Can tech make doctors aware of their unconscious biases?

Researchers aim to develop technology that alerts clinicians when their verbal and nonverbal cues affect interactions with patients.

Researchers at the University of Washington School of Medicine will lead an effort to develop technology that alerts clinicians when their unconscious biases might be affecting how they interact with patients.

The five-year, $2.8 million project is being funded by the National Library of Medicine. Andrea Hartzler, associate professor of biomedical informatics and medical education, is the lead investigator. 

“We all have unconscious biases we aren’t aware of, called implicit biases,” Hartzler said. “Our goal is to develop a training tool that can make clinicians aware how these biases may be affecting their behavior while they’re meeting with patients and help them improve their care.”

The researchers include Janice Sabin, an associate professor of biomedical informatics and medical education and an expert on implicit bias, and Wanda Pratt, a professor at the UW Information School who works to develop technology tools to help patients gather and manage their personal health information. They will work with Nadir Weibel, an associate professor of computer science and engineering at the University of California, San Diego (UCSD). He designs computer systems that monitor and analyze how people behave in different social situations.

pictures of Andrea Hartzler, Janice Sabin and Wanda Pratt
From left, Andrea Hartzler, Janice Sabin, and Wanda Pratt are leading the UW's efforts in this research.

“We have explicit belief systems that we are aware of and often openly express, and we think we follow those beliefs at all times,” Sabin said. “But we have subconscious attitudes and beliefs — implicit biases — that we’ve absorbed from our upbringing and culture that can drive our behavior without us knowing it.”

Previous research has shown that healthcare providers often treat certain groups of patients differently.  Emergency room doctors have been shown to treat chest pain in women less urgently than they do in men, resulting in a delayed diagnosis of heart disease and higher in-hospital death rates among women.  

Similar studies show that pediatricians are less likely to prescribe medications for post-surgical pain to African-American children than to white children. And numerous studies have shown that healthcare providers view obese patients as being undisciplined and lacking willpower, leading these physicians to assume obese patients are less likely to comply with their recommendations.

UW researchers have long led the study of unconscious bias, developing the Implicit Association Test, a widely used tool.

In the new project, UW and UCSD researchers hope to adapt an approach called social signal processing. In this approach, verbal and nonverbal cues will be monitored and analyzed to understand the nature of social interactions. These cues include the tone and speed of speech, body language, gestures and facial expressions.

Data suggests this study of doctor-patient interactions effectively detects when physicians’ cues signal interest and warmth, for instance, by maintaining eye contact with the patients, or signal a lack of interest by looking away or by frequently interrupting the patient.

Most of these studies, however, have relied on video recordings of doctor-patient interactions. This requires researchers to watch each video and note the cues that may affect the relationship – a laborious, time-consuming process.

Researchers in the UnBIASED (Understanding Biased patient-provider Interaction And Supporting Enhanced Discourse) project hope to develop an automated system, using technology developed in the Weibel Lab. It not only records audio and video of a social interaction, but also automatically documents eye direction, posture and other nonverbal social cues.

Ultimately the goal is to devise a system that provides clinicians real-time feedback that indicates how they are relating to different types of patients, Weibel said.

“We may not be able to tell you that you’re biased against a particular group, but we will be able to tell you that you behave differently when you’re interacting with people from this group,” he said. “That would be enough of a cue for you to reflect and realize that there might be something you need to do to change how you behave and interact.”

Awareness is growing about implicit biases and how these negatively affect many aspects of society, including medicine, Pratt said.

“We want to see if there are some ways we can fix that, if there are ways we can make people aware of these biases in the clinic at the moment when they may be affecting the quality of healthcare.”

For details about UW Medicine, please visit

Tags:healthcare equity

UW Medicine