Artificial intelligence (AI) has the potential to have a more significant impact on healthcare in the future compared to current online symptom checkers. However, for these “AI doctors” to enhance patient satisfaction, they might need to adopt a more personalized approach than traditional human doctors. A recent study indicates that when AI doctors remember detailed social information about their patients, it leads to higher satisfaction levels among those patients, provided they are given control over their privacy.
According to a study conducted by researchers at Penn State, AI doctors can enhance patient satisfaction. The study revealed that when AI systems remember more social details about their patients, satisfaction levels increase, but this is effective only if patients can manage their privacy settings.
The team’s findings were published in the journal Communication Research.
“People often perceive AI doctors as impersonal and uniform,” stated S. Shyam Sundar, who is the Evan Pugh University Professor and the James P. Jimirro Professor of Media Effects at Penn State. “Our research indicates that it’s crucial for these AI systems are designed to not only discuss a patient’s medical background but also to personalize their interactions by recalling non-medical details, like the patient’s job and interests. Ultimately, AI doctors could manage telehealth appointments as effectively as human physicians.
To investigate if a doctor’s understanding of a patient’s personal or medical background enhances patient satisfaction, researchers engaged 382 online participants in conversations with a medical chatbot during two sessions spaced approximately two weeks apart. Participants believed they were interacting with either a human doctor, an AI doctor, or an AI-assisted physician.During the initial appointment, the “doctor” — actually a programmed script designed by researchers for uniformity — engaged with patients on various subjects including their diet, fitness levels, lifestyle choices, sleep patterns, and mental well-being. They also inquired about personal details such as the patient’s job, family relationships, eating habits, and preferred activities. After this discussion, the doctor provided general suggestions regarding nutrition, exercise routines, and strategies for mental health care.
At the second appointment, the doctor either remembered details from the patient’s medical or personal history or prompted them to remind him of those specifics. Following this exchange, similar recommendations were provided again.
During the initial appointment, health professionals provided guidance and offered 50% of patients the chance to have their visit documented and their details saved on an online platform at the end of their final session. Following this, participants filled out an online survey to evaluate their satisfaction with the service.
The study revealed that patients rated AI doctors more favorably when these virtual doctors remembered personal social details, as long as they ensured privacy controls were in place before wrapping up the appointment. In contrast, human doctors did not need to remember either social or medical details for patients to perceive a strong connection with them.
“When an AI doctor remembers details about a patient’s social life, it makes the patient feel that the doctor is more attentive to their individual needs, which boosts satisfaction. However, this only works well if the patient has control over their privacy,” explained Cheng “Chris” Chen, who is the lead author and an assistant professor of communication design at Elon University. Chen completed his doctorate in mass communications at Penn State. “This finding was unexpected because AI systems usually treat all information equally; however, patients interpret it differently. They view it as a sign that the doctor is making an extra effort to remember personal details.”
Sundar mentioned that the effort required from patients is significant.
“Patients prefer having control over their privacy when interacting with AI systems,” he stated. “As long as I can manage my own data, I welcome your insights into my social life and appreciate the effort involved.”
The researchers believe this study highlights important considerations for designing AI tools in healthcare.
“Remembering details about patients’ social lives can enhance satisfaction, foster compliance, and lead to improved health results,” Cheng explained.
Co-author Joe Walther, who holds the Bertelsen Presidential Chair, emphasized these points further.
As a distinguished professor of communication at the University of California, Santa Barbara, the research delves into broader issues concerning the nature of knowing someone or the feeling of being understood by others or even by technology.
“When I tell a student that her assignment was superior to many others, it’s merely a numerical comparison,” Walther explained. “However, if I mention that her work has improved since earlier in the semester, she realizes that I truly understand her; she’s not just another statistic. This is similar for patients: Am I merely defined by my most recent lab results, or do they see me as an individual?”
