The learning divide: A visual comparison highlights the potential pitfalls of relying on AI for “easy answers” versus the proven benefits of diligent study and engagement, as a new study suggests. Image (and typos) generated by Nano Banana.
Source
The Register
Summary
A new study published in PNAS Nexus finds that people who rely on ChatGPT or similar AI tools for research develop shallower understanding compared with those who gather information manually. Conducted by researchers from the University of Pennsylvania’s Wharton School and New Mexico State University, the study involved over 10,000 participants. Those using AI-generated summaries retained fewer facts, demonstrated less engagement, and produced advice that was shorter, less original, and less trustworthy. The findings reinforce concerns that overreliance on AI can “deskill” learners by replacing active effort with passive consumption. The researchers conclude that AI should support—not replace—critical thinking and independent study.
Key Points
Study of 10,000 participants compared AI-assisted and traditional research.
AI users showed shallower understanding and less factual recall.
AI summaries led to homogenised, less trustworthy responses.
Overreliance on AI risks reducing active learning and cognitive engagement.
Researchers recommend using AI as a support tool, not a substitute.
by Tadhg Blommerde – Assistant Professor, Northumbria University
Estimated reading time: 5 minutes
In an era dominated by AI, educators are finding innovative ways to guide students. This image, inspired by a “Dr. Strange-Syllabus,” represents a pedagogical approach focused on fostering self-reliance and critical thinking, helping students to navigate the complexities of AI and ultimately trust their own capabilities. Image (and typos) generated by Nano Banana.
There is a scene I have witnessed many times in my classroom over the last couple of years. A question is posed, and before the silence has a chance to settle and spark a thought, a hand shoots up. The student confidently provides an answer, not from their own reasoning, but read directly from a glowing phone or laptop screen. Sometimes the answer is wrong and other times it is plausible but subtly wrong, lacking the specific context of our course materials. Almost always the reasoning behind the answer cannot be satisfactorily explained. This is the modern classroom reality. Students arrive with generative AI already deeply embedded in their personal lives and academic processes, viewing it not as a tool, but as a magic machine, an infallible oracle. Their initial relationship with it is one of unquestioning trust.
The Illusion of the All-Knowing Machine
Attempting to ban this technology would be a futile gesture. Instead, the purpose of my teaching became to deliberately make students more critical and reflective users of it. At the start of my module, their overreliance is palpable. They view AI as an all-knowing friend, a collaborator that can replace the hard work of thinking and writing. In the early weeks, this manifests as a flurry of incorrect answers shouted out in class, the product of poorly constructed prompts fed into (exclusively) ChatGPT, and a complete faith in the response it generated. It was clear there was a dual deficit: a lack of foundational knowledge on the topic, and a complete absence of critical engagement with the AI’s output.
Remedying this begins not with a single ‘aha!’ moment, but through a cumulative, twelve-week process of structured exploration. I introduce a prompt engineering and critical analysis framework that guides students through writing more effective prompts and critically engaging with AI output. We move beyond simple questions and answers. I task them with having AI produce complex academic work, such as literature reviews and research proposals, which they would then systematically interrogate. Their task is to question everything. Does the output actually adhere to the instructions in the prompt? Can every claim and statement be verified with a credible, existing source? Are there hidden biases or a leading tone that misrepresents the topic or their own perspective?
Pulling Back the Curtain on AI
As they began this work, the curtain was pulled back on the ‘magic’ machine. Students quickly discovered the emperor had no clothes. They found AI-generated literature reviews cited non-existent sources or completely misrepresented the findings of real academic papers. They critiqued research proposals that suggested baffling methodologies, like using long-form interviews in a positivist study. This process forced them to rely on their own developing knowledge of module materials to spot the flaws. They also began to critique the writing itself, noting that the prose was often excessively long-winded, failed to make points succinctly, and felt bland. A common refrain was that it simply ‘didn’t sound like them’. They came to realise that AI, being sycophantic by design, could not provide the truly critical feedback necessary for their intellectual or personal growth.
This practical work was paired with broader conversations about the ethics of AI, from its significant environmental impact to the copyrighted material used in its training. Many students began to recognise their own over-dependence, reporting a loss of skills when starting assignments and a profound lack of satisfaction in their work when they felt they had overused this technology. Their use of the technology began to shift. Instead of a replacement for their own intellect, it became a device to enhance it. For many, this new-found scepticism extended beyond the classroom. Some students mentioned they were now more critical of content they encountered on social media, understanding how easily inaccurate or misleading information could be generated and spread. The module was fostering not just AI literacy, but a broader media literacy.
