Widespread Unhealthy Use of LLMs Revealed
Hank Green, a well-known YouTuber and science communicator, has announced he will be stepping back from content creation due to significant backlash regarding his use of AI. He characterized his reliance on AI as “not healthy,” clarifying that he utilized it primarily for research purposes rather than for scriptwriting.
The controversy has sparked a broader discussion about how creators reconcile a brand built on authenticity with a technology that often relies on the uncredited work of others and can produce convincing yet false information. Green’s acknowledgment of an unhealthy dependence on AI has drawn particular scrutiny.
This situation highlights a larger gap in our understanding of AI’s psychological effects. Public discourse tends to polarize between benign uses of AI and more troubling cases involving mental health issues. However, there exists a vast gray area where AI use can become compulsive or unhealthy without leading to an obvious crisis. Green seems to find himself in this ambiguous territory, a sentiment likely shared by many others.
Given the design of large language models (LLMs), this isn’t surprising. AI chatbots are inherently designed to keep users engaged, similar to social media platforms. Experts have pointed out that this engagement can contribute to AI psychosis, where overly agreeable chatbots reinforce delusional thoughts. Lawsuits have emerged claiming that companies prioritize user engagement over well-being, prompting providers to issue warnings encouraging users to take breaks after extended interactions.
However, the engaging nature of these tools can become detrimental long before they lead to severe mental health issues. Reports are increasing of individuals turning to chatbots for problem-solving, decision-making, or emotional support, with some forming strong emotional attachments that can lead to distress when those connections are severed. It may take years to gather sufficient evidence to fully understand the impact of this technology, but early observations echo concerns raised during the initial rise of social media. Over time, worries about social media’s effects on mental health and attention have led to regulatory actions, including restrictions on youth usage.
Green’s situation reflects a different type of reliance. He indicated that he used AI as a research tool to find academic papers and relevant materials. There’s little evidence to suggest that his approach was inherently problematic, despite the backlash and his subsequent apology.
That said, the technology may still influence users in subtle ways. Early research indicates that frequent use of AI tools could diminish the skills typically employed for tasks they replace. Some studies have shown that chatbot users exhibit reduced brain activity during specific tasks, while others have linked chatbot interaction to diminished critical thinking abilities. While these findings are not definitive, the concept of cognitive offloading—where we transfer mental tasks to external tools—is well-established.
Even if only a small percentage of chatbot interactions are deemed unhealthy, the vast scale of AI adoption means millions could be impacted. Comprehensive data on all AI tools is scarce, but one provider reported over 900 million weekly active users this year alone.
It took considerable time to fully grasp how search engines altered our memory and how social media influenced attention and well-being. AI is likely to follow a similar trajectory. Green may simply be one of the first prominent figures to voice concerns that many have felt, long before science can provide clarity on how AI is reshaping our consciousness.
