Lucas J. Wiese

Demand for an AI-literate workforce has surged, in large part to counter a growing skills gap. Meanwhile, expertise in ethical and governance dimensions of AI is increasingly deemed crucial to handle various organizational, regulatory, and social concerns. However, the focus of AI literacy efforts to date has been primarily technical. This paper helps close this gap by providing the first large-scale analysis of AI ethics and governance skills sought by employers in the labor market. Drawing on more than four million job postings for AI-related professions over the years 2018-2023, we provide an empirically-grounded characterization of AI ethics and governance competencies and perform associated descriptive analyses. We find that professionals with AI ethics and governance competencies are requested by employers to hold diverse skill sets, covering technical, managerial, and regulatory domains, though the two professions remain distinct. Moreover, the demand for expertise in these domains has grown rapidly, both in absolute terms and as a proportion of AI-related job postings. More than 100,000 professionals with expertise in AI ethics and governance are now requested annually, with the concentration highest in the financial and information sectors. These findings can help individuals, employers, and institutions of higher education better design job requirements, educational programs, and individual learning pathways, closing the career competency gap in AI ethics and governance.

This chapter reviews and critiques common practices in the instruction of Artificial Intelligence (AI) ethics. We view AI ethics as an investigation – or at least a useful label – for efforts that minimise the negative impacts of AI and promote positive outcomes. Towards this aim, ‘AI ethics education’ is one prominent (or at least attempted) strategy. In this chapter, we highlight how educational attempts to promote AI ethics might not lead to the substantive goals of AI ethics very much at all.

The potential of AI technology to transform human life, well-being, and daily work is faced with numerous risks and challenges yet to be fully accounted for. However, the complexity of AI ethics makes it hard to pin down what to teach, how to teach it, and how to assess its effectiveness. Drawing on an educational perspective, this paper presents a systematic literature review and qualitative analysis of the early years of AI ethics education as a formalized field to analyze whether its future trajectory is aligned with educational best practices. Our review highlights core challenges in AI ethics education and the content, assessment, and pedagogy used in real interventions over recent years. We find that efforts to teach AI ethics do helpfully draw on a holistic view (as opposed to a narrow view), and utilize progressive pedagogies like case studies and group projects that aim to meaningfully challenge students’ ethical reasoning skills in applied practices. However, many real- world AI ethics teaching interventions do not leverage well-supported assessment techniques known to support student learning; rather, assessment is conducted primarily for research evaluative purposes. This gap in rigorous assessment raises implications for researchers and practitioners, as responsible development and use of AI will be stymied if educators cannot successfully determine whether students have truly learned relevant AI ethics content or skills.

Abstract Recent advancements in artificial intelligence (AI) and machine learning (ML) have driven research and development across multiple industries to meet national economic and technological demands. Consequently, companies are investing in AI, ML, and data analytics workforce development efforts to digitalize operations and enhance global competitiveness. As such, evidence-based educational research around ML is essential to provide a foundation for the future workforce as they face complex AI challenges. This study explored students’ conceptual ML understanding through a scientific argumentation framework, where we examined how they used evidence and reasoning to support claims about their ML models. This framework lets us gain insight into students’ conceptualizations and helped scaffold student learning via a cognitive apprenticeship model. Thirty students in a mechanical engineering classroom at Purdue University experimented with neural network ML models within a computational notebook to create visual claims (ML models) with textual explanations of their evidence and reasoning. Accordingly, we qualitatively analyzed their learning artifacts to examine their underfit, fit, and overfit models and explanations. It was found that some students tended toward technical explanations while others used visual explanations. Students with technically dominant explanations had higher proficiency in generating correctly fit models but lacked explanatory evidence. Conversely, students with visually dominant explanations provided evidence but lacked technical reasoning and were less accurate in identifying fit models. We discuss implications for both groups of students and offer future research directions to examine how positive pedagogical elements of learning design can optimize ML educational material and AI workforce development.

The release and widespread use of generative artificial intelligence causes concern for the future of teaching and learning. Since the release of ChatGPT, some institutions released guidance on its use in education, while other institutions waited for the technology to mature. This study is contextually situated during the Fall 2023 semester at a single university; Unique because the university had not published LLM guidance yet, but the technology had been out long enough for students to become familiar with its use. Through the conceptual lens of Teacher Noticing This study examined (a) whether faculty saw the potential use of LLMs for teaching and learning, and (b) how they responded to the rapid impact of LLMs in the classroom before university-standard guidance. Via document analysis, we found that despite LLM chatbots being widespread for roughly 9 months before the Fall semester, only a third of faculty acknowledged its use in the classroom. Faculty took three positions toward it: encouraged, discouraged, and prohibited. As found in qualitative analysis, most of the language was precautionary and discouraging. Through the lens of Teacher Noticing, we suggest that this is worrisome since faculty beliefs seemed to be mismatched with the enthusiasm and excitement of AI from students. Only a few months later, the university encouraged the use of creatively incorporating LLMs in the classroom to foster learning and increase students’ awareness of the limitations of the tools. In a technology department especially, instructors falling behind the curve of digital literacy may impact students’ satisfaction with their education. Future work should be done to understand how university guidance impacts faculty beliefs and how that translates to pedagogical techniques and learning outcomes.