From Blind Trust to Critical Confidence
What this experience has taught me is that student overreliance on AI is often driven by a lack of confidence in their own abilities. By bringing the technology into the open and teaching them to expose its limitations, we do more than just create responsible users. We empower them to believe in their own knowledge and their own voice. They now see AI for what it is: not an oracle, but a tool with serious shortcomings. It has no common sense and cannot replace their thinking. In an educational landscape where AI is not going anywhere, our greatest task is not to fear it, but to use it as a powerful instrument for teaching the very skills it threatens to erode: critical inquiry, intellectual self-reliance, and academic integrity.
Tadhg Blommerde
Assistant Professor Northumbria University
Tadhg is a lecturer (programme and module leader) and researcher that is proficient in quantitative and qualitative social science techniques and methods. His research to date has been published in Journal of Business Research, The Service Industries Journal, and European Journal of Business and Management Research. Presently, he holds dual roles and is an Assistant Professor (Senior Lecturer) in Entrepreneurship at Northumbria University and an MSc dissertation supervisor at Oxford Brookes University.
His interests include innovation management; the impact of new technologies on learning, teaching, and assessment in higher education; service development and design; business process modelling; statistics and structural equation modelling; and the practical application and dissemination of research.
An academic experiment unfolds: Visualizing the stark differences in engagement and performance between students who used AI and those who did not, as observed by one professor. Image (and typos) generated by Nano Banana.
Source
Gizmodo
Summary
A study by University of Massachusetts Amherst professor Christian Rojas compared two sections of the same advanced economics course—one permitted structured AI use, the other did not. The results revealed that allowing AI under clear guidelines improved student engagement, confidence, and reflective learning but did not affect exam performance. Students with AI access reported greater efficiency and satisfaction with course design while developing stronger habits of self-correction and critical evaluation of AI outputs. Rojas concludes that carefully scaffolded AI integration can enrich learning experiences without fostering dependency or academic shortcuts, though larger studies are needed.
Key Points
Structured AI use increased engagement and confidence but not exam scores.
Students used AI for longer, more focused sessions and reflective learning.
Positive perceptions grew regarding efficiency and instructor quality.
AI integration encouraged editing, critical thinking, and ownership of ideas.
Researchers stress that broader trials are required to validate results.
Diverse strategies in action: English professors are developing unique and personalised methods to encourage original thought and deter the misuse of AI in their classrooms. Image (and typos) generated by Nano Banana.
Source
Yale Daily News
Summary
Without a unified departmental policy, Yale University’s English professors are independently addressing the challenge of generative AI in student writing. While all interviewed faculty agree that AI undermines critical thinking and originality, their responses vary from outright bans to guided experimentation. Professors Stefanie Markovits and David Bromwich warn that AI shortcuts obstruct the process of learning to think and write independently, while Rasheed Tazudeen enforces a no-tech classroom to preserve student engagement. Playwriting professor Deborah Margolin insists that AI cannot replicate authentic human voice and creativity. Across approaches, faculty emphasise trust, creativity, and the irreplaceable role of struggle in developing genuine thought.
Key Points
Yale English Department lacks a central AI policy, favouring academic freedom.
Faculty agree AI use hinders original thinking and creative voice.
Some, like Tazudeen, impose no-tech classrooms to deter reliance on AI.
Others allow limited exploration under clear guidelines and reflection.
Consensus: authentic learning requires human engagement and intellectual struggle.
The irony of a digital dilemma: Students caught using AI to cheat are now turning to the same technology to craft their apologies. Image (and typos) generated by Nano Banana.
Source
The New York Times
Summary
At the University of Illinois Urbana–Champaign, over 100 students in an introductory data science course were caught using artificial intelligence both to cheat on attendance and to generate apology emails after being discovered. Professors Karle Flanagan and Wade Fagen-Ulmschneider identified the misuse through digital tracking tools and later used the incident to discuss academic integrity with their class. The identical AI-written apologies became a viral example of AI misuse in education. While the university confirmed no disciplinary action would be taken, the case underscores the lack of clear institutional policy on AI use and the growing tension between student temptation and ethical academic practice.
Key Points
Over 100 Illinois students used AI to fake attendance and write identical apologies.
Professors exposed the incident publicly to promote lessons on academic integrity.
No formal sanctions were applied as the syllabus lacked explicit AI-use rules.
The case reflects universities’ struggle to define ethical AI boundaries.
Highlights the normalisation and risks of generative AI in student behaviour.
