AI Attachment Style Test (EHARS)

The Experiences in Human-AI Relationships Scale (EHARS) was developed in 2025 by psychologists Fan Yang and Atsushi Oshio to study how human attachment dynamics apply to conversational artificial intelligence. While traditional attachment theory examines emotional bonds between people, rapid advances in natural language processing have led many users to form psychological expectations of artificial agents. EHARS does not assume that artificial systems have emotions or inner lives; instead, it investigates the human user's personal responses, expectations, and interpersonal habits when communicating with an interactive system.

The first dimension measured by the scale is AI Anxiety. This facet captures the degree to which an individual seeks emotional reassurance, validation, or signs of closeness from an artificial intelligence. People with higher scores on this dimension may feel unsettled when an AI provides brief, neutral, or seemingly cold answers. They may also find themselves frequently asking the system for affirmations of support or personal connection. A lower score generally reflects emotional neutrality, where the user does not seek emotional validation or worry about how the system responds to them.

The second dimension is AI Avoidance, which gauges a person's reluctance to share personal feelings or develop emotional closeness with an artificial agent. Individuals scoring high in avoidance prefer to keep interactions strictly functional, pragmatic, and emotionally guarded. They often feel uncomfortable disclosing private thoughts or treating the system as a confidant. Conversely, those with low avoidance scores are more open to sharing personal experiences, treating conversations with AI as a safe or comfortable space to express internal thoughts and feelings.

Research suggests these dimensions are separate rather than opposite poles of one scale. A person can score high on both, low on both, or differently on each. Low scores on both may reflect little emotional involvement with AI, although these questions cannot tell why. The researchers also found that AI attachment responses do not directly mirror attachment patterns in human relationships. The EHARS is not a measure of feelings of loneliness, and its two bars should be read as a profile of responses toward a particular conversational tool rather than as a fixed personal type.

The population comparison markers displayed on your chart are based on published research data from a formal validation sample of 242 adult AI users. In that study, the sample mean was 2.26 for AI anxiety and 3.95 for AI avoidance on a seven-point scale. These markers are estimates rescaled from the published sample means, not validated percentiles or universal norms. The sample was relatively young and mostly female, and the original validation was conducted in Chinese. The English items supplied in the paper were not validated; this test further paraphrases them for an educational adaptation. Because the site uses a five-point response scale, your comparison with the seven-point source means is approximate. Use the reference lines as context for reading your two scores, not as a judgment about how unusual your responses are.

This test is provided for educational and self-exploratory purposes only and is not intended to serve as a psychological diagnosis, mental health evaluation, or measure of human interpersonal functioning. It does not evaluate whether conversational AI can or should replace relationships with other people. This assessment is an independent adaptation based on the published Experiences in Human-AI Relationships Scale and is not affiliated with or endorsed by its original authors or their academic institutions.

References

• Yang, F., & Oshio, A. (2025). Using attachment theory to conceptualize and measure the experiences in human-AI relationships. Current Psychology, 44, 10658–10669.

Frequently Asked Questions

Take Test