Is AI making us dumb?

A recent experiment conducted at MIT's Media Lab has some experts worrying that the stampede toward Large Language Models could be dulling and dumbing us down.

The MIT experiment that has educators worried

Ever wondered what might happen if 54 students were tasked to write SAT-style essays, one third using only their brains, another third using Google search, and the rest using ChatGPT?

The MIT study “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task” has revealed the concerning neural and behavioural consequences…

Research scientist Nataliya Kosmyna reported participants showing "fewer widespread connections between different parts of their brains; less alpha connectivity, which is associated with creativity; and less theta connectivity, which is associated with working memory." (source)

Relying on AI leads to "cognitive debt", making it harder to build your own thinking skills.

Some ChatGPT users in the study felt "no ownership whatsoever" over their essays, and 15 of the 18 students in this group couldn't quote from what they'd supposedly written.

Brilliantly bland: The perils of thinking the same

Just as concerning as dulled brains is dulled diversity of thought. AI users' essays converged on remarkably similar ideas.

For example, when asked about philanthropy, the ChatGPT group uniformly argued in favour, while the other two groups included critiques.

As Kosmyna puts it, "Average everything everywhere all at once – that's kind of what we're looking at here."

This mirrors a 2024 Cornell University study which found that AI users from different cultures had to fight the tendency of the platforms to favour Western norms.

When writing about food and pop culture, all participants were prompted with Western tropes, like pizza and Shaquille O'Neal, despite half of them hailing from India.

The creativity conundrum

While AI proponents promise an unleashing of creative potential, at least some early research suggests the opposite.

A 2024 Santa Clara University study challenged subjects with creative problems like "How could you make a jigsaw puzzle more engaging?"

Those using Brian Eno's 1975 ‘Oblique Strategies’ cards (containing creativity prompts like “ask your body” and “try faking it!”) came up with more diverse solutions than ChatGPT users.

Computer scientist Max Kreminski found users gradually "cede their original thinking" as they slip into "curationist mode" – simply selecting from AI suggestions rather than generating original thoughts.

So what now?

The MIT study has limitations. It involved only 54 participants and its results are not yet peer reviewed.

Writing for The Conversation, University of South Australia education experts Vitomir Kovanovic and Rebecca Marrone suggested some of the results may have derived from the structure of the study itself.

The solution isn't banning AI in education and workplaces, but deploying it strategically.

Balancing the benefits

Kovanovic and Marrone argue, "Knowing when, where and how to use AI is the key to long-term success."

University students who are allowed to use GPTs for essays could temper these grades with good old-fashioned pen and paper assessments. 

Job applicants could face a final, face-to-face, device-free interview that dives deep into their understanding of key concepts.

Perhaps the question isn't whether AI makes us smarter or dumber - it's whether we'll think with it rather than let it think for us.

Indeed, in a world of artificial intelligence, authentic human thinking might become our most valuable skill.

Yours in numbers,

-Adam S