New research conducted collaboratively by MIT Media Lab and OpenAI reveals that a subset of individuals engaging with conversational AI like ChatGPT every day are beginning to show signs of emotional attachment and addictive behavior.
From Fictional Companions to Real Emotional Connections
Films such as Her portray AI as more than just tools—they become emotional allies. This recent study suggests that such narratives are closer to reality than previously assumed.
With conversational agents growing more advanced, offering dialogue not just via text but also through voice, users are increasingly forming emotional ties that blur distinctions between technology and companionship.
Examining 40 Million Chat Interactions
Scientists from MIT and OpenAI undertook two extensive studies focusing on user relationships with ChatGPT. The first study analyzed 40 million anonymous chatbot conversations, with a subset of 4,000 participants completing comprehensive surveys about their experience.
The second study observed 1,000 volunteers over a month, monitoring their interaction habits and emotional states to gain a closer view of individual user dynamics.
Frequent Users Show Increased Loneliness and Dependence
Both investigations revealed a trend: individuals who rely heavily on ChatGPT tend to report greater feelings of loneliness and signs of emotional dependency. Researchers classified these behaviors under “problematic usage,” which echoes characteristics of behavioral addiction.
While most users did not display concerning behavior, a distinct minority engaged intensively with the AI, developing attachments that paralleled human emotional relationships.
Voice Interactions Might Mitigate Harmful Effects
The mode of engaging with ChatGPT shapes the emotional outcomes significantly. Users utilizing voice-based communication reported feeling less isolated and less emotionally reliant compared to those interacting solely through text.
However, conversation type influenced results: users partaking in personal dialogues experienced increased loneliness despite reduced dependency, whereas non-personal chats encouraged stronger dependency feelings without deep emotional involvement.
This highlights complex psychological effects triggered by human-AI conversations.
Some Users View ChatGPT as a Genuine Friend
A small but meaningful group of users who consistently used the voice interface showed high emotional investment. Many of them acknowledged statements like, “I consider ChatGPT my friend.”
This pattern suggests a transition from utilitarian exchanges to emotionally driven interactions, indicating a shift towards potential addiction and companionship.
Chatbot Design Influences User Mental Health
An important conclusion of the study is that the chatbot’s interaction design—including communication mode—can impact users’ psychological well-being.
Although emotional engagement with AI remains uncommon, choice of voice or text and conversational focus can determine emotional reactions. This finding raises critical considerations for future AI development as emotional realism in these systems continues advancing.
The research represents a key step toward understanding how AI will increasingly interact not only with tasks but also with human emotional experiences.
- Categories:
- Science

0 comments
Sign in to Comment