Through the prism of artificial intelligence, the latest language models such as ChatGPT provoke numerous debates, particularly regarding their impact on our brain activity. A recent column suggests a decline in cognitive abilities, but a more rigorous study reveals different results. Through this analysis, we will discover how the use of ChatGPT, instead of harming our brain, could actually lighten our cognitive load and redirect our mental resources toward more complex tasks.
As part of a study aimed at evaluating the cognitive impact of various technological tools when writing an essay, the results show a refreshing perspective on the use of language models like ChatGPT. Contrary to claims of decreased brain activity, the study highlights the facilitative role of these technologies. The experiment measured the effects of using ChatGPT compared to a traditional search engine and work without technological aids.
The conditions of the study: a journey into cognition
The researchers established three distinct conditions for this study:
- The exclusive use of a Large Language Model (LLM) like ChatGPT.
- Searching for information using a traditional search engine.
- The use of personal knowledge without any external aids.
The participants were equipped with EEG headsets to measure their brain data throughout the experience. These measurements allowed observing the results on the cognitive load, that is, the mental effort required to carry out a task.
Surprising results: reduced cognitive load
One of the most interesting results of this study indicates that using a search engine decreases cognitive load compared to a state of work without assistance. In comparison, using ChatGPT leads to an even more pronounced reduction in this cognitive load. This phenomenon could be interpreted as an outsourcing of certain mental tasks.
The research showed that the more sophisticated the tool, the less intense the neural connectivity is. Rather than being a cause for alarm, this observation underscores the efficacy of technological tools that allow us to optimize our mental resources, thus redirecting energy towards more complex executive functions such as planning and software structuring.
The educational implications of using ChatGPT
The study raises pertinent questions about the role of such tools in the educational environment. It highlights the idea that the order of exposure to technologies influences our cognitive engagement. For instance, the results show that participants who first used ChatGPT and then found themselves without assistance exhibited lower neural connectivity. Conversely, those who transitioned from a situation without technology to using an LLM saw their connectivity improve.
This raises concerns about introducing these tools at the beginning of the learning process. It is therefore essential to reflect on the methods and contexts in which LLMs can be integrated constructively.
Trust in language models: a vigilance issue
Meanwhile, another study addresses the relationship we have with LLMs and how it influences our critical thinking. The idea is that the more we know how to trust these tools, the less we mobilize our ability to analyze and argue. This potential dependence on LLMs could impact how we process information. The example of a calculator illustrates this point well: while it provides reliable results, an LLM can generate errors.
Toward a thoughtful support of AI usage
The insights drawn from this research are not merely theoretical; they carry real implications for educational policies. Banning the use of AIs does not seem to be a viable solution, as highlighted by Albert Moukheiber. The idea is rather to frame the use of technologies like ChatGPT, to clarify their functioning, and to identify contexts where their use is beneficial.
It is crucial to incorporate a form of digital literacy to avoid two main pitfalls: on one hand, the reproduction of existing inequalities, and on the other hand, the sometimes hyperbolic discourse of the tech industry, which influences educational policies. The goal is not simply to use them, but to collectively decide how we wish to integrate these tools into our collective practice.
