Students need more than AI skills—they need accountable AI literacy

As colleges and universities in Canada welcome students back to campus, many will encounter generative artificial intelligence (GenAI) as a regular part of their academic experience. Students may utilize tools like ChatGPT and Gemini for tasks such as brainstorming, summarizing readings, refining writing, and crafting job applications. The urgent question now is not whether students will employ AI, but how universities will teach them to use it judiciously and responsibly.

Canada's government has introduced its national artificial intelligence strategy, AI for All, which intertwines AI with public trust, economic opportunity, and Canadian sovereignty. This strategy also encompasses a National AI Literacy Initiative targeting postsecondary students and educators. However, mastering the use of an AI system is not equivalent to attaining AI literacy.

Students require more than the capacity to formulate effective prompts or expedite tasks; they need accountable AI literacy—the ability to elucidate why AI was employed, evaluate its output, and assume responsibility for the resulting work.

While teaching students how to communicate with AI might prove beneficial, it should enhance rather than supplant human judgment. Prompt engineering, the process of devising and refining instructions to elicit a useful response from GenAI, can be a legitimate educational tool. Structured prompting enables students to deliberate, test, and refine their interactions with AI, thereby making the rationale behind their actions transparent.

However, even a well-crafted response can still be inaccurate, biased, or inadequately supported. GenAI systems may generate falsehoods with unwarranted confidence, fabricate sources, obscure uncertainty, or produce claims that students lack the requisite knowledge to assess. Consequently, we must exercise caution in depending on AI systems as reliable tutors or collaborators.

The question of accountability arises: to whom and for what are students accountable? The answer includes individuals who read, evaluate, or rely on their work, such as instructors, classmates, employers, clients, and members of the public. In professional contexts, AI-assisted communication may impact patients, employees, customers, or communities that had no input into the decision to utilize AI.

Responsibility begins with factual accuracy but extends beyond that. Students should be able to explain the use of AI in an assignment, how the AI's claims were verified, whether confidential information was input, and why the final work meets the standards of the course or profession.

UNESCO's AI competency framework for students integrates practical knowledge with ethics, human agency, and responsible citizenship. Students should possess sufficient subject matter expertise to recognize weak or misleading outputs. Claims that carry significant consequences necessitate verification rather than mere acceptance due to their authoritative appearance.

Disclosure requirements entail identifying AI's contribution while retaining personal accountability for the final work. AI literacy also involves understanding when not to utilize AI. Entering personal data into commercial platforms, delegating high-stakes decisions, or automating assignments intended to cultivate foundational skills may be inappropriate, even when the technology can perform such tasks.

Accountability should not amount to shifting all risk onto individual students. Universities must establish clear expectations and devise assessments that render students' reasoning visible. Rather than solely relying on AI-detection software, instructors can request students to document revisions, justify their choices, and reflect on the extent to which AI improved or hindered their work.

Pilot studies exploring the integration of GenAI into educational assessments propose that instructors classify tasks based on whether AI is prohibited, allowed, or actively employed. This approach provides students with clearer guidelines than either strict prohibitions or ambiguous permissions. Universities, colleges, and schools must also ponder the platforms they introduce into classrooms.

Mandating the use of a commercial AI product may expose students' data and amplify universities' reliance on private technology providers. AI companies bear responsibility for system design, data practices, and claims regarding the reliability of their products. Ensuring students can verify AI outputs does not absolve technology providers of their accountability for faulty systems.