Researchers from the University of Westminster, with support from University College London (UCL), have published a new paper that explores the use of chatbots and virtual assistants for health conditions.
The paper – entitled ‘Health chatbots acceptability moderated by perceived stigma and severity: a cross-sectional survey’ was published 8 December in SAGE Digital Health journal.
Between May and June 2019, the researchers conducted an online study, advertised using Facebook, for UK citizens. A total of 237 participants completed the study with the majority female (73.4%), aged over 45 years old (65.0%), and educated with a degree or higher (54.9%).
Using a factorial simulation experiment, the researchers’ analysis suggests that for “health conditions that are perceived as highly severe such as cancers, people would be less likely to use AI algorithms”, however they suggest “patients were more inclined to discuss highly stigmatising conditions with AI chatbots than a GP”.
The researchers aimed to assess how perceived stigma and severity of various health issues are associated with the acceptability for three sources of health information and consultation: an automatic chatbot, a GP, or a combination of both. They also aimed to further understanding and research into the use and effectiveness of AI symptom checking technology for differing conditions.
One of the conclusions from the researchers, included: “For policymakers and digital service designers to recognise the limitation of health chatbots”, and they called for “future research about the use of AI in healthcare to establish a set of health topics most suitable for chatbot-led interventions and primary healthcare services”.
Dr Tom Nadarzynski, lead author of the study from the University of Westminster, said: “Many AI developers need to assess whether their AI-based healthcare tools such as symptoms checkers or risk calculators are acceptable interventions. Our research finds that patients value the opinion of healthcare professionals, therefore implementation of AI in healthcare may not be suitable in all cases, especially for serious illnesses.”
Read the full paper in the SAGE Digital Health journal. The authors included: Oliver Miles, Robert West and Tom Nadarzynski.
