Recently, my editor Chelsea expressed a compelling concern regarding the rise of AI. Unlike typical journalistic worries about job loss, copyright infringement, or dull writing, she highlighted a unique issue: the potential loss of joy from experiencing those delightful “aha!” moments. “For me, it’s almost a physical sensation, like a wave of euphoria washing over my brain,” she noted.

Her thought-provoking question was: If we increasingly delegate idea generation to AI, will our dopamine rush from solving problems diminish? What else might our brains be missing if we experience fewer of these revelatory moments?

As it turns out, these “aha!” moments provide much more than momentary joy. Research shows that they can fundamentally alter our brains, enhance our learning, and contribute to long-term health. Thankfully, even in an AI-centric world, there are steps we can take to mitigate losses, aside from forgoing tools like ChatGPT altogether.

Chelsea’s vivid description aligns well with scientific findings. As Carola Salvi, a researcher at John Cabot University in Italy, explains, while not every insight triggers a dopamine release, numerous studies indicate that dopamine plays a crucial role in these eureka moments.

For instance, research by Martin Tick and his team at the Medical University of Vienna in 2018 demonstrated that individuals solving problems designed to spark “Eureka” moments displayed noticeable changes in dopamine-related brain activity during fMRI scans. Activity peaks in the midbrain coincided with the sensation of saying, “Ah!” In contrast, when participants reached conclusions without any prior hunch, brain activity significantly dropped.

These “aha!” moments are not just pleasurable; they also offer considerable cognitive advantages related to learning and memory. Salvi suggests that they serve as an internal “selection signal,” allowing accurate and satisfying solutions to stand out. Thus, the brain, possibly aided by dopamine, marks these insights as significant.

This theory makes sense, given that ideas perceived as “aha!” are generally deemed more accurate than others. However, it’s essential to note that while “eureka!” moments offer useful signals, not all ideas that feel right are valid. Empirical evidence supports the role of sudden insights, or even “What the heck!” moments, in enhancing memory retention. Essentially, the emotional thrill Chelsea spoke of activates areas in the brain that help solidify memories of those moments. Brain scans during these insights indicate transformative changes in the neural pathways involved in memory and vision, linked to how effectively individuals recall learned information later.

“From an evolutionary standpoint, this makes perfect sense,” argues Salvi. “When your brain uncovers a beneficial new pattern, it’s crucial for that information to become ingrained.” Hence, the “aha!” moment acts as a tagging mechanism for valuable insights.

This brings us back to AI. By excessively relying on large-scale language models (LLMs) for generating ideas and solutions—even for minor dilemmas—are we depriving ourselves of essential learning and memory opportunities?

For insight, I reached out to Hannah Critchlow, a neuroscientist from the University of Cambridge and the author of The 21st Century Brain: How to Future-Proof Your Mind in the Age of AI.

She cited a fascinating study comparing neural activity in a group of 18 participants tasked with essay writing using only their cognitive abilities, with assistance from a search engine, or through ChatGPT. Those utilizing AI exhibited consistently lower brain activity compared to those relying solely on Google or their own intellect. Over four sessions conducted across four months, participants using ChatGPT faced challenges in accurately citing their work and displayed decreased performance across neurological, linguistic, and behavioral metrics.

Although the small sample size warrants caution, these findings highlight a potential paradox: while LLMs may seem to facilitate swift insights, they might inadvertently hinder long-term learning and memory retention.

So, how can we counter this trend without completely dismissing ChatGPT and similar tools? Critchlow emphasizes research indicating that collaborative idea discussions—held in non-competitive settings—can lead to greater flexibility of thought. Brain waves often synchronize during such exchanges.

This observation sheds light on the unique cognitive value human interactions provide, which cannot be duplicated by AI. Facilitating opportunities for brain synchronization proves advantageous. Critchlow asserts that a brain’s synchronization with others can be predictive of future cognitive health. “This synchronization may help guard against dementia and significantly influences adolescents’ ability to learn and bond with peers,” she concludes.

In essence, the solution is not merely to diminish our engagement with LLMs but to bolster human connections. Critchlow argues that educational institutions should foster a learning environment that prioritizes small, interactive group settings. “Perhaps paradoxically, these advanced tools are illustrating that our species’ success hinges on our capacity to connect and communicate. By sharing ideas and collaborating, we can unlock those gratifying ‘aha!’ moments, allowing us to solve problems collectively for the betterment of humanity.”

For those resonating with Chelsea’s sentiments, a simple takeaway emerges: while it might be tempting to lean on LLMs for quick insights, actively engaging your mind to discover answers autonomously not only boosts your immediate dopamine levels but also enhances your long-term learning and cognitive health.