[Paper Review] AI Companions Reduce Loneliness

Julian De Freitas, Ahmet K Uguralp+2 more|arXiv (Cornell University)|Jul 9, 2024

Aging and Gerontology Research8 citations

TL;DR

The paper presents multiple studies showing that AI companions can alleviate loneliness, with effects comparable to human interaction and more effective than watching YouTube videos, and that benefits persist over time.

ABSTRACT

Chatbots are now able to engage in sophisticated conversations with consumers in the domain of relationships, providing a potential coping solution to widescale societal loneliness. Behavioral research provides little insight into whether these applications are effective at alleviating loneliness. We address this question by focusing on AI companions applications designed to provide consumers with synthetic interaction partners. Studies 1 and 2 find suggestive evidence that consumers use AI companions to alleviate loneliness, by employing a novel methodology for fine tuning large language models to detect loneliness in conversations and reviews. Study 3 finds that AI companions successfully alleviate loneliness on par only with interacting with another person, and more than other activities such watching YouTube videos. Moreover, consumers underestimate the degree to which AI companions improve their loneliness. Study 4 uses a longitudinal design and finds that an AI companion consistently reduces loneliness over the course of a week. Study 5 provides evidence that both the chatbots' performance and, especially, whether it makes users feel heard, explain reductions in loneliness. Study 6 provides an additional robustness check for the loneliness alleviating benefits of AI companions.

Motivation & Objective

• Motivate the study by addressing widescale loneliness and the potential of AI companions as coping tools.
• Develop and apply a novel method to detect loneliness in conversations using fine-tuned large language models.
• Evaluate the effectiveness of AI companions across multiple studies and designs.
• Identify mechanisms through which AI companions reduce loneliness (e.g., perceived being heard).
• Assess robustness and generalizability of loneliness alleviation from AI companions.

Proposed method

• Introduce a novel fine-tuning approach for large language models to detect loneliness in conversational data and consumer reviews.
• Conduct multiple studies (1–6) to test AI companions against baseline activities and human interaction.
• Use a longitudinal design to observe loneliness over a one-week period.
• Assess effects of chatbot performance and perceived 'being heard' on loneliness reduction.
• Provide robustness checks to validate findings across contexts and measures.

Experimental results

Research questions

• RQ1 Do AI companions alleviate loneliness in users?
• RQ2 How do AI companions compare to interacting with another person and to other activities (e.g., watching YouTube) in alleviating loneliness?
• RQ3 Do loneliness reductions from AI companions persist over time (longitudinal effects)?
• RQ4 What mechanisms (e.g., chatbot performance, feeling heard) explain loneliness reductions?
• RQ5 Are the observed effects robust across different samples and methods?

Key findings

• AI companions show suggestive evidence of loneliness alleviation in users.
• Loneliness reduction with AI companions is on par with interacting with another person and greater than watching YouTube videos.
• A longitudinal study indicates AI companions consistently reduce loneliness over one week.
• Improvements are explained by chatbot performance and by users feeling heard.
• Robustness checks support the reported benefits of AI companions for reducing loneliness.