Imagine that you spent years developing a medication. You evaluated it through multiple clinical trials and found promising results. In your very last trial, the medication failed to deliver its expected benefits. What would you do? You would go back to the beginning, review each step, determine what might have gone wrong, and explore ways to avoid such results in the future. You would apply your critical thinking skills as you examine errors and oversights. You would exercise your judgment to determine all possible underlying reasons for such an outcome.
The results of two major clinical trials conducted by Novartis and Novo Nordisk should remind us that scientific predictions made by humans can fail. The pharmaceutical companies developed two medications aimed at reducing cardiovascular events. Both medications produced promising biomarker results and passed phase 1 and phase 2 trials. Unfortunately, the expected reduction in cardiovascular events did not occur in phase 3. The failure of both medications was announced recently at the annual meeting of the American Heart Association in November 2026. So, why should we trust AI-generated conclusions if human judgment and critical thinking may fail in highly controlled experimental environments?
What AI Can and Can't Do
Artificial Intelligence is not, by definition, an educational or research tool. AI is human-designed, and AI platforms are advanced computer systems designed to perform tasks that require human intelligence. Over time, some AI systems can learn a task and improve their performance from experience and data they are fed. Amazingly, they can personalize our learning and teaching experiences. They can simplify learning materials, convert content into questions and answers, and even create summaries and mock exams. The important point that deserves our attention is that AI platforms do not personalize experiences unless we provide our information. Providing information is a condition we must uphold. In other words, human intelligence remains indispensable.
In an academic setting, this information may include course materials, literature reviews, or data collected for research studies. Such platforms may accelerate learning and teaching processes; however, they cannot fully replace human judgment and critical thinking.
Critical Thinking and the 5A Framework
For students pursuing degrees in health fields, critical thinking is not simply a skill limited to academic performance. It is a skill that continues to develop into a lifelong responsibility that influences healthcare services. Critical thinking is needed to evaluate evidence, identify accurate resources, further appraise those resources, make informed clinical decisions, solve complex clinical problems, identify gaps and limitations, and conduct research to generate clinical facts.
When these are performed professionally and applied critically, we can provide high-quality healthcare services. This is referred to as the Quality Enhancement Plan (QEP) framework, a framework that we need to follow when making decisions. QEP is a 5-trait framework that emphasizes critical thinking through Asking, Acquiring, Appraising, Applying, and Assessing.
Can AI help us follow the QEP? Yes, it can. We value what AI can do for us when making clinical decisions across different health fields. Think about the last time you used an AI platform to generate a response to a prompt.
- The prompt ”you” provided served as your Ask.
- The machine scanned various resources to generate a response, and this step served as your Acquire.
- Either you asked the machine to appraise, or “you” completed the appraisal yourself. If the machine completed the appraisal, did “you” verify whether the resources were reliable and relevant, or whether they even existed?
- Moving to Apply, the machine suggested a way to translate the findings into practice. Did “you” critically evaluate whether the suggestion was relevant and applicable?
- Lastly, the AI-generated response included multiple weaknesses and gaps. Did “you” verify and reflect on how AI-identified weaknesses influenced the findings?
By breaking the process down into its components, we can see how AI and humans interact to complete the process. AI can simulate some aspects of critical thinking; however, human intelligence remains essential. It is essential for innovative tasks and scenarios where machines fail to replicate human reasoning. Thus, your critical judgment is crucial when you Ask, Appraise, Apply, and Assess.
Although AI helps speed up the decision-making process, its contribution to the QEP framework is ideally limited to delivering resources and suggesting implications you cannot rely on solely. AI is our accelerator, but it is never the decision-maker. It is our responsibility not to let machines guide our decisions.
How AI Fails Without Human Review
I appreciate how Dr. David Stuckler1, a professor of policy analysis and public management at Bocconi University, emphasizes that AI platforms can be a primary source of errors in research and evidence synthesis. He outlines different failure modes when AI is used without critical evaluation. Confidence without clarity is a common failure mode. AI-generated responses often sound convincing and exciting, even when they are not entirely accurate. Such responses create a sense of confidence without a clear understanding of a topic.
AI platforms praise our ideas and suggest further actions that sound interesting and accurate. If we do not check whether those suggestions are relevant and accurate, we fall into a failure mode of going down a rabbit hole of inaccurate conclusions. Since AI systems are not research systems, the foundational knowledge needed to synthesize evidence is not built-in. Those machines aim to satisfy prompts without a firm system trained on how scientific reasoning should happen. Dr. Stuckler states, “AI platforms may turn small mistakes into deep structural failure, where logic breaks down1.”
When to Allow AI in the Classroom
As an educator, I always ask myself: Am I using Artificial Intelligence (AI) wisely? Why do I allow AI usage in a course while prohibiting it in another? Where is AI taking academia and evidence synthesis?
Different skills students develop, acquire, and strengthen in different courses continue to guide my “Allow or Prohibit” decisions. When courses are foundational or require critical thinking and quality writing, I am more hesitant to allow AI. Today, calculators have made our mathematical calculations easier and faster. On the other hand, they have weakened our ability to recall mathematical tables in the blink of an eye. Moreover, fewer of us understand the underlying mathematical processes that calculators perform.
AI is a much more sophisticated system than a calculator. AI not only affects our mental effort to recall multiplication facts, but also reduces our cognitive engagement when we perform a task2. Cognition and intelligence work together to help us make our judgments; both shape the complex knowledge decision-making process2.
We cannot overlook the fascinating roles that AI plays in education and healthcare. But let us consider a collaborative approach between humans and AI that does not compromise our intelligence or critical thinking skills3. AI machines excel at computational tasks. They analyze more than 300 billion words from numerous resources in a split second4. Those machines gather information, extract data, process data, and summarize evidence.
Creative thinking, practical knowledge, intuitions, critical evaluation, and moral reasoning are fundamental human capacities. We all- learners, educators, and healthcare providers- are responsible for controlling the evolving dynamics of human-machine interactions. We are responsible for continuing to identify tasks that should be uniquely human and those that can be completed by AI machines. After all, making context-sensitive judgment remains the ultimate responsibility of healthcare providers3.
Resources:
- Prof. David Stuckler. Do NOT Use AI for Research Until You Watch This. Youtube. 2026. February. Available from: https://youtu.be/N6dv_YfK9zc?si=WBLBvWE6-YKHBnn3
- Siemens G, Marmolejo-Ramos F, Gabriel F, Medeiros K, Marrone R, Joksimovic S, et al. Human and artificial cognition. Comput Educ Artif Intell. 2022;3:100107. Doi: 10.1016/j.caeai.2022.100107
- Ifeanyichukwu A, Vaswani V. Human and artificial intelligence: a review of competencies, collaboration, and ethical implications. Palgo J Educ Res. 2025;10(1):1-8. Doi:10.5281/zenodo.15292956
- Lim J. Generative artificial intelligence: what it is, what it is not and what it can be for the United Nations. United Nations; 2023. Retrieved from https://www.un.org/en/un-chronicle/generative-artificial-intelligence-what-it-what-it-not-and-what-it-can-be-united
