Researchers from the University of Twente, MST and Politecnico di Milano conducted a pilot study to explore whether a GPT-controlled social robot can support patients with medical information in a hospital setting. The first results are cautiously positive: patients and caregivers accept the technology. The research focuses on technical, organizational and ethical feasibility.
Health care systems are under increasing pressure. Due to staff shortages and a growing demand for care, the accessibility of care is under pressure. Clear and effective patient communication remains essential, especially in chronic conditions. Digital technology can help with this, but it also raises questions about reliability and trust.
The paper is published in the journal Frontiers in Digital Health.
Exploring AI with a ‘face’
In that context, scientists from the University of Twente, together with health care professionals from Medisch Spectrum Twente, investigated whether a GPT-controlled social robot can inform patients about their condition and treatment. The system consists of a physical social robot with a face, facial expressions and speech capabilities. It can answer questions through natural conversation with the patient.
The study indicates that this physical presence was accepted by both patients and caregivers. Patients experienced the conversation as accessible and pleasant. “This should not be interpreted as evidence that care quality improves,” emphasizes lead researcher Jan-Willem van ‘t Klooster. “We investigated whether such a system can function in practice, not whether it already improves care.”
Tested in the clinic, not just in the lab
The research began with a lab study, but was then tested in the hospital’s daily practice. A total of 21 patients with osteoarthritis and 7 health care professionals spoke with the robot. Both patients and health care providers rated the system positively in terms of ease of use and acceptance. According to Van ‘t Klooster, this is important: “Acceptance is a first step. Then you can investigate whether such a technology really contributes to better information provision, therapy adherence or time savings for health care providers.”
A crucial part of the research was the way in which AI was used. The GPT technology was not given free access to the internet, but was only allowed to use information from pre-approved, doctor-validated medical websites. In this way, the researchers wanted to limit the risk of incorrect or fabricated answers (hallucinations).
“The debate is often about whether you should use AI in health care,” says Van ‘t Klooster. “We show that it’s mainly about how you set it up. By setting clear boundaries, control remains in the hands of health care professionals.”
Teamwork between technology, care and behavior
The project was very much a team effort, bringing together expertise from behavioral sciences and clinical practice. In addition to researchers from the University of Twente, health care providers, designers and international partners were also involved.
“It is precisely this collaboration that makes this kind of research possible,” says Van ‘t Klooster. Follow-up research remains necessary, including knowledge transfer and long-term use. An investigation into the language level to be used is currently taking place.
More information
Jan-Willem J. R. van ‘t Klooster et al, A GPT-reinforced social robot for patient communication: a pilot study, Frontiers in Digital Health (2026). DOI: 10.3389/fdgth.2025.1653168
Key medical concepts
Clinical categories
The content is provided for information purposes only.