Welcome to the hospital and how can I assist you today?

Robots enabled by AI are working well with patients and medical staff in China

When patients arrive at the Anhui Provincial Hospital in eastern China, they are greeted by a humanoid robot that can answer non-emergency questions from visitors and patients in Mandarin and 22 Chinese dialects. 

Developed using machine learning from more than 53 professional medical books and a vast amount of patient data, the robot is ‘clever’ enough to answer questions about physicians’ schedules in 47 departments, and navigate to more than 600 locations within the hospital. 

Its ability to correctly answer the 260 most-frequently asked questions is an impressive 90.81%. And, at the end of each day, human reception nurses update its database to further improve its accuracy. 

With its big blue eyes and peaceful expression, Xiaoyi, as the robot has been dubbed by its creators, is the face of our medical future. At Anhui, it is just one example of artificial intelligence (AI) being used to reduce the nursing workload, and to help doctors with diagnosis and treatment. 

These sorts of innovations are changing the way we live and work, and exactly how they are doing so is becoming an important question for researchers, says  Shan-Ling Pan, a professor and deputy head (research) of the school of information systems, technology and management at UNSW Business School. 

At a time when work places around the world are increasingly adopting AI, we can only realise its true value if we understand how it connects with us socially and institutionally, says Pan.

“From media to curriculums within universities – we’re all embracing the potential and the opportunities that AI can afford. We are interested in the extent to which AI is having an impact on the way we work,” he adds. 

'They build rules and boundaries around the work they can delegate to the machines, and the work which requires unique human capacity'

SHAN-LING PAN

Mimicking humans

In their research paper,  Artificial Intelligence in Healthcare Robots: A Social Informatics Study of Knowledge Embodiment,  Pan and fellow researchers, L. G. Pee of Nanyang Technological University in Singapore and Lili Cui of the Shanghai University of Finance and Economics, examined how healthcare professionals are working with AI at Anhui Provincial Hospital, and have found that humans and AI (including software and hardware robots) are collaborating in professional knowledge work – analysis, diagnosis and problem solving – with AI mimicking humans in many ways. 

For example, an imaging diagnostic system now being used by the hospital integrates medical-grade display hardware and image processing software to visually analyse CT images taken of patients’ lungs, a job previously done by radiologists. 

The robot can identify pulmonary nodules with high sensitivity, analyse their surroundings and automatically classify them as malignant or benign and recommend diagnosis to human radiologists. 

So far, it has diagnosed about 11,800 CT images with an accuracy of 94.9%. The final diagnosis, however, is still made by human staff. 

The medical professionals use the robot to assist or augment their work, says Pan. 

“They build rules and boundaries around the work they can delegate to the machines, and the work which requires unique human capacity,” he  adds.

The intersection between humans and AI comes at the point of decision-making, Pan says. For example, the software robot takes in a patient’s basic physical details, processes them and makes a recommendation, but a doctor makes the final decision about what action to take. 

Job security

Importantly, the researchers found that despite predictions of large-scale job losses as a result of automation, staff at the hospital were not concerned about their job security. 

Instead, says Pan, most of the workers considered the robots to be complementing rather than competing with them. For example, the hospital uses an electronic health records system equipped with voice recognition that allows doctors to dictate their patient observations directly into the system rather than to a nurse who would have had to transcribe the notes. The dictation is turned into structured text entries with a 97% accuracy rate. 

“From the workers’ perspective, laborious work is reduced,” says Pan. “They welcome the automation and assistance offered by AI robots. They can now focus on more value-adding work.”

'They are competitors to human physicians to the extent that they are viewed as alternative physicians offering second opinions'

SHAN-LING PAN, L. G. PEE & LILI CUI

 Giving humans ‘superpowers’ 

In their paper, the researchers looked at how the knowledge embodied in robots working with  humans can be applied and how it can generate value. They conceptualised four ways humans interact with AI in the workplace.

First there are the 'co-operators', robots that reduce human physical work rather than cognitive effort.  Next are the 'collaborators', which at Anhui Hospital  help medical staff by easing the mental effort needed to make a diagnostic decision – for example, by making a recommendation about a CT scan. 

The third category, 'competitors', can complete work autonomously. The reception robot Xiaoyi, for example, can interact and handle inquiries just like a human nurse does. 

Interestingly, the researchers observed that many patients found it fun to use Xiaoyi, something not usually associated with human reception nurses. They noted it would be interesting to see how patients’ perception changed when the novelty wore off.

Finally, there are 'coopetitors', AI with the capacity to complete work autonomously and give humans 'super powers'. For example, Anhui Hospital is testing a medical diagnosis system that offers physicians a second opinion on a diagnosis. 

“They are competitors to human physicians to the extent that they are viewed as alternative physicians offering second opinions. At the same time, they cooperate and augment the work of human physicians by suggesting treatment options based on an enormous database of medical cases and knowledge,” the researchers say.

The researchers concluded that competitors and coopetitors had the greatest social impact. Multiple social actors were aware of their presence, and their human-like form rendered them as one of the social actors in knowledge work, rather than simply a technology tool used by human workers.