Effects of AI on the Human Brain in Education

Introduction

Technology leaders and educators are increasingly debating how artificial intelligence (AI) might be reshaping human cognition, especially in education. For example, former Google executive Mo Gawdat recently warned on The Diary of a CEO podcast that as we lean on AI tools, we risk “atrophying” our own cognitive abilities – essentially a “use it or lose it” effect. He fears that if AI handles more of our thinking tasks, human brains may lose practice in critical skills like problem-solving and focus. These concerns are echoed by psychologists and neuroscientists who note parallels with how overuse of digital tech can weaken attention and memory. At the same time, many experts also see AI’s promise: personalized tutors like Khan Academy’s Khanmigo or GPT-powered assistants could enhance learning, adapt to student needs, and even free teachers to focus on deeper mentorship. This report examines both sides – potential benefits and risks – of AI’s impact on brain function, cognition, and learning. We draw on recent expert commentary and emerging research to explore how AI use in education might affect attention span, memory, critical thinking, and overall cognitive development. A comparison of major viewpoints and study findings is provided in a summary table.

AI, Attention Span, and Focus in the Classroom

One concern is that AI tools – by providing instant answers and constant stimuli – could further erode students’ attention spans. A 2025 survey of academic leaders found 66% believe generative AI will diminish student attention spans, with 24% expecting a major impactelon.edu. This aligns with broader research on digital media: heavy exposure to interactive technology and rapid information flow has been linked to shortened attention and increased distractibility in young peoplefrontiersin.orgfrontiersin.org. For instance, students in one study could only focus for ~6 minutes in the presence of digital distractions like messaging or social mediafrontiersin.org. Neuroscientists note that frequent multitasking with devices correlates with physical brain changes – e.g. reduced gray matter in areas involved in attention regulationfrontiersin.org. The constant novelty and alerts from AI-driven apps can lead to “continuous partial attention,” training the brain to flit between tasks rather than sustain deep focusfrontiersin.orgfrontiersin.org. Over time, this may increase mental fatigue and mimic attention-deficit symptomsfrontiersin.org.

On the other hand, AI might also improve engagement if used thoughtfully. Intelligent tutoring systems can personalize content and pacing to each learner, potentially holding student attention better than one-size-fits-all lecturesfrontiersin.org. In a Harvard experiment, an AI physics tutor actually left students more engaged and motivated than a traditional classnature.com. Educators report that some AI tools can captivate students by adapting to their interests – e.g. relating math problems to a student’s favorite sport or hobby – thereby maintaining focuspressbooks.pubpressbooks.pub. The key appears to be how AI is integrated: passively consuming quick AI answers could encourage mental autopilot, whereas interactive, well-designed AI platforms might keep students actively involved. The prevailing view among experts is that while AI can boost short-term engagement, sustaining true attention requires guided use. Many college leaders emphasize the need to train students in mindful use of AI, so that instant answers do not replace the slow, effortful processes that build concentrationelon.eduelon.edu.

Memory, Knowledge Retention, and Cognitive Offloading

AI tools also raise questions about memory and learning retention. Psychologists have long observed the “Google effect” – the tendency to forget information that is easily looked up onlinethedecisionlab.com. Relying on AI as an external brain could extend this effect. Offloading cognitive tasks to digital assistants means we may encode and store less information internally, affecting long-term memory formation. Researchers use terms like “digital amnesia” or “digital dementia” to describe memory decline from excessive tech usefrontiersin.org. For example, routinely using GPS for navigation has been linked to declines in hippocampal-dependent spatial memory, since our brains get less practice forming cognitive mapsfrontiersin.org. Similarly, studies found that people who frequently offload tasks to smartphones (like saving phone numbers or reminders) show weaker recall ability over timefrontiersin.org. As one review put it, “when individuals rely on digital devices to store information, they are less likely to remember it,” leading to measurably lower memory performancefrontiersin.org.

