A new study out of the University of British Columbia is highlighting the risk factors for AI chatbot addiction. With the popularization of large language models (LLMs), chatbots have become a part of everyday life for tens of millions of people—but researchers are worried about the lack of safeguards to protect users. “In AI chatbot addiction, one feature that we have highlighted in the paper is anthropomorphization of the chatbot that can make users feel like they’re personally having relationships with the chatbot,” Dr. Dongwook Yoon said. An associate professor of computer science and a senior author on the newly published study, Yoon argues that the more people rely on these chatbots for emotional support and companionship, the worse the problem will get. The researchers analyzed more than 300 Reddit users across 14 subreddits who discussed their experiences with AI addiction. The symptoms of chatbot addiction are similar to those of a gambling addition or a gaming addiction, the study reports. The dominating symptom reported by users is salience, or the activity dominating one’s thoughts and behaviours. Withdrawal, emotional disregulation, and cognitive change are among other commonly recorded symptoms. The UBC researchers say not all hope is lost, though. “If we can implement some safety-related regulations, then we can direct or guide the companies so they can implement content moderations or safer design features or user education which will be beneficial for even the company’s long-term sustainability,” Yoon explained. The researchers suggested a combination of design changes, like reminders the chatbot isn’t human, and more AI literacy could help users struggling with chatbot use. Ethical responsibility is being increasingly discussed in user experience circles, and the study hopes to inform companies and practitioners on AI chatbot addiction and best practices for risk monitoring.
