A practical guide to staying in control of your brain
The debate on generative artificial intelligence, and particularly on tools like ChatGPT, is often polarized. On one hand, techno-optimists describe it as a revolutionary force that will unleash human potential; on the other, critics portray it as a threat that will atrophy our cognitive abilities. As an association focused on digital sustainability, Sloweb takes a position of critical openness: technology is neither good nor bad, but its impact depends on how we choose to use it. Users must not be passive passengers, but conscious pilots, or even better, digital activists.
A recent MIT study, titled “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task”, sparked a wave of sensational headlines. Many media outlets hastily concluded: “ChatGPT makes us dumber.” But this simplification, however catchy, does not do justice to the study’s complexity and, above all, hides its most promising discovery. Analyzing this study through the lens of social and cognitive sustainability reveals a much more nuanced and encouraging truth.
The MIT study: what does it really say?
Researchers at the MIT Media Lab conducted an experiment to measure the cognitive impact of using a Large Language Model (LLM) during essay writing. They divided 54 participants into three groups:
- LLM Group: Could only use ChatGPT to write.
- Search Engine Group: Could use any website but not LLMs.
- “Brain Only” Group: Could not use any external tools.
By monitoring brain activity with electroencephalography (EEG) and analyzing the texts produced, some clear trends emerged.
The results of the first three sessions showed that brain connectivity systematically decreased with increasing external support: the “Brain Only” group showed the strongest and broadest neural networks, followed by the Search Engine group, and finally, the LLM group, which showed the overall weakest neural coupling.
Behaviorally, the LLM group showed significant difficulty: 83% of participants could not correctly cite a sentence from the essay they had just written, and perceived a lower sense of “ownership” of the work compared to the other groups.
The clickbait trap and “Cognitive debt”
It is precisely from these data that the alarmist headlines arise. But does lower brain activity really mean “getting dumber”? Not necessarily. Our brain is intrinsically geared towards energy saving. The use of an external tool that simplifies a task leads to what is called “cognitive offloading”: the brain delegates part of the work, reducing the immediate load.
The real risk, highlighted by the study’s title itself, is not an instant decrease in intelligence, but the accumulation of “cognitive debt”. Like financial debt, postponing mental effort today can lead to greater costs in the long run: atrophy of critical skills, reduced creativity, and increasing dependence on external tools. The passive and uncritical use of an LLM leads us precisely in this direction.
The surprising discovery: there’s a better way to use AI
The most fascinating part of the study (the one systematically ignored by the media) emerges in the fourth session. Here, the groups were reversed: those who had only used their brain were invited to use the LLM (“Brain-to-LLM” group), and vice versa (“LLM-to-Brain” group).
The results were surprising and overturned the pessimistic narrative:
- The LLM-to-Brain group, accustomed to AI assistance and then left to themselves, showed weaker neural connectivity, less engagement of alpha and beta networks (associated with planning and working memory), and a tendency to reuse the specific vocabulary of the LLM they had been exposed to, confirming the risk of “cognitive debt.”
- But it is the Brain-to-LLM group that gives us real hope. These participants, after writing essays based solely on their own faculties, when they used the LLM for the first time, showed an explosion of brain connectivity across all frequency bands. Their brains reactivated large occipito-parietal and prefrontal nodes, demonstrating greater memory recall and more extensive cognitive engagement.
This is not the chronicle of a brain shutting down, but of a brain empowering itself. The act of reprocessing autonomously created content with an LLM forced participants to a higher level of integration: they had to compare their original ideas with AI suggestions, evaluate, discard, integrate. This “hybrid cognition” process proved to be neurologically richer and more stimulating than passive AI use from the outset.
Towards cognitive sustainability: practical tips
The MIT study’s discovery is not a condemnation, but a user manual. It tells us that it is not the tool itself that makes the difference, but our approach. We can choose to accumulate cognitive debt or invest in cognitive growth. To do this, however, we must become active and conscious users.
Here are some strategies we can implement immediately:
- Brain 1st
Start the creative process with your mind, generating ideas or a draft based on your knowledge; subsequently, use AI primarily for revision and improvement (grammar, fluidity, reorganization, expansion), maintaining authorship of your ideas and formulating more effective prompts. - Fact-checking
For effective AI use, diversify tools (LLM for synthesis and ideas, then verify with other sources) and always manually fact-check the output, as AI can generate inaccurate or distorted information. - Critical prompt engineering
To develop “Critical Prompt Engineering” and promote active engagement with AI, it is essential not to passively accept answers. Instead of simply asking for an essay on a topic, it is advisable to formulate questions that stimulate critical thinking, such as requesting pros and cons, authoritative sources, identification of biases, or stylistic reformulations. It is also useful to request different perspectives on the same topic. In parallel, it is important to focus on the learning process, not just the final product, by practicing metacognition (“am I really learning or just delegating?”) and actively discussing with others to foster understanding and memorization. - Artificial sparring partner
Use AI as a “sparring partner” to strengthen critical thinking through questions and debates, and create personalized “quizzes” to test comprehension and memory after studying a topic.
In conclusion
The true challenge of digital sustainability is not a mere rejection of technological innovation, but the ability to govern it with wisdom and foresight. The MIT study, in particular, offers us an illuminating perspective: if used with critical intelligence and a deep awareness of its potential and limitations, artificial intelligence (AI) can transcend the role of a mere tool to become a true training ground for our brain.
AI, far from replacing human cognitive abilities, can act as a catalyst for the development of new skills and the enhancement of existing ones. Through interaction with intelligent systems, we can be stimulated to think in new ways, solve complex problems, process information more quickly, and develop more refined critical thinking. This means that AI does not make us mentally lazy, but on the contrary, pushes us to overcome our limits, explore new frontiers of knowledge, and improve our learning and adaptation capabilities.
Therefore, it is essential to adopt a proactive and informed approach to integrating AI into our daily and professional lives. We must learn to discern information, critically evaluate sources, formulate pertinent questions, and use AI as an ally to enhance our intellectual faculties, rather than as a substitute for our judgment. In this way, AI will not be an obstacle to our evolution, but an engine for continuous and sustainable intellectual growth.
And then, of course, the cure for everything always remains the same: reading many good books.