Intended for healthcare professionals


Using artificial intelligence to assess clinicians’ communication skills

BMJ 2019; 364 doi: (Published 18 January 2019) Cite this as: BMJ 2019;364:l161
  1. Padhraig Ryan, research fellow in health informatics1,
  2. Saturnino Luz, chancellor’s fellow2,
  3. Pierre Albert, PhD candidate2,
  4. Carl Vogel, associate professor in computational linguistics3,
  5. Charles Normand, Edward Kennedy chair of health policy and management1,
  6. Glyn Elwyn, professor4
  1. 1Centre of Health Policy and Management School of Medicine, Trinity College Dublin, Dublin, Ireland
  2. 2Usher Institute of Population Health Sciences and Informatics, University of Edinburgh, Edinburgh, UK
  3. 3School of Computer Science and Statistics, Trinity College Dublin, Dublin, Ireland
  4. 4Dartmouth Institute for Health Policy and Clinical Practice, Lebanon, New Hampshire, USA
  1. Correspondence to: G Elwyn glynelwyn{at}

Most doctors have never had their communication skills formally assessed and do not know how they compare with their peers. Glyn Elwyn and colleagues explain how AI might facilitate this and help improve interactions with patients

Artificial intelligence (AI) has been defined as the capability of a machine to mimic intelligent human behaviour.1 To limited extents, AI has arrived. We can give orders to our smartphones and talk to devices such as smart speakers and ask them to update us about the day’s weather and traffic. They don’t perform perfectly, yet the ability to understand and respond to human speech is advancing rapidly. How long might it be before speech recognition, machine learning, and other developments in AI will offer tools to medicine, and how might those tools offer insights into what happens between clinicians and patients?

Novel ways to manage practice tasks

AI research involves the development of “intelligent” computer agents. Traditionally, AI encoded existing knowledge about the world and thereby relied on prespecified human expertise. The hard coding of information into AI algorithms was typically a lengthy process. An alternative approach, known as machine learning, relies less on prior assumptions and enables computers to develop algorithms based on repeated trials and errors. Although this still requires expert human input, the time to develop AI algorithms is now often much shorter. Machine learning is prominent in tasks such as creating the equivalent of eyesight for computers, enabling self driving cars for instance.

Machine learning has potential to have a big effect on medicine,23 and AI applications are beginning to emerge in healthcare. Many clinicians may prefer to use their voice rather than keyboard and mouse to interact with technology. Some, such as radiologists, already interact with digital systems using voice and physical gestures, and studies suggest important productivity gains,45 although …

View Full Text

Log in

Log in through your institution


* For online subscription