# A conversational agent as a virtual therapist for patients diagnosed with schizophrenia: A preliminary study

## TL;DR

A conversational agent called Terabot was tested as a virtual therapist for schizophrenia patients, showing some promise but needing improvements in empathy and personalization.

## Contribution

This is the first preliminary study exploring a conversational agent's acceptability as a virtual therapist for schizophrenia patients.

## Key findings

- Patients found Terabot's clarity and therapeutic exercises useful, but reported moderate empathy and emotional understanding.
- Technical issues like freezing and stuttering were reported, along with a lack of personalization and dialogic flexibility.
- Facilitators observed patient engagement and openness to the technology despite its limitations.

## Abstract

Schizophrenia is a complex mental disorder characterized by disturbances in thinking, emotions, and behavior. Emotional dysregulation is a significant clinical issue. Recent years have seen growing interest in the use of conversational agents in mental health care, including for schizophrenia. This study explores the use of a conversational agent as virtual therapist (Terabot) designed to support the recognition and regulation of three emotions (anger, shame, and fear) in hospitalized individuals with schizophrenia. The pilot study aimed to assess the acceptability of this innovative intervention, focusing on protocol implementation, technical performance, and patient experience.

The study employed a mixed-methods design involving 35 inpatients diagnosed with schizophrenia (ICD-10: F20.0–F20.9). Quantitative evaluation was based on the Brief Psychiatric Rating Scale (BPRS) and a structured acceptability survey, while qualitative thematic analysis was conducted on patients’ open-ended responses and facilitators’ notes. Research procedures followed COREQ standards.

The study involved 35 patients; 34 completed the quantitative scales, and 32 completed the acceptability survey. The quantitative findings suggest that patients generally have a positive attitude toward the analyzed items, particularly Terabot’s clarity of communication, friendly appearance, and the usefulness of its therapeutic exercises, whereas scores for empathy and emotional understanding were only moderate. The qualitative analysis revealed three main areas of patient experience: (1) the therapeutic relationship with Terabot, including both perceived strengths and clear limitations in relational depth, (2) the perceived usefulness of therapeutic exercises, especially relaxation techniques, and (3) Terabot’s physical features, such as its appearance and voice. Patients highlighted contemporary nature of the tool, as well as its limitations, including technical issues (such as freezing and stuttering), rigid communication, insufficient personalization of exercises, and a lack of empathy. Facilitators’ observations confirmed these limitations while also noting patient engagement and openness to technological innovation.

This preliminary study suggests that Terabot may be an acceptable and potentially useful tool for supporting therapeutic work with patients with schizophrenia. However, the findings must be interpreted with caution, given the study’s exploratory design, the time-limited therapeutic intervention, and technical limitations. Enhancements in relational responsiveness, dialogic flexibility, and the range of therapeutic exercises are necessary before drawing more definitive conclusions. Moreover, controlled clinical trials are required to evaluate Terabot’s effectiveness and its capacity to foster a stronger therapeutic alliance.