AI chatbots and tutors could amplify this cognitive offloading. A 2024 MIT Media Lab study (the first to scan brains of ChatGPT users) found that participants who wrote essays with AI assistance retained significantly less of what they wrote and showed reduced brain activation in memory-related regionstime.com. EEG recordings revealed that the AI-assisted group had markedly lower engagement of deep memory processes (reflected in weaker alpha/theta brain waves) and later “remembered little of their own essays,” in contrast to those who wrote without AItime.com. In effect, the AI users outsourced so much thinking to ChatGPT that they failed to integrate the knowledge or form memories of it. This neural evidence confirms what earlier behavioral studies suggested: cognitive offloading can be a double-edged sword. It frees up mental resources in the moment, but at the risk of shallower processing and weaker retentionmdpi.commdpi.com. Even educators see this anecdotally – students who rely on AI to summarize readings or solve problems may finish tasks faster, but often struggle to recall or apply the material later.

To counteract these effects, experts recommend approaches that keep the learner mentally active. For instance, using AI in a “prompt-and-probe” manner – where the student must explain, verify, or build on the AI’s output – could ensure the brain still encodes knowledge. The MIT study offered a hopeful sign: participants who first wrote essays on their own and then used AI showed increased brain connectivity and strong memory engagement when integrating the AI’s inputtime.com. This suggests that AI can enhance learning if used to supplement (rather than replace) active thinkingtime.com. The challenge for education is finding that balance: leveraging AI’s vast information access and memory storage without causing our own memory muscles to atrophy. Neuroscientists like Dr. Nataliya Kosmyna (lead author of the MIT study) urge caution especially for young students, warning that “developing brains are at the highest risk” if we introduce AI too early as a crutchtime.com. Simply put, if an AI always remembers and tells the answer, a child’s brain gets far less practice encoding facts or retrieving knowledge from memory – crucial skills for cognitive development.

Critical Thinking and Problem-Solving: Improved or Impaired?

Perhaps the biggest debate is whether AI tools are improving students’ critical thinking or subtly undermining it. On one side, AI proponents argue that intelligent tutors and assistants can foster higher-order thinking by personalizing learning. Adaptive AI systems can pose questions at just the right difficulty, give hints rather than answers, and encourage students to explain their reasoning. Such scaffolding could strengthen critical thinking by guiding students through problems while still requiring their input. For example, Khan Academy’s GPT-4 based tutor Khanmigo is designed to coach students through solutions step-by-step instead of outright giving the answer. Teachers report that when it works, the AI asks probing questions and prompts reflection, which can deepen students’ understanding and analytical skillspressbooks.pubpressbooks.pub. There is evidence that well-implemented AI tutors can emulate many benefits of one-on-one human tutoring – a context known to yield significant gains in critical thinking and problem-solving abilitynature.comnature.com. A recent randomized trial demonstrated that a carefully designed AI tutor (built with pedagogical best practices) led college students to learn more material in less time without sacrificing comprehension, suggesting students were in fact engaging meaningfully with the contentnature.com. Moreover, 75% of academic leaders in 2025 predicted that GenAI tools will improve students’ research and analytical skills by helping them gather and evaluate information more efficientlyelon.edu. In principle, AI could handle rote aspects of assignments and free students to focus on big-picture critical thinking (e.g. interpreting results, crafting arguments) – essentially serving as a cognitive partner that amplifies human reasoning.

However, a growing body of research and real-world observations raises alarms that over-reliance on AI may erode critical thinking instead of enhancing it. The concern is that if students become accustomed to AI “doing the thinking” for them, they may skip the mental struggle that builds reasoning skills. In education and cognitive science circles, this is sometimes termed “AI-induced cognitive atrophy (AICICA),” referring to a decline in core cognitive abilities from excessive dependence on AIpmc.ncbi.nlm.nih.gov. Dergaa et al. (2024) articulate this as the “use it or lose it” principle applied to the mind: if a student always opts for the convenient AI-provided solution over wrestling with a problem, their capacity for independent analysis, creativity, and judgment could weaken over timepmc.ncbi.nlm.nih.govpmc.ncbi.nlm.nih.gov. Notably, the conversational, on-demand nature of modern AI makes it a more immersive crutch than tools of the past. While using a calculator or Google requires specific queries and returns static results, engaging a chatbot tutor is more interactive and persuasive – it can lead students to trust its answers without verification, further reducing their habit of critical evaluationslejournal.springeropen.comslejournal.springeropen.com.

