Re-imagining Critical Thinking in Higher Education in the Age of Artificial Intelligence (Part 2)
Did you miss Part 1 of this article series? Find it here. This article is part 2 of 4 of the series. Follow me for details of all the parts.
Artificial Intelligence Is Changing the Nature of Learning, Not Just the Tools of Learning
The conversation about artificial intelligence in higher education has largely focused on what students can do with AI. A far more important question, however, is what AI is doing to students’ thinking. This distinction is critical because education is fundamentally concerned with cognitive development rather than task completion. Universities do not exist to help students complete assignments efficiently; they exist to cultivate the intellectual capabilities that enable graduates to solve problems independently, make sound judgments, adapt to uncertainty, and generate new knowledge, especially at postgraduate level throughout their lives. Artificial intelligence therefore demands more than a technological response; it demands a re-examination of how human learning actually occurs.
Learning is often misunderstood as the accumulation of information. Contemporary research in cognitive psychology, neuroscience, and the learning sciences paints a very different picture. Learning is an active biological and cognitive process in which the brain constructs, reorganizes, and strengthens neural pathways through sustained engagement with ideas. Knowledge is not transferred into the mind like files copied onto a computer; it is actively built through questioning, reasoning, reflection, feedback, and repeated application. This process requires mental effort. Indeed, the very struggle associated with learning is not a sign of failure but evidence that meaningful learning is taking place.
One of the most recent concepts explaining this phenomenon is what Robert and Elizabeth Bjork describe as “desirable difficulties.” Their research demonstrates that learning strategies requiring greater cognitive effort often produce stronger long-term understanding than approaches that make learning appear easy (Bjork & Bjork, 2011). Activities such as recalling information from memory, solving unfamiliar problems, explaining concepts to others, and applying knowledge in new contexts create productive challenges that strengthen learning. Students frequently perceive these activities as difficult because they demand sustained mental engagement. Yet neuroscience suggests that such effort promotes learning by reinforcing neural connections associated with memory retrieval and conceptual understanding.
This insight has profound implications for artificial intelligence. AI significantly reduces many of the desirable difficulties through which expertise develops. A student struggling to synthesize competing academic perspectives can now obtain a coherent synthesis within seconds. A learner wrestling with the structure of an argument can request an instantly polished essay. Programming errors can be corrected automatically. Mathematical solutions can be generated without requiring students to identify underlying principles. While these capabilities undoubtedly improve efficiency, they may also limit opportunities for the cognitive effort that transforms novice learners into independent thinkers.
The same argument emerges from another principle in educational psychology: retrieval practice. Brown, Roediger, and McDaniel (2014), drawing upon decades of empirical research, demonstrate that actively retrieving knowledge from memory produces substantially deeper learning than repeatedly reviewing information. The act of recalling concepts, solving problems without immediate assistance, and reconstructing knowledge strengthens long-term retention while improving learners’ ability to transfer understanding to unfamiliar situations. Put simply, students learn more effectively when they are required to think rather than just recognize information.
Artificial intelligence changes this equation. When answers become immediately available, learners may bypass retrieval altogether. Instead of reconstructing knowledge through deliberate reasoning, they increasingly consult AI systems capable of generating persuasive responses almost instantaneously. The immediate result is greater efficiency. The long-term consequence, however, may be weaker conceptual understanding because the cognitive processes responsible for strengthening learning are exercised less frequently.
Closely related is the concept of metacognition, the capacity to think about one’s own thinking. Effective learners continuously monitor what they know, identify gaps in understanding, evaluate the quality of their reasoning, and regulate their learning strategies accordingly. Metacognition enables individuals to distinguish genuine understanding from superficial familiarity. It is widely regarded as one of the strongest predictors of academic success and lifelong learning. Generative AI presents both opportunities and risks for metacognitive development. Used well, AI can support reflection by providing alternative explanations, generating feedback, or exposing learners to competing viewpoints. Used uncritically, however, it can create an illusion of understanding. Students may mistake AI-generated explanations for personal comprehension simply because the output appears coherent and convincing. They may recognize an argument without ever constructing it themselves. This distinction is subtle but fundamental.
Understanding produced by external systems should never be confused with understanding developed through internal cognitive effort.
