Check out our podcast on this topic, “What’s Next”, on the CANeLearn YouTube Channel
AI in Education: What to expect next…
A number of years ago, I was having a heated exchange with the director of another organization. At the time, I was CEO of a non-profit educational organization that was offering online courses for students at the high school level. What I knew distinguished our online courses was that we employed a truly socially constructivist methodology where students were central to the learning, where collaboration and authentic learning were intrinsic to the process. It was hard work for the teachers who had to put it all together; for the students who were new to the approach, it required more effort than they had ever been expected to put in. This other director was highlighting a survey of high school students who were asked if they preferred lectures to project-based learning, and, not surprisingly, most students preferred lectures. Having once upon a time been a teenager myself, their responses made sense. More work and more effort, or less work and less effort? Duh. I told this to the other director, who scoffed at it. “Just stick to what works,” he said as he walked away from me, totally convinced he had made his point.
What research says: Freeman et al. (2014), in a large meta-analysis of 225 studies in undergraduate STEM education, found that active learning increased examination performance and reduced failure rates compared with traditional lecturing. Chen and Yang’s (2019) meta-analysis of project-based learning also found a medium-to-large positive effect on student achievement when compared with traditional instruction.
As I reflect on that exchange, I have to say I agree with him. Most people will opt for the easy way out most of the time, and there have been times when I did the same thing. Maybe it is human nature; I won’t comment on that, but there is something I know with absolute certainty, and that is learning requires effort. And this brings me in a very roundabout fashion to the point of this blog. AI is about to make things even easier for us. Isn’t that wonderful? Less work. Less effort. AI agents are entering the workforce, and while there are numerous bugs to be worked out, there is no doubt they are here to stay. Recently, Gemini announced that soon we will all have our own agent that will do internet searches for us, gather all of the necessary data, synthesize it, and collate it into a cozy output. Previously, Google searches using programmer-developed algorithms that incorporated both personal and corporate biases were the bane of educators, as they made it too easy for students to plagiarize data; now, educators have to consider that not only will AI search, but it will also do all the work for the learner. Gemini, the developers are proud to say, will get up and do things. Needless to say, it will do it the way it has been designed to do, but ultimately, the user will barely notice.
What research says: Research on “desirable difficulties” supports the claim that meaningful learning often requires productive effort rather than ease. Bjork and Bjork (2020) argue that some learning conditions feel harder in the moment but improve long-term retention and transfer. This matters in an AI context because Lee et al. (2025) found that higher confidence in generative AI was associated with less critical thinking effort among knowledge workers using GenAI tools.
Are any red lights going off yet? No? Think about this. We will all get the same answers, assume the same premises as fact, and conclude what we have been told to conclude. Will this lead us to groupthink? Maybe. I don’t know, but it scares me. Actually, it terrifies me because this is not the future; this is now.
What research says: Alon-Barkat and Busuioc (2023) examined human-AI decision-making and warned that algorithmic advice can introduce new biases in the interaction between people and automated systems. Lee et al. (2025) also found that GenAI shifts critical thinking toward verification, integration, and stewardship, which means students need to be taught how to interrogate AI outputs rather than simply receive them.
There is good news, however, on the horizon. Our educational leaders have recognized the urgency and the danger, and they are mobilizing as they never have before to ensure that our children acquire the necessary skills needed to thrive in an AI-driven world. Core to these skills is the ability to think critically, to drill down to undeniable core facts, to prompt an LLM significantly to generate alternate perspectives, to identify the “provocations” in the AI-generated text, and to move to one’s own conclusions. This whole new push will see students graduate as independent thinkers who rely on their intellectual skills to guide them through a jungle filled with AI agents ready to get up and do things.
What research says: Abrami et al. (2015), in a meta-analysis published in Review of Educational Research, found that critical thinking can be taught, but that effective instruction requires deliberate strategies such as dialogue, authentic or situated problems, and mentoring. Halpern (1998) also argues that critical thinking must be taught in ways that support transfer across domains, including metacognitive monitoring and deliberate practice.
Of course, we all know nothing is being done except in isolated schools. We never really taught critical thinking in our schools, and we are not in any position to start now.
What research says: Abrami et al. (2015) found that critical-thinking gains were associated with specific instructional conditions, not vague aspirations. In the AI era, this matters because students who are not taught to question sources, evaluate claims, test assumptions, and examine their own reasoning will be poorly equipped to use AI as a tool instead of just as an intellectual replacement.
For a fun take on this post’s topic, check out New Rule: Grad GPT (Bill Maher on YouTube)
Bibliography
Abrami, P. C., Bernard, R. M., Borokhovski, E., Waddington, D. I., Wade, C. A., & Persson, T. (2015). Strategies for teaching students to think critically: A meta-analysis. Review of Educational Research, 85(2), 275–314. https://doi.org/10.3102/0034654314551063
Alon-Barkat, S., & Busuioc, M. (2023). Human–AI interactions in public sector decision making: “Automation bias” and “selective adherence” to algorithmic advice. Journal of Public Administration Research and Theory, 33(1), 153–169. https://doi.org/10.1093/jopart/muac007
Bjork, R. A., & Bjork, E. L. (2020). Desirable difficulties in theory and practice. Journal of Applied Research in Memory and Cognition, 9(4), 475–479. https://doi.org/10.1016/j.jarmac.2020.09.003
Blumenfeld, P. C., Soloway, E., Marx, R. W., Krajcik, J. S., Guzdial, M., & Palincsar, A. (1991). Motivating project-based learning: Sustaining the doing, supporting the learning. Educational Psychologist, 26(3–4), 369–398. https://doi.org/10.1080/00461520.1991.9653139
Chen, C.-H., & Yang, Y.-C. (2019). Revisiting the effects of project-based learning on students’ academic achievement: A meta-analysis investigating moderators. Educational Research Review, 26, 71–81. https://doi.org/10.1016/j.edurev.2018.11.001
Freeman, S., Eddy, S. L., McDonough, M., Smith, M. K., Okoroafor, N., Jordt, H., & Wenderoth, M. P. (2014). Active learning increases student performance in science, engineering, and mathematics. Proceedings of the National Academy of Sciences, 111(23), 8410–8415. https://doi.org/10.1073/pnas.1319030111
Halpern, D. F. (1998). Teaching critical thinking for transfer across domains: Dispositions, skills, structure training, and metacognitive monitoring. American Psychologist, 53(4), 449–455. https://doi.org/10.1037/0003-066X.53.4.449
Kapur, M. (2008). Productive failure. Cognition and Instruction, 26(3), 379–424. https://doi.org/10.1080/07370000802212669
Lee, H.-P. H., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025). The impact of generative AI on critical thinking: Self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (CHI ’25), Article 1121, 1–22. Association for Computing Machinery. https://doi.org/10.1145/3706598.3713778
Source note on Gemini: The description of Gemini’s Deep Research Agent is supported by Google’s developer documentation, which states that the agent can autonomously plan, execute, and synthesize multi-step research tasks. This is not included in the scholarly bibliography because it is an official product source rather than a scholarly article.