"this group performed worse than their counterparts"
Relying on AI tools could harm your ability to think, learn, and remember, a new study has found.
Academics at MIT Media Lab observed students using large language models (LLMs) such as ChatGPT and reported poorer memory and weaker brain activity compared to those using traditional methods.
Over four months, researchers tracked 54 university students as they wrote essays using three different approaches – ChatGPT, Google, and no external tools.
Brain activity was recorded using electroencephalogram (EEG) scans.
The researchers, who published their study in ‘Your brain on ChatGPT‘, concluded:
“The group who used AI had a measurable impact and had a likely decrease in learning skills.”
Despite early benefits from AI-assisted writing, the ChatGPT group showed the lowest performance across “neural, linguistic, scoring” metrics.
The researchers wrote: “Even though the benefits were initially apparent, this group performed worse than their counterparts in the brain-only group at all levels.”
Students who used ChatGPT reportedly struggled to recall what they had written and felt less connected to their work.
By contrast, the group which didn’t use any AI tools produced more original ideas and showed higher engagement.
The Google users showed “moderate engagement”, but the most cognitive activity was observed in the group who completed the task without help.
Brain scans also showed a curious contrast: those who had not used AI earlier exhibited increased brain activity when later trying out ChatGPT.
Researchers believe this was because they attempted to actively integrate the tool into what they already knew.
However, students who had used ChatGPT from the beginning continued to show reduced brain activity, even when later told to write essays without it.
Students using ChatGPT said they experienced fewer obstacles when answering questions, performing with “reduced friction” compared to the Google group.
But researchers issued a stark warning about this apparent ease of use.
“This convenience came at a cognitive cost, diminishing users’ inclination to critically evaluate the LLM’s output or ‘opinions’ (probabilistic answers based on the training datasets).
“This highlights a concerning evolution of the ‘echo chamber’ effect: rather than disappearing, it has adapted to shape user exposure through algorithmically curated content.”
The study also raised concerns about corporate influence, warning:
“What is ranked as ‘top’ is ultimately influenced by the priorities of the LLM’s shareholders.”
Loading ...