Multiple studies in 2023–2025 lend empirical weight to these worries. A systematic review by Zhai, Wibowo & Li (2024) found that students’ over-reliance on AI dialogue systems (like generative chatbot assistants) consistently correlated with poorer critical thinking and decision-making outcomesslejournal.springeropen.comslejournal.springeropen.com. In the reviewed studies, when students uncritically accepted AI-generated content (often due to AI’s authoritative tone or their own difficulty judging its reliability), they showed declines in analytical reasoning and a tendency to favor quick heuristic answersslejournal.springeropen.com. Similarly, a large-scale 2024 study in the Journal of Intelligence tested people’s performance on critical thinking tasks with and without AI support. It reported that heavy users of AI tools scored significantly lower on independent critical thinking assessments than light users, even after controlling for education levelmdpi.commdpi.com. The authors attributed this to cognitive offloading: those who habitually deferred to AI had fewer opportunities to practice evaluating information and forming arguments on their ownmdpi.commdpi.com. Strikingly, the negative effect appears most pronounced for open-ended, creative thinking tasks. One finding was that users with AI assistance produced a “less diverse set of outcomes” for a given problem, whereas those without AI came up with more varied ideasgizmodo.com. The uniformity of AI-guided answers suggests a kind of convergence in thinking, potentially stunting creativity and divergent problem-solvinggizmodo.com.

Real-world experiments by industry have also detected this critical-thinking deficit. Researchers at Microsoft and Carnegie Mellon University (2025) observed that knowledge workers who grew confident in an AI’s capabilities would often “take their hands off the wheel,” engaging less in oversight or independent thoughtslashdot.org. When these workers trusted the AI to handle a task, their involvement became perfunctory – they experienced a “perceived enaction of critical thinking” (feeling like the job was done when the AI answered) without actually scrutinizing the solutiongizmodo.com. The study warned that such patterns, especially for low-stakes or routine tasks, could lead to “long-term reliance and diminished independent problem-solving” skillsslashdot.org. Tellingly, when participants did not trust the AI, they stayed mentally sharper and showed more confidence in their own ability to evaluate and improve the workgizmodo.com. This implies that maintaining a healthy skepticism and active role when using AI is critical to preserving one’s analytical faculties.

Classroom trials show how fine the line is between AI as a tutor versus a cheat-sheet. In an early pilot of Khanmigo in New Jersey schools, educators noted that the AI sometimes “helped too much,” simply giving students the answer to a math problem instead of guiding their thinkingpressbooks.pub. When the bot provided direct answers, students skipped the reasoning process, prompting officials to intervene – they want AI to ask leading questions or give hints, not outright solutions, precisely to protect the development of critical thinkingpressbooks.pubpressbooks.pub. Khan Academy tweaked the system after discovering this issue, yet in testing it still occasionally solved a problem for the student, illustrating the ongoing challengepressbooks.pubpressbooks.pub. These anecdotes underscore a core point made by psychologists: convenience can be cognitively costly. If an AI tutor is not carefully constrained, students may take the path of least resistance (why struggle with a problem if the AI will just answer it?), and in doing so, miss out on the very struggle that trains the brain. By contrast, when AI is used to complement human thought – for example, generating ideas that students must then critique, or showing solutions that students must explain – it can actually stimulate critical thinking. The MIT “Brain on ChatGPT” experiment ended on a hopeful note in this regard: the group that had to rewrite their essays without AI after initially relying on ChatGPT performed poorly (unable to recall or reason deeply about the content), but the group that started without AI and later incorporated ChatGPT showed improved creative thinking with no loss of critical engagementtime.comtime.com. In short, sequence and pedagogy matter – the best outcomes may come from using AI to enrich learning after students have attempted to think for themselves.

