Testing times: AI in the classroom

Higher education assessment models must foster critical thinking and align learning with real-world demands

Educational institutions, which briefly experimented with Open Book Examinations (OBE), are steadily returning to traditional closed-book formats. At the same time, workplaces are moving in the opposite direction – incorporating AI-based automation in coding, technical writing, creative production, and more. While classrooms increasingly restrict access to information and automated tools during assessments, industries place growing value on the ability to use these tools effectively. Students often feel uncomfortable with this disconnect and question whether assessment methods in the classroom truly reflect the realities of the professional world.

To better understand this disconnect, it is important to recognise that in any organisation (industry or academia), processes are in place to ensure high-quality output. In industry, the output is products or services; in academia, it is educated individuals. While advanced AI tools can enhance industrial output, their indiscriminate use in education risks diluting the very quality of the academic outcomes – students’ learning and understanding. Assessment strategies need to be refined so they better address the students’ concerns.

Earlier, traditional closed-book, invigilated examinations dominated assessment systems in universities. During that period, OBEs were promoted as a way to provide more authentic summative assessments that reflect the real-world expectations. These exams help students synthesise available information to answer questions that assess higher-order thinking skills. Additionally, this approach can reduce test anxiety and promote academic honesty.

However, the rise of artificial intelligence tools such as ChatGPT and Gemini has pushed examinations back to square one, with faculty members favouring closed-book, time-bound, and invigilated formats. This shift underscores the tension between technological innovation and earlier innovations in assessment. While industries can leverage technological innovations to enhance efficiency, the field of education must resist the temptation to let AI technology dictate assessment practices. After all, education should equip students with the ability to think critically and solve problems. Generating assessment responses using AI tools can significantly hinder the development of these skills.

The more important question, then, is not simply whether assessments should be designed on an open-book or closed-book format, or how the use of AI-based tools can be restricted. Rather, we should ask what skills can be assessed to help students thrive in an AI-dominated world, how AI tools can be used ethically, and how the assessments should be designed.

The process behind the answers

In a world where AI can quickly generate answers, the skills that matter most are the ability to evaluate, critique, and synthesise information. AI can suggest solutions, but it cannot truly bring together a student’s personal experiences, local realities, or ethical concerns. Instead of asking students to summarise a research paper or case study, we can ask them to examine two conflicting responses generated by AI and decide which is more convincing and why. Although AI can produce arguments, it still lacks the judgment required to question its own assumptions or identify their logical flaws.

In addition to what needs to be assessed, it is equally important to focus on how the assessments should be designed such that the focus is not only on the final answer but on the thinking behind that answer. It is important to pay attention not just to the product but also to the process of assessment. Such an approach values the documentation of thinking, revision, and meta-cognitive reflection, alongside the outcome. Process journals, drafts, and structured peer discussions can show how students rethink ideas, respond to feedback, and improve their work over time. These aspects of learning are difficult for AI to imitate.

Another useful approach is to adopt multiple methods of assessment, including viva voce, presentations, and classroom discussions. When students explain their ideas in real time, answer unexpected questions, and defend their views, they show a genuine understanding and ownership of their learning—something a machine cannot easily replicate.

These choices of skills, methods, and assessment designs may suffice for now, but we must recognise that AI models will continue to evolve, bringing new challenges. The rapid pace at which the machines learn compared to humans means that technology will constantly push the frontiers, requiring us to adapt continuously. The role of the educators is to anticipate these changes and develop strategies that align with technological advancements. Rather than attempting to prevent the use of AI, we should accept that AI is here to stay and design assessments that evaluate how effectively students collaborate with it as an intellectual tool.

The writer teaches at Azim Premji University

(Disclaimer: The views expressed above are the author's own. They do not necessarily reflect the views of DH.)