The work of educational psychologist John Sweller further illuminates this challenge. Sweller’s Cognitive Load Theory argues that human working memory possesses limited capacity and that effective instructional design should reduce unnecessary mental demands while preserving the cognitive effort essential for learning (Sweller, 1988). Importantly, Sweller does not advocate eliminating intellectual challenge altogether. Rather, he distinguishes between cognitive load that interferes with learning and cognitive effort that promotes it. Artificial intelligence has enormous potential to reduce unproductive cognitive burdens. For example, by automating repetitive administrative tasks, summarizing large volumes of information, or providing immediate formative feedback. These applications can free learners to focus on higher-order reasoning. Problems arise, however, when AI eliminates the productive cognitive work that students themselves must perform to develop expertise.
Educational researchers Paul Kirschner and Carl Hendrick have similarly cautioned against misconceptions about learning that prioritise convenience over cognition. Their work consistently demonstrates that expertise develops through deliberate practice, sustained attention, feedback, and repeated engagement with increasingly complex problems rather than through passive exposure to information (Kirschner & Hendrick, 2020). Learning, they argue, cannot be outsourced. Technologies may support it, accelerate aspects of it, or enrich it, but they cannot replace the biological and cognitive processes through which understanding is constructed.
Perhaps the most directly relevant concept for the AI era is cognitive offloading. Risko and Gilbert (2016) define cognitive offloading as the use of physical actions or external technologies to reduce the mental effort required for cognitive tasks. Humans have always offloaded aspects of cognition. We write notes instead of memorizing everything, use calculators instead of performing lengthy arithmetic mentally, and rely on GPS rather than memorizing complex routes. These practices are not inherently problematic. Indeed, they often enhance productivity and allow cognitive resources to be redirected toward more sophisticated tasks. The challenge arises when cognitive offloading replaces rather than complements higher-order thinking. Artificial intelligence differs fundamentally from earlier technologies because it increasingly performs tasks involving analysis, evaluation, synthesis, writing, reasoning, and creative generation, functions that lie at the heart of university education.
The critical question is therefore no longer whether students should use external tools. Rather, it is which aspects of cognition should remain fundamentally human.
Recent analyses by the Organisation for Economic Co-operation and Development (OECD) reinforce this concern. The OECD argues that generative AI should not be viewed primarily as a substitute for human intelligence but as a technology requiring education systems to place greater emphasis on analytical thinking, critical reasoning, ethical judgment, collaboration, and creativity, precisely because these uniquely human competencies become more valuable as AI capabilities expand (OECD, 2024). The future graduate will not compete with artificial intelligence by memorizing more information than machines. Graduates will distinguish themselves through their capacity to evaluate AI outputs, challenge assumptions, navigate ambiguity, integrate diverse perspectives, and make decisions grounded in ethical and contextual understanding.
The implications are profound. Artificial intelligence is not just changing how students learn; it is changing what it means to learn. The challenge for higher education is therefore not preserving traditional teaching methods but ensuring that technological efficiency does not come at the expense of intellectual development. Universities must intentionally design learning experiences that preserve the cognitive effort necessary for critical thinking while harnessing AI’s extraordinary capacity to enhance teaching and research. The objective should not be frictionless learning. It should be meaningful learning.
The New Crisis Nobody Is Talking About – The Outsourcing of Human Thinking
Public debate about artificial intelligence in higher education has been dominated by one issue: plagiarism. Universities have invested heavily in AI detection software, revised academic integrity policies, redesigned assessment regulations, and debated whether students should be permitted to use generative AI in coursework. While these discussions are understandable, they may also be distracting institutions from a much deeper educational crisis.
The greatest threat posed by artificial intelligence is not that students will submit work they did not write. It is that they may gradually stop doing the thinking that universities exist to develop.
This distinction fundamentally changes the nature of the debate. Academic integrity has traditionally focused on authorship. Did the student produce the submitted work? Was the assignment completed honestly? Were appropriate sources acknowledged? These remain essential questions. Yet in the age of artificial intelligence, they are no longer sufficient. A student may fully disclose AI use, comply with institutional guidelines, submit technically original work, and still engage in remarkably little independent thinking. The assignment may satisfy formal requirements while failing to achieve the educational purpose for which it was designed. This possibility exposes a critical weakness in many current university policies. Institutions are concentrating on protecting the originality of academic products while paying comparatively little attention to protecting the originality of intellectual processes. Yet universities exist primarily to cultivate the latter.
