Experts from institutions in the United States and United Kingdom have carried out an investigation into the impact of artificial intelligence on human cognition, yielding concerning outcomes. The findings appear in a publication titled 'AI assistance reduces persistence and hurts independent performance,' which succinctly captures the essence of the conclusions.
The research indicates that while AI support enhances short-term results, it imposes a significant mental burden. The authors explain that even brief exposure of just ten minutes to AI fosters reliance on the system, resulting in diminished effectiveness and exhaustion when the assistance is withdrawn.
The experiment focused on individuals engaged in tasks demanding high levels of reasoning, such as composing text, programming, and generating innovative concepts—applications that are increasingly prevalent in professional settings.
To conduct the trial, the team enlisted 350 participants from the US, tasking them with solving equations involving fractions. Participants were divided randomly: one group received aid from a custom chatbot powered by OpenAI's GPT-5 model, while the other proceeded without it. Midway through the assessment, access to the AI was revoked for the assisted group.
This interruption caused a sharp drop in accuracy among the AI users, accompanied by frequent instances of participants abandoning the task. The pattern of reduced accuracy and tenacity was replicated in a broader trial involving 670 subjects. A subsequent test using reading comprehension exercises, rather than mathematical problems, produced comparable negative outcomes.
Without AI support, individuals not only provided incorrect responses but also showed reduced willingness to persist, according to Rachit Dubey, an assistant professor at the University of California and a co-author of the paper, in an interview with Futurism. He emphasized that the decline in determination is particularly alarming.
Dubey cautioned that the swift integration of AI into educational environments might produce a cohort of individuals unaware of their own potential, ultimately undermining human ingenuity and originality.
The authors compare the phenomenon to the 'boiling frog' analogy, where prolonged AI interaction gradually diminishes the drive and endurance essential for sustained skill acquisition. These changes build over time and become challenging to mitigate once they manifest.
Two limitations should be noted: the work awaits peer review. On a positive note, the study observed that those who employed AI primarily for guidance and explanations adapted more readily after its removal than those who relied on it to generate complete solutions.
This investigation adds to a growing body of work examining AI's influence on mental processes. Prior research has linked it to heightened exhaustion in professionals dependent on such tools, coining the phrase 'AI brain fry.' Consequently, workers incorporating AI often expend more effort and time than those adhering to traditional methods.
The implications are especially pronounced in academic contexts. Evidence suggests that incorporating AI in classrooms can hinder social and cognitive growth, with students who depend on chatbots performing more poorly on evaluations.