Is AI Making Us Stupid? Exploring Cognitive Offloading, Memory and Brainpower

Introduction: The AI Brainpower Question

Guardian Science Weekly peers Maddie and Ian Sample explore a timely concern: does AI make us less intelligent by taking over thinking tasks? The discussion situates AI as a tool that can write essays, summarize information, and provide solutions in seconds, raising questions about creativity, critical thinking, and cognitive effort in the AI era.

The MIT Study and Cognitive Debt

The episode centers on MIT research where 54 students were assigned to write three essays under three conditions: using a large language model (ChatGPT), using only a search engine, or relying on their own brains. EEGs, linguistic analyses, and post-task interviews were used to assess cognitive engagement. The researchers observed reduced brain connectivity and lower cognitive engagement in AI-assisted and search-assisted groups, with the ChatGPT group showing the most pronounced effects. Interviews suggested participants using AI struggled to recall specifics from their essays, a phenomenon described as cognitive debt. “Over these 4 months of writing these essays, the large language model users consistently underperformed at neural, linguistic, and behavioural levels.” – Ian Sample

Some researchers have challenged the certainty of a cognitive debt, noting study design caveats, but the findings still illuminate how technology reshapes brain activity during thinking tasks.

Memory, the Google Effect, and Offloading

The conversation extends to digital amnesia or the Google effect—the tendency to remember where information is found rather than the information itself. The panel shares personal anecdotes about outsourcing memory to devices and discusses MRI findings showing memory stores shifting when reminders are available externally. Cognitive offloading allows the brain to conserve resources for more complex tasks, but it may reduce memory for stored content itself. Sam Gilbert explains that offloading isn’t new, tracing back to Socrates’s worry about writing diminishing oral memory, and argues that technology changes how we use our brains rather than eroding innate ability.

Desirable Difficulty, Education, and the Extended Mind

The hosts discuss Robert Bjork’s desirable difficulty, the idea that challenging practice enhances long-term learning. They compare memorization drills, problem-solving, and the use of calculators in math education, noting that foundations matter but tools can support growth when used appropriately. The extended mind concept is raised, suggesting that phones and devices may extend cognitive capacity if used thoughtfully and with attention to preserving core knowledge in the brain. “There is a fundamental link between something being hard and your ability to understand it deeply,” says the discussion, highlighting the value of cognitive effort alongside external aids.

Practical Takeaways: When to Offload and When to Practice

The episode offers practical considerations for students, workers, and lifelong learners: use AI to handle repetitive or time-consuming tasks while retaining opportunities to problem-solve and recall information to develop robust mental models. The conversation notes GPS and digital tools can boost confidence and accessibility but at the potential cost of navigation knowledge and memory. The panel closes with reflections on motivation and the joy of problem-solving, warning against a future where rewards of effort diminish if we rely exclusively on AI for thinking. “There is a reason so many of us love puzzles,” Ian Sample concludes, underscoring the value of cognitive challenge.

Final Thoughts

The experts acknowledge the nuanced landscape: AI can support memory and learning or blunt cognitive skills if misused. The takeaway is to design assessments and tasks that incentivize thinking, maintain a knowledge foundation, and consciously decide when to offload and when to train the brain through difficult work.