Original writing is not the same as original thinking.
An essay can be original because it has never previously existed, while the intellectual work underpinning it has been largely performed by artificial intelligence. Conversely, a student may produce an entirely authentic argument that builds thoughtfully upon existing scholarship, demonstrating deep reasoning, careful judgment, and genuine intellectual growth. The educational value lies not simply in the finished document but in the quality of thinking that produced it. This distinction is perhaps the most important educational question raised by generative AI. Higher education has assessed written assignments as proxies for cognitive development. Essays, dissertations, laboratory reports, policy analyses, and research projects were never valuable simply because they produced text. They were valuable because the process of constructing them required students to analyse evidence, compare competing ideas, organize arguments, reflect critically, solve problems, and communicate reasoning. Writing served as evidence that thinking had occurred.
Artificial intelligence disrupts this relationship. Today, it is increasingly possible to generate sophisticated academic products without engaging fully in the intellectual processes they were designed to measure. The educational risk therefore lies not in the technology itself but in confusing the production of knowledge artifacts with the development of human understanding. This phenomenon might be described as cognitive outsourcing, the progressive transfer of higher-order intellectual responsibilities from human learners to intelligent systems. Unlike traditional cognitive offloading, which largely involved routine memory or computational tasks, cognitive outsourcing reaches into the very capacities that define university education: interpretation, synthesis, evaluation, critical reflection, creativity, and judgment.
The consequences may emerge gradually rather than dramatically. Students may become less comfortable grappling with ambiguity because AI rapidly supplies answers. They may invest less time evaluating competing arguments because AI conveniently synthesizes them. They may lose confidence in developing independent positions because machine-generated responses appear more articulate than their own early drafts. Over time, intellectual perseverance, the willingness to wrestle with difficult ideas may quietly decline. This concern finds support in broader research on expertise development. Experts differ from novices not because they possess more information alone, but because they have developed richer conceptual frameworks, stronger pattern recognition, more refined judgment, and greater capacity to navigate uncertainty. These qualities emerge through years of deliberate practice, reflection, feedback, and sustained engagement with complex problems. They cannot simply be downloaded or generated on demand.
Ironically, artificial intelligence makes independent thinking more valuable, not less.
As AI systems become increasingly capable of generating information, summarizing evidence, drafting reports, and automating routine analysis, human comparative advantage shifts toward those capacities machines cannot reliably replicate. Ethical reasoning. Contextual judgment. Intellectual humility. Curiosity. Moral imagination. The ability to ask questions no algorithm has been instructed to answer. The wisdom to know when evidence is insufficient. The courage to challenge consensus. These are not residual skills left over after technological progress. They are becoming the defining competencies of leadership in an AI-mediated world.
This has profound implications for universities across Africa. The continent’s greatest resource is not artificial intelligence itself but its people, the creativity, resilience, diversity, and intellectual potential of one of the world’s youngest populations. If higher education unintentionally encourages graduates to depend on AI for thinking rather than empowering them to think with AI, Africa risks becoming a consumer of algorithmic intelligence developed elsewhere rather than a producer of contextually grounded knowledge and innovation. The challenge is therefore much larger than regulating AI use in examinations or detecting machine-generated assignments. Universities must redesign educational systems around a more fundamental objective: ensuring that every encounter with artificial intelligence strengthens rather than substitutes the learner’s own reasoning.
The future will not belong to graduates who can just prompt AI more effectively than others. It will belong to those who can critically evaluate its outputs, identify its limitations, recognise its biases, integrate its insights with human experience, and exercise the wisdom to know when the machine is wrong. In that sense, the defining purpose of higher education has not changed. It has become even more important. The central question is no longer whether students are using artificial intelligence. The central question is whether artificial intelligence is still being used in ways that preserve, deepen, and ultimately expand humanity’s capacity to think.
Part 3 will explore why protecting critical thinking matters even more for Africa. Stay tuned and follow the conversation.
Author
Dr. Gilbert A. Ang’ana is the Dean of School of Leadership, Business, and Technology at Pan Africa Christian (PAC) University whose research focus is on values-based and collaborative leadership and governance in Africa. He is a leadership, governance and policy consultant, teacher and coach and the Founder of Accent Leadership Group and Accent Global Initiative (Think Tank)
