MIT's Media Lab AI Assistant Study

Researchers from MIT's Media Lab conducted a study to monitor the brain activity of writers.

Why it matters: They discovered that users of ChatGPT exhibited minimal brain engagement and consistently performed poorly in neural, linguistic, and behavioral aspects.

• The investigation involved 54 participants from the Boston area, aged 18 to 39, who were divided into three groups.

• Each group was tasked with writing multiple SAT essays using different resources: OpenAI's ChatGPT, the Google search engine, or no tools at all.

Context: The findings suggest that the use of large language models (LLMs) may be detrimental to learning; however, the sample size for this study is relatively small.

• Nonetheless, Nataliya Kosmyna, the primary author of the paper, felt it was important to share these findings to address concerns that society's increasing reliance on LLMs for quick answers might compromise long-term brain development.

Between the lines: In contrast, participants in the no-tools group exhibited the highest level of neural connectivity, particularly in the alpha, theta, and delta bands, which are associated with creativity, ideation, memory load, and semantic processing.

• This group was more engaged and curious, taking ownership of their work and reporting higher satisfaction with their essays.

Yes, but: The third group, which utilized Google Search, also reported high satisfaction and demonstrated active brain function.

• This distinction is significant as many people now turn to AI chatbots for information instead of traditional search engines like Google.

The other side: Through practical AI literacy training, we have found that employees who utilize LLMs and other AI tools, such as Grammarly and Co-Pilot, can boost their productivity by more than one hour each day.

• For instance, AI tools like ChatGPT are not meant to replace search engines; instead, they are designed to automate research-intensive tasks.

• This is a subtle but essential distinction.

Be smart: When engaging interns or associates with research tasks, most knowledge workers do not directly use what is given.

• Instead, they write their synopsis based on the information provided.

• Including fact-checking studies, statistics, and other key information.

• The same should be done when using an AI assistant.

Go deeper: The problems discovered by MIT's Media Lab are all correctable with proper AI Literacy training.

• Want to find out what that looks like for your company?

• Reach out to Todd Moses & Co today to learn more.