The Problem with AI Literacy

Almost every piece of AI curriculum for faculty we have seen, including ones we have developed, start with “AI Literacy” and we understand why. The UNESCO framework positions AI learning as universal across four dimensions. Similarly, the popular EDUCAUSE model is defined as follows, “AI Literacy in Teaching and Learning (ALTL) involves understanding the fundamentals of how AI works.” The authors go on to identify four key universal areas of concern: Technical, Evaluative, Practical, and Ethical. All of these things are important and we don’t take issue with any of them in particular, but we do think the term “AI Literacy” as a singular concept is inaccurate and counterproductive. 

Let us consider the closest analogs from recent history. Media literacy and computer literacy are two commonly understood concepts that we have emphasized in higher education for some time. However, we do not treat either literacy framework as a singular concept. Media literacy means something quite different for instructors of journalism, political science, and when teaching in general education. Similarly, computer literacy for a computer engineering student is quite distinct from the same concept for a nursing student. 

We need a similarly nuanced approach to AI literacy for different disciplines. AI literacy for a computer science student will mean exposure to how models are trained and their capabilities for generating code. An English student probably does not need these capacities, but an understanding of literary form and writing capabilities would be essential. 

We urge practitioners in higher education to move toward a pluralistic “AI Literacies” approach to the technology that reflects its varied impact on different fields. This is a linguistic move, but it is an important one that recognizes the disciplinary differences of the technology.