Securing Academic Integrity: Strategies to AI-Proof Exam Questions

While AI tools can enhance the educational experience, they also pose significant risks to the integrity of traditional assessment methods. The ability of AI-powered applications to generate, solve, and even predict exam questions has raised concerns among faculty about the fairness and effectiveness of exams as a measure of student learning.

The Need to Design Exam Questions Differently

As AI advances, the need to design exam questions that can effectively assess a student's true understanding, critical thinking, and problem-solving skills has become more pressing than ever. The traditional question formats, once considered robust, are now vulnerable to being easily deciphered by AI tools, leaving educators searching for innovative strategies to safeguard the authenticity of their assessments.

This article explores practical recommendations for "AI-proofing" exam questions—methods that ensure the reliability of assessments in an era where AI capabilities are increasingly sophisticated. By adopting these strategies, educators can create exams that challenge students to engage deeply with the material, demonstrate genuine comprehension, and apply knowledge in ways that AI cannot easily replicate.

Example Approaches and Questions

To create final exam questions that are less likely to be easily answered by students using AI tools like ChatGPT, consider incorporating the strategies below.

1. Application-Based Questions:

• Example: "Given a scenario where a hospital's genetic counseling program integrates cultural beliefs, describe how you would approach counseling a patient from a collectivist culture. What specific strategies would you employ, and why?"

• Rationale: These questions require students to apply theoretical knowledge to specific, real-world situations, which involves critical thinking rather than just factual recall.

2. Analysis and Interpretation:

• Example: "Analyze the ethical implications of implementing autonomous vehicles in urban environments. How might this technology conflict with the existing infrastructure and societal norms, and what adjustments could be made to ensure a smooth integration?"

• Rationale: Analysis requires students to engage deeply with the material and interpret it in their own words, making it more difficult for AI to generate direct answers.

3. Case Studies and Scenarios:

• Example: "An urban planner is tasked with redesigning a neighborhood prone to flooding. The local community believes that traditional methods like sandbagging are the best solution. How should the planner address these beliefs while introducing modern flood mitigation techniques?"

• Rationale: Case studies demand a synthesis of knowledge and personalized solutions, which cannot be easily reproduced by AI without understanding the specific course context.

4. Comparison and Contrast:

• Example: Compare and contrast the use of symbolism in 19th-century British literature and 20th-century American literature. How do the cultural contexts of these periods influence the authors' use of symbolic imagery?"

• Rationale: This requires students to draw distinctions and similarities, which necessitates deeper understanding rather than just information retrieval.

5. Reflective Questions:

• Example: "Reflect on your own potential biases as a future healthcare professional. How might these biases affect your interactions with patients from different cultural backgrounds, and what steps can you take to mitigate them?"

• Rationale: Reflective questions prompt personal insights, which AI cannot easily generate, making it more difficult to provide a generic answer.

6. Creative and Original Thought:

• Example: "Propose a new communication strategy for addressing misinformation on social media platforms that has not been covered in class. Justify your proposal with evidence from the course material."

• Rationale: This type of question requires original thinking and creativity, encouraging students to go beyond the information provided in the course.

7. Multi-Part Questions:

• Example: "First, summarize the main ethical dilemmas in medical imaging. Then, apply these dilemmas to a hypothetical situation involving a patient who is hesitant about medical imaging due to religious beliefs. Finally, suggest a plan to address the patient's concerns while maintaining ethical standards."

• Rationale: Multi-part questions require comprehensive responses that build on each other, making it challenging for AI to generate complete and coherent answers.

These strategies encourage students to demonstrate a deep understanding of the material, apply their knowledge in novel ways, and engage in critical thinking—areas where AI tools may struggle to provide comprehensive answers.

These questions and examples were compiled or adapted from AI responses to prompts entered in ChatGPT 4. The prompts were aimed at producing suggestions for AI-proof final exam questions.