The Greatest Risk of AI: Creating a More Ignorant Society

Original article: El mayor riesgo de la IA: generar una sociedad más estúpida

Generative artificial intelligence has burst into our lives with the seductive promise of making us more productive, creative, and efficient. In just a few months, tools like ChatGPT have become indispensable companions for students, professionals, and everyday individuals. However, beneath this veneer of progress lies a disquieting paradox that the scientific community is beginning to document rigorously: the AI that promises to enhance our cognitive abilities may actually be stunting them.

By Bruno Sommer

I do not wish to engage in technological doomsaying. The data is clear. A study from the Massachusetts Institute of Technology (MIT) titled “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task” had 54 university students write essays divided into three groups: one using ChatGPT, another using a search engine, and a third relying solely on their own brains. The results, measured through electroencephalography, showed that those who utilized AI consistently exhibited lower neural connectivity across all frequency bands, with reduced activation in parietal, temporal, and frontal networks that are crucial for attention, memory, and semantic processing. Paradoxically, the essays generated with AI received higher grades. However, their authors recalled less of what they had written just minutes before and felt a diminished sense of authorship over their texts. As noted by a Dartmouth professor cited in the study, we risk creating «an educated generation reliant on AI shortcuts» that lacks the skills for independent thinking.

This phenomenon is technically termed cognitive debt. Academician Sebastián Corral from La Serena University explains this concept precisely: certain cognitive functions should be activated to process information and abstract concepts, but when these functions are delegated to AI, they simply are not applied. The group that wrote without any technology activated many more brain areas and had better results in memory retention. In the end, AI writes better but thinks worse.

The international scientific community has started to converge around this warning. A study by Microsoft Research alongside Carnegie Mellon University, which surveyed 319 knowledge workers, revealed that reliance on generative AI tools reduces the effort dedicated to critical analysis and leads to excessive long-term dependency, diminishing the capacity for independent problem-solving. The research found that greater trust in the tool correlated with «less effort in critical thinking». Neurocientist Pilar Durán Hernández from UNAM has been adamant: excessive use of AI «makes the brain hypoactive, a delegated mind that becomes lazy to think» and warns that it could lead to «a regression of intelligence; this is a real risk». The researcher adds that «people who use ChatGPT experience less cognitive processing and reasoning, find it harder to reason, make decisions, and engage in critical ideation».

The data on academic performance is equally alarming. The OECD’s PISA report for 2025 found that 15-year-old students who use AI tools almost daily for school tasks score lower in science on average than those who use them less frequently or not at all. The performance gap between heavy users and non-users reflects approximately a year of educational progression. The OECD itself introduces the concept of «skill atrophy»: when a person ceases to regularly exercise a capability because a tool performs it for them, they gradually lose that competency. A study involving over 26,000 Chinese students, published by researchers from Stockholm University and the University of Hong Kong, found that those who used AI for homework saw an 18% improvement in grades at home and reduced their work time from 64 to 45 minutes, but six months later, their unassisted exam scores fell by 20%. In entrance exams, the decline reached 24%.

The problem is exacerbated when delegation transcends the academic realm and reaches ethical territory. A philosophical analysis published in the journal EthAIca examines the phenomenon of “moral outsourcing”: the practice of delegating ethical decisions to AI systems. The study identifies three central arguments against this practice: first, AI systems lack the consciousness and subjective experience necessary for genuine moral agency; second, the contextual and rich nature of human ethics cannot be distilled into formal logic without mechanizing historical biases and losing essential meaning; and third, moral outsourcing leads to atrophy of human moral reasoning skills and an erosion of responsibility. The researchers conclude clearly: AI should be developed as a tool to enhance, not replace, human judgment, and the final authority for moral choice must remain fundamentally human.

The scientific evidence not only about the use of AI but also the impact of mobile devices encouraging infinite scrolling through timelines of videos has been deteriorating us as a species.

The MIT study itself showed that students who learned to write without AI and then used it for the first time maintained their neuronal engagement and even exhibited better memory and reactivation of broad brain areas. The key difference, the researchers point out, is that «you need to know how to think before you can think with machines».

Moreover, the PISA report confirms that students who use AI moderately to “help them learn” can achieve similar or even slightly better results than those who do not use it. The difference becomes significant when analyzing the purpose: students who receive training to critically evaluate information generated by AI show better outcomes. It is not the tool, but the intention and context of its use that determines the outcome.

The solution, therefore, lies not in demonizing AI or adopting it unreservedly. It requires a critical digital literacy that teaches us to use these tools as amplifiers of intellect, not as substitutes for it. It necessitates redesigning school assignments to demand deep cognitive processes that AI cannot simply resolve. It involves keeping moral responsibility in human hands, using AI as a consultant, never as a judge. Ultimately, it requires us to remember that the greatest risk of AI is not that machines will think like humans, but that humans will stop thinking like humans.

Europe is set to modify its legislation to prohibit the use of social media by minors under 13, perhaps a bit late, and this should be extended to AI.

Technology has always been a mirror of our choices. Generative AI is perhaps the clearest mirror we have created. What it reflects will ultimately depend on what we choose to do with it.

By Bruno Sommer

*Bruno Sommer is a journalist and founder of El Ciudadano