Cognitive Development, Learning Outcomes, and the Brain

From a developmental perspective, AI’s influence on the growing brain is a major concern. Children and adolescents have high neural plasticity, meaning their cognitive skills are shaped by how they are exercised. If AI changes what and how young people learn, it could literally shape their brain development. Educators and neuroscientists worry that overexposure to AI guidance might stunt the formation of independent learning capabilities. As one child psychiatrist cautions, heavy reliance on AI for schoolwork in formative years can weaken “the neural connections that help you access information, remember facts, and be resilient” – in essence, training the brain to be dependenttime.com. This could manifest as future learners who struggle when an AI is not present to prompt them. Some early evidence hints at this: a 2023 study noted that kids who frequently used AI for homework had more trouble solving new problems on their own, suggesting a lack of adaptive problem-solving strategies (though more research is needed for firm conclusions).

Conversely, AI also offers opportunities to enrich cognitive development. Well-designed AI educational tools can expose students to experiences that might be hard to replicate in a traditional classroom. For example, adaptive learning platforms use AI to customize lessons to a student’s level, which can prevent boredom for advanced learners and frustration for those who need reinforcementfrontiersin.orgfrontiersin.org. By keeping the difficulty “just right,” AI tutors aim to keep students in a zone of proximal development – a state known to maximize cognitive growth. Immediate feedback from AI can also correct misunderstandings in real time, preventing small knowledge gaps from wideningfrontiersin.org. There is evidence that such tailored support boosts mastery and retention of material. In one controlled trial, college students using an AI tutor showed significantly higher learning gains on assessments compared to peers in an active-learning course, with the AI group covering more content in less timenature.com. They also reported feeling more motivated, not less, by the AI-based learning experiencenature.com. These results suggest that, under the right conditions, AI can accelerate learning and perhaps even improve certain cognitive skills (like problem-solving speed or pattern recognition) by providing intensive practice and differentiated instruction. Indeed, a majority (91%) of academic leaders expect GenAI tools to “enhance and customize learning,” and many believe AI will increase student creativity and improve writing skills through personalized guidanceelon.eduelon.edu.

That said, the quality of AI-mediated learning is crucial. If the AI simply spoon-feeds answers (as in the problematic Khanmigo cases), students may score well in the short term but fail to develop robust understanding or transferable skills. This reflects the paradox educators see: AI can raise test performance while lowering the very cognitive effort that makes learning durable. For instance, an analysis in Frontiers in Cognition pointed out that children need to struggle and learn from mistakes to build strong problem-solving abilities, and AI can inadvertently short-circuit this process if it always jumps in with the solutionfrontiersin.org. The same review noted that interactive AI toys did enhance certain cognitive skills in young children (like vocabulary and pattern recognition), but children who became overly reliant on the toy for answers showed less improvement in independent exploration and perseverancefrontiersin.orgfrontiersin.org. In other words, guided discovery through AI can be beneficial, while over-guidance can hinder the development of grit and curiosity.

From a neurological standpoint, researchers are just beginning to study how AI-driven learning impacts the brain. Early neuroimaging work (e.g. the MIT EEG study) indicates that how we use AI dramatically changes brain activity. Solving a problem without AI engages a wide network of brain regions (frontal lobes for executive function, associative areas for memory, etc.), whereas doing the same with an AI assistant shows significantly reduced activity in those networkstime.comtime.com. Over time, consistently lower activation could imply less strengthening of those neural circuits. Some experts compare this to the way GPS navigation usage correlates with reduced activation (and even structural changes) in brain areas responsible for spatial orientationfrontiersin.orgfrontiersin.org – essentially, the brain “prunes” skills it doesn’t use. However, it’s not all or nothing: brain plasticity means that if AI is used in a way that supplements human thought (for example, the hybrid approach in the MIT study’s second phase), it might stimulate new kinds of cognitive activity or make learning more efficient without net losstime.comtime.com. Research in this area is nascent, but ongoing studies are examining whether alternating between analog thinking and AI assistance could produce the best outcomes for brain development.

