Abstract:

Assignments in higher education have traditionally been evaluated based on the final products students submit, whether those products take the form of written responses, computer code, problem solutions, or project deliverables. The growing availability of generative artificial intelligence tools challenges this long-standing model of assessment. When AI systems can generate technically correct answers, functional code, or well-structured written explanations, the final artifact alone may no longer provide reliable evidence of student learning. As a result, instructors face an important question: what aspects of student work should be evaluated when AI systems can produce the visible output of an assignment?

This paper examines how assessment practices may need to evolve in courses where students have access to generative AI tools. Rather than focusing exclusively on the final product, instructors can design assignments that make students’ reasoning processes visible through documentation, testing, and revision activities. In this approach, assessment emphasizes how students interact with AI systems, how they evaluate AI-generated output, and how they apply disciplinary knowledge to refine or correct that output.

Assessing prompting and response evaluation introduces new opportunities for measuring critical thinking, problem formulation, and reflective reasoning. For example, students can be asked to submit records of their prompts alongside explanations of why those prompts were constructed in particular ways, what aspects of the AI response were useful or problematic, and how the output was modified to meet disciplinary expectations. Such artifacts make the reasoning behind AI-supported work visible to instructors and provide a richer basis for evaluating learning than the final artifact alone.

The paper discusses how assignments can be structured to support this process-oriented form of assessment, including strategies for documenting prompts, requiring verification of AI output, and incorporating reflective explanations into grading criteria.

These approaches allow instructors to evaluate how students engage with AI systems rather than simply evaluating the results those systems produce. By focusing on the processes of prompting, interpretation, and revision, educators can adapt assessment practices to AI-supported learning environments while maintaining meaningful standards of academic evaluation. In doing so, assessment shifts toward evaluating how students frame problems, interpret AI-generated information, and apply disciplinary knowledge to refine the results. This shift encourages a model of evaluation that prioritizes reasoning and judgment over simple artifact production.

Keywords:

Assessment design, student learning assessment, prompt evaluation, higher education, educational measurement, generative artificial intelligence.