Is AI Making Us Stupider? New Scientific Insights on Cognitive Impact

Researchers are investigating whether the integration of generative artificial intelligence into daily cognitive tasks leads to “cognitive atrophy,” a decline in the brain’s ability to perform independent problem-solving. According to reports from Science et Vie and La Presse, the central concern is that outsourcing mental effort to tools like ChatGPT may weaken the neural pathways responsible for critical thinking and memory.

Cognitive Offloading and Neural Plasticity

The phenomenon of “cognitive offloading”—delegating mental tasks to external devices—is not new, but the scale of generative AI has accelerated the process. Science et Vie reports that when a user relies on AI to structure an argument or solve a complex problem, the brain avoids the “cognitive friction” necessary for learning. This lack of effort can lead to a reduction in the activation of the prefrontal cortex, the area associated with executive function and decision-making.

This shift is linked to neural plasticity, the brain’s ability to reorganize itself based on usage. If specific mental skills are no longer exercised, the brain may prune those connections to save energy. Researchers cited by La Presse suggest that this does not necessarily make a person “stupid” in a general sense, but it may create a dependency where the user becomes unable to perform the task without the digital intermediary.

The Boundary Between AI Assistance and Human Decision

Mazarine M. Pingeot, in an interview with La Semaine, addresses the difficulty of distinguishing between an AI-generated suggestion and a human’s own decision. Pingeot highlights a risk of “cognitive hyper-connection,” where the speed of AI output bypasses the human reflective process. This creates a loop where the user accepts an AI’s output not because it is verified, but because the effort to challenge it is too high.

The danger is not the tool itself, but the erosion of the “critical distance” required to evaluate information. According to Les Echos, this hyper-connection can lead to a state where the user’s cognitive processes are mirrored by the AI, effectively narrowing the scope of original thought and limiting the ability to synthesize information from diverse, non-digital sources.

Comparative Impact on Learning and Memory

The impact of AI varies based on how the technology is applied to a task. Data analyzed by Noovo Info and La Presse suggests a contrast between “active” and “passive” AI use:

• Passive Use: Users who request a final answer or a completed text exhibit a higher risk of cognitive decline in the specific skill being outsourced.
• Active Use: Users who use AI as a Socratic tutor—asking the AI to guide them through a problem rather than providing the solution—may actually enhance their learning by receiving immediate, personalized feedback.

This distinction suggests that the “stupidity” associated with AI is not an inevitable biological outcome, but a result of how the interface is used. The risk is highest in educational settings where the goal is the development of a skill, rather than the mere production of a result.

Institutional and Research Next Steps

Researchers are now attempting to quantify the long-term effects of this shift through longitudinal studies. The goal is to determine if the cognitive skills lost to AI offloading can be recovered through targeted mental exercise or if the dependency creates a permanent shift in cognitive architecture. Currently, there is no consensus on a standardized “cognitive baseline” to measure this decline across different age groups.