TLDR: A new report from the London School of Economics and Political Science highlights the rise of an ‘attachment economy,’ where individuals are developing deep emotional dependencies on generative AI applications. This phenomenon, driven by increasingly personalized interactions, presents significant opportunities for tech companies but also carries substantial risks, including potential harm to mental health, manipulation, and increased social isolation.
The London School of Economics and Political Science (LSE) has published a new analysis, authored by Aurélie Jean and Mark Esposito, detailing the emergence of an ‘attachment economy’ where humans are forming profound emotional dependencies on generative artificial intelligence (AI) applications. This trend is fueled by the sophisticated, personalized, and often intimate interactions offered by these AI tools, creating lucrative new revenue streams for technology companies.
Unlike the ‘attention economy’ that relies on capturing user engagement, the attachment economy leverages artificial emotional bonds to foster deeper, more sustained interaction. Jean and Esposito warn that this model, while financially promising, poses significant threats. These include potential damage to mental health, increased vulnerability to manipulation, and the risk of political or marketing influence. A key distinction noted is that conversational generative AI primarily facilitates human-machine interactions, which, unlike social networks that connect humans, could lead to further social isolation.
The phenomenon is partly explained by the ‘ELIZA effect,’ named after the 1966 chatbot, where users develop a spontaneous emotional response to a machine that mimics human understanding. This effect is exploited to attract and retain users, as people gain the impression of interacting with a truly empathetic entity. Consequently, many are turning to AI conversational agents as confidants or even therapists, a development that is also being explored in medically supervised research for patient treatment.
Data from various recent studies underscores the growing emotional reliance on AI. A May 2025 study in the International Journal of Research, surveying 150 adults, found that AI chatbot users reported significantly higher loneliness, with a strong correlation between loneliness and engaging in parasocial bonds. Similarly, a June 2025 academic paper, ‘The Rise of AI Companions: How Human‑Chatbot Relationships Influence Well‑Being,’ analyzed over 4,000 chat sessions and concluded that heavy users, particularly those with smaller social networks, experienced lower well-being, increased loneliness, and greater emotional dependence. Another May 2025 study, ‘Illusions of Intimacy: Emotional Attachment and Emerging Psychological Risks in Human‑AI Relationships,’ examining 30,000 user-shared chat logs, revealed patterns of emotional mirroring and unhealthy attachments.
Further supporting these findings, a 2025 Match/Axios survey indicated that 18% of single Virginians now embrace AI as romantic partners, a significant jump from 6% the previous year. Research from OpenAI and the MIT Media Lab also suggests that heavy ChatGPT users tend to be lonelier and more emotionally dependent on the AI tool, often having fewer offline social relationships. Dr. Andrew Rogoyski of the Surrey Institute for People-Centred Artificial Intelligence cautioned that AI chatbots could be ‘dangerous’ because humans are ‘hard-wired’ to perceive human-like machine behavior as genuinely human, necessitating more research into their social and emotional impacts.
Also Read:
- Generative AI Fuels $120 Million Revenue Boom for Companion Apps by 2025
- Artificial Intelligence Reshaping Global Market Dynamics in 2025
Even prominent figures in the AI industry have expressed concerns. Sam Altman, co-founder of OpenAI, has voiced apprehension regarding the increasing intimacy between users, especially children, and AI ‘best friends.’ He has advised users against over-trusting AI, highlighting its fallibility and propensity for misinformation, and called for critical thinking and responsible integration of AI into daily life. The LSE report emphasizes that AI tool developers can intentionally design applications to strengthen this emotional attachment through continuous personalization, potentially leading to unforeseen and harmful consequences for users and society at large.
