses of the Likert scale questions (in a previous iteration) to understand the structure of the measure, then used regression analyses to compare pre- and post-scores on the scale and subscales, and included demographic data (gender identity, program, prior training in human subjects research) to understand any differences in perception among various populations. Here, we present qualitative findings from students’ open-ended responses, collected at the beginning (pre) and end (post) of the semester, which explored their perspectives and discussions on the application of AI in healthcare and personnel selection problems. The responses were coded thematically and compared pre- to post- to track evolving perceptions. Themes were derived from the responses, with some including societal/ethical/organizational risks, human oversight, fairness, privacy and others. We will discuss the framework used to develop the curriculum, curriculum activities, findings from our analyses, and recommendations for future AI course design for engineering students.