HCC: Small: Mind Perception in AI Companionship: Testing the Assumptions of Social Theories

Artificial intelligence (AI) is increasingly a part of everyday life for functional purposes (like interpreting x-rays or recommending entertainment), and also for companionship (like chatting or even just sitting together). Companionship is a positive state of close connection with someone or something that unfolds over time and is valued for itself.

Companionship is important to human life because it enhances well-being---for instance through reduced loneliness, enhanced emotional resilience, and finding elevated meaning in life. Evidence shows that people can see AI agents as mindful entities, and scientists and technologists often assume that creating more human-like, seemingly mindful AI, is necessary to foster companionship benefits.

However, there are no scientific studies demonstrating that seeing a machine as a “someone” actually enables or enhances companionship benefits. Without a full understanding of the link between machine-mind perception and companionship outcomes, we may be developing and using technologies that carry unnecessary risks to privacy and may even diminish well-being.

This project determines how to best measure the notion of mind perception, companionship, and well-being in human-AI relations. A series of studies then assesses the assumed link between perceiving AI systems as mindful entities and their efficacy as companions across different kinds of AI applications. By answering the fundamental question of whether mind perception plays a role in AI-companionship benefits, the work will ultimately help technologists make better decisions about AI design, public health officials make better decisions about AI policies, and everyday users make better decisions about whether and how they want to interact with AI companions.

To accomplish the desired outcomes, the project pursues three objectives: 1) Develop and validate measurements for AI mind perception, companionship, and relational benefits; 2) build a data-driven model of relationships between those variables; and 3) test the model in short- and long-term human-AI relations. Objective 1 will be achieved by analyzing public conversations about AI companions, generating and evaluating self-report measurement tools, validating existing measurements for use in human-AI contexts, and exploring behavioral indicators of mind perception, companionship, and well-being.

Objective 2 will be achieved through studies designed to identify direct or indirect relationships between mind perception, companionship, and well-being—experiments test the causal influence of mind perception on companionship experiences and subjective well-being. Objective 3 will be achieved by longitudinally testing the identified causal effects (via real-time surveys of experience) over short-term and long-term companionship interactions.

This work advances the science of social-psychological processes and AI companionship. It comes at a time when companionship, as a key component in fulfilled human life, is increasingly addressed by social AI. This project lays the evidential groundwork to determine whether and how current theories of human mind perception apply to AI companionship.

The research advances understanding of whether or not we must see someone in the machine for them to meaningfully contribute to human well-being.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.