Finally, psychological impacts deserve attention as part of the brain’s overall well-being. Some instructors note that when students outsource work to AI, they often exhibit lower confidence and motivation – they haven’t wrestled with the material, so they feel less accomplishment. A May 2025 Harvard study found that while AI helpers made people more productive in writing tasks, they felt less motivated and reported lower enjoyment in the processtime.com. There is also a social-cognitive angle: if students engage more with AI tutors and less with peers or teachers, could that affect communication skills or empathy? Early indications (from studies of social robot use) suggest heavy AI interaction can subtly alter social cognition – for instance, some users become frustrated when human teachers don’t provide the instant feedback an AI does, which can impact patience and interpersonal skills. These areas warrant further research, but they underscore that AI’s impact on the “learning brain” is multifaceted – encompassing attention, memory, reasoning, motivation, and emotional development.

Benefits vs. Concerns: Key Viewpoints and Findings (Table)

To summarize the landscape of expert perspectives and study results, the table below compares major benefits attributed to AI in education with the corresponding concerns or findings that temper those benefits. It highlights how reputable sources – from peer-reviewed studies to surveys of educators – converge or diverge on AI’s cognitive impact.

Source / Study (Year)	Key Findings / Viewpoint
Mo Gawdat (2023) – Tech Executive (Diary of a CEO interview)	Warns that over-reliance on AI could atrophy human cognition. Suggests that as AI handles more tasks, “our own abilities deteriorate” from disusegizmodo.com. Advocates maintaining human critical-thinking practice alongside AI.
MIT Media Lab EEG Study (Kosmyna et al., 2024) – “Your Brain on ChatGPT”	Using ChatGPT for writing caused a 47% drop in neural engagement (vs. writing without AI). AI-assisted writers “consistently underperformed at neural, linguistic, and behavioral levels,” showing weaker memory and less original thoughttime.comtime.com. Indicates AI use can impair deep cognitive processing, especially in younger brainstime.com.
Harvard RCT on AI Tutor (Kestin et al., 2025) – Scientific Reports	College students with a custom AI tutor learned significantly more in less time than those in an active-learning classnature.com. They reported higher engagement and motivation. This provides empirical evidence that an AI tutor (when designed with proper pedagogy) can enhance learning outcomes and even attention.
Dergaa et al. (2024) – Frontiers in Psychology (Opinion)	Introduces “AI-Chatbot Induced Cognitive Atrophy (AICICA)” – the hypothesis that excessive chatbot use leads to decline in critical thinking, analysis, and creativitypmc.ncbi.nlm.nih.gov. Especially in education, young users may favor convenient AI answers over deep comprehension, potentially hindering development of cognitive facultiespmc.ncbi.nlm.nih.govpmc.ncbi.nlm.nih.gov.
Zhai et al. (2024) – Systematic Review in Smart Learning Environments	Reviewed studies on student over-reliance on AI dialogue systems. Found that when students unquestioningly accept AI outputs, it often impairs decision-making and reasoningslejournal.springeropen.comslejournal.springeropen.com. Students favored fast AI solutions (heuristics) over slow, reflective thinking, highlighting a tendency to use AI as a cognitive shortcut at the expense of skill-building.
Microsoft & CMU Study (2025) – Knowledge Workers & AI (via Gizmodo)	Surveyed 319 professionals; found greater trust in AI led to less critical oversight. Those confident in AI frequently “let go” of tasks, exhibiting a false sense of doing critical thinking while actually relying on the AIgizmodo.com. AI-assisted work also yielded less diverse solutions, which researchers interpreted as a sign of reduced creative critical thinking in usersgizmodo.com.
Khanmigo Classroom Pilot (2023) – Khan Academy & Newark Schools	Demonstrated AI’s double-edged nature. Benefit: Khanmigo provided personalized support, pinpointing student misconceptions and sparking engagement in many cases. Concern: It sometimes simply gave answers, which hindered students’ critical thinking development according to teacherspressbooks.pubpressbooks.pub. Khan Academy adjusted the tool to focus more on guiding rather than solving.
Elon Univ. Survey of Educators (2025) – 337 Higher-Ed Leaders	Optimistic on benefits: 91% say AI will enhance & personalize learning; 75% say it will improve student research skills; 66% foresee increased creativityelon.eduelon.edu. Simultaneously worried: 92% think students will become over-dependent on AI, and 66% expect diminished attention spans due to AI useelon.edu. 95% also anticipate challenges with academic integrityelon.edu.
Frontiers Review (Shanmugasundaram et al., 2023) – Digital Tech & Cognition	Notes that AI-powered educational tools can enhance cognitive development by personalizing learning and providing instant feedbackfrontiersin.orgfrontiersin.org. However, cautions that children’s over-reliance on AI to solve problems may short-circuit the learning of problem-solving skills and reduce opportunities to build independent thinkingfrontiersin.org. Recommends balanced tech use.

