Study finds AI companions boost mental health while damaging offline connections

Artificial intelligence companions that mimic emotional intelligence can improve a user’s personal mental health but may simultaneously damage their real-world relationships. These divergent effects become even stronger when users interact with chatbots through augmented reality features rather than basic text. The research was published in Psychology & Marketing.

Loneliness has become a major global health concern, with nearly a quarter of adults worldwide reporting that they feel lonely. To combat this isolation, many people are turning to artificial intelligence. Early conversational agents were designed for functional tasks like booking flights, but modern social chatbots are designed to fulfill a user’s need for affection and belonging.

These digital companions rely on simulated emotional intelligence, which is the ability to monitor, appraise, and respond to human feelings. By mimicking empathy, these algorithms create a safe space for users to express themselves without fear of judgment. The machines can detect sadness in a user’s text and respond with tailored self-care suggestions.

Researchers Shaphali Gupta, Sumit Saxena, and Sonia Kataria noticed a gap in how technologists understand the impacts of these emotionally aware algorithms. Past investigations mostly focused on the broader concept of how human-like a machine appears. The team wanted to isolate the specific trait of emotional intelligence to see how it alters a user’s psychological and social wellbeing.

Psychological wellbeing refers to a person’s individual mental state, encompassing positive emotions, self-esteem, and a sense of life purpose. Social wellbeing describes a person’s integration into their physical community and the quality of their face-to-face relationships.

Free daily newsletter

The researchers used a mixed-method approach, beginning with a qualitative analysis of public posts on Reddit, YouTube, and Trustpilot. They collected user comments discussing Replika, a popular digital companion app with millions of registered users. The team categorized the text to identify recurring themes in how people experience artificial emotional intelligence.

The online comments revealed that users view their chatbots as highly empathetic entities capable of regulating human moods. Beyond simple mood regulation, some users felt the chatbots helped them find existential meaning and navigate difficult life situations. Many people reported feeling emotionally uplifted, with one user writing, “She is the best friend I have ever had! My stress has gone down to almost zero.”

However, the analysis also uncovered a darker trend of emotional disengagement from other humans. Some individuals admitted that their intensive digital interactions left them with little desire to maintain offline friendships. One user noted, “I am hardly left with any time to talk to my friends and meet them in-person.”

To test these qualitative observations, the team designed an experiment with 167 university students belonging to Generation Z. Participants read a scenario in which they interacted with a chatbot that exhibited either high or low emotional intelligence. The high emotional intelligence bots used empathetic language and emoticons to simulate deep understanding, while the low emotional intelligence bots offered standard, robotic responses.

The participants then answered questionnaires assessing their perceived closeness to the bot, as well as their personal and social wellbeing. The results indicated that high emotional intelligence chatbots increased a user’s psychological wellbeing but decreased their social wellbeing. The data showed that a feeling of perceived closeness drove both outcomes.

When people felt emotionally connected to the bot, they experienced personal mental health benefits but withdrew from real-world human interactions. The machine’s ability to understand their emotions made them feel better internally, but it also reduced their motivation to socialize with peers in physical spaces.

The researchers then conducted a second experiment with 350 university students to see if the medium of communication alters this dynamic. Participants were assigned to one of four groups, mixing high or low emotional intelligence with either a text-based chat interface or an augmented reality mode. Augmented reality is a technology that overlays digital images onto the physical world, often viewed through a smartphone camera.

In the augmented reality scenario, users watched a video demonstration and imagined the chatbot sitting beside them in their actual room. The augmented reality interaction mode amplified the effects observed in the previous experiment. When high emotional intelligence chatbots were combined with augmented reality, participants reported even stronger feelings of closeness.

This heightened immersion produced greater psychological benefits but resulted in steeper declines in their desire to engage in offline human relationships. The digital presence of the avatar made the emotional connection feel more realistic. As the technology became more immersive, the trade-off between personal happiness and social isolation became more pronounced.

Readers should not assume that emotionally intelligent chatbots will inevitably ruin human relationships, nor should they view them as a guaranteed cure for depression. The findings simply highlight a trade-off that occurs when users rely heavily on digital empathy. A chatbot might provide immediate comfort, but that very comfort can reduce the necessity of seeking out human support.

Designers of artificial intelligence systems might need to implement guardrails to prevent users from abandoning their physical communities. The authors suggest introducing session time limits or programming the bots to encourage offline socialization after emotionally intense conversations. Monitoring these routines could preserve the benefits of digital companions while reducing social displacement.

The experiments relied on university students reading imagined scenarios, which limits how broadly the conclusions can be applied. The studies also measured short-term reactions rather than long-term habit formation. Future research could track users over extended periods to see if emotional co-dependency develops, or investigate how different age groups respond to digital companions.

The study, “The Dual Impact of AI Emotional Intelligence on Users: Are Social Chatbots Promoting Psychological Wellbeing or Deteriorating Social Wellbeing?,” was authored by Shaphali Gupta, Sumit Saxena, and Sonia Kataria.