Table: Major viewpoints on AI’s impact on cognition in education – a comparison of key benefits and concerns (with sources).

Conclusion

AI’s expanding role in education presents a nuanced picture for the human brain. On one hand, AI tutors and personalized learning systems hold immense potential to enrich education – improving engagement, tailoring instruction to individual needs, and even boosting learning efficiency and outcomes. Cutting-edge studies show that when thoughtfully implemented, AI can act as a cognitive amplifier, helping students reach insights and mastery that might be hard to achieve at scale with traditional methodsnature.com. On the other hand, mounting evidence and expert commentary urge caution: the very convenience and power of AI may invite an over-reliance that undermines crucial cognitive skills – from attention and memory to critical thinking and creativitypmc.ncbi.nlm.nih.govtime.com. The human brain thrives on challenge and active processing; if AI too readily removes those elements, we risk creating learners who know more answers but understand less deeply. As Mo Gawdat and others have warned, we must be mindful not to let our natural mental “muscles” atrophy in the age of intelligent machines.

The emerging consensus is that the impact of AI on the brain depends largely on how we use it. Used passively or as a shortcut, AI can encourage mental laziness and shallow learning. Used actively and strategically, AI can augment human cognition, providing rich feedback and freeing up humans to focus on higher-order thinking. Educators, psychologists, and neuroscientists alike recommend a balanced, guided approach to AI in learningfrontiersin.orgelon.edu. This means designing AI tools that promote engagement (not just answers), teaching students digital literacy and metacognitive skills (so they remain critical consumers of AI output), and continuing to value “unplugged” learning activities that exercise the brain in traditional ways. In practical terms, it might involve class assignments where students must first attempt problems without AI before consulting a chatbot, or using AI’s suggestions as a starting point for discussion rather than the final word. Ongoing research – including brain scans, classroom trials, and longitudinal studies – will be vital to understand the long-term neurological effects.

In sum, AI in education is neither a panacea nor a doom – it is a powerful tool that can either sharpen the mind or make it complacent, depending on context. The human brain has remarkable adaptability; with wise integration of AI, we can hopefully leverage that adaptability to expand cognitive potential rather than diminish it. As one review emphasized, “education is not about the dazzle of modern tools but about fostering human connections and thinking”pressbooks.pub. Keeping that principle at the forefront will help ensure that artificial intelligence serves as a partner to human intelligence – one that challenges us, supports us, and ultimately helps us grow smarter and more capable, without eroding the very abilities that define our humanity.

Sources: The information in this report is drawn from a range of expert analyses and studies, including peer-reviewed research in Frontiers in Psychology, Scientific Reports, and Frontiers in Cognition; a 2024 MIT Media Lab study on AI and brain activitytime.comtime.com; surveys of educators and tech leaderselon.edugizmodo.com; and commentary from cognitive scientists and educational psychologists. These sources are cited in-text with brackets (e.g.,time.com) corresponding to the full references. All point toward a critical conclusion: the impact of AI on the student brain will be what we make of it – and now is the time to shape that impact responsibly.

Sources