Chatbots provide millions with COVID-19 information every day, but they can be improved - here’s how

The coronavirus outbreak occurred in our modern, highly connected, information-dense world. Yet, dissemination of accurate, up-to-date information about the spread of the disease remains a challenge.

Traditional media (radio, TV and print channels) have shrinking audiences and the best require subscriptions for access. Several local and regional authorities have made text alerts available, but these are available just to those who register and aren't available in every locality. Younger audiences prefer social media over traditional channels but many social media channels have been challenged by fake news and privacy breaches, and aren’t always fully reliable.

Enter Conversational AI, also known as Chatbots, a young technology that allows information access through text or voice-based interaction, that is proving its worth during the coronavirus crisis and showing how this new communications channel can be leveraged in the years ahead by a range of groups and institutions.

How chatbots are fighting coronavirus

Chatbots have been a natural choice for disseminating health information during the coronavirus crisis. Advances in Natural Language Processing (NLP) have enabled conversational AI technologies and widened their reach, leading to tools such as Siri, Alexa, and Google Home that are part of many consumers' every day lives. The intuitive interface of chatbots presents a low-friction approach to disseminate critical information to vast populations. And Chatbots, like websites, are available 24/7.

Additionally, chatbots offer strong potential for curated information. The information can be customized to the needs and symptoms of the individual. Response to specific questions can be provided in an interactive manner, more rapidly than traditional online search methods. The information is also adaptable to local guidelines and regulations, based on the location of the user.

Many companies and organizations are leading the charge in deploying chatbots to provide COVID-19 information. The two most authoritative voices of the pandemic, WHO and CDC, have also included chatbots in their websites to provide up-to-date information to billions on the spread of the disease and its symptoms. Many governments are also launching chatbots to provide validated information to their citizens.

Anyone with access to the chatbots can do so from a range of devices (online computer, smart phone, or analogue phone in some cases). Widespread adoption of chatbots for COVID-19 information can also reduce the burden on hospital call centres. The chatbots’ interactive symptom checking feature can potentially lower the volume of cases in urgent care and emergency care.

Challenges to address
While chatbots have their benefits, there are some deficiencies that must be addressed to maintain the public’s trust in the information the chatbots provide. For instance, it is possible to receive inconsistent responses to symptom checking from the multitude of chatbots for COVID-19. Presented with identical symptoms, some chatbots provide inconsistent results. Some chatbots urge the user to get immediate care, while others suggest taking rest at home.

That inconsistency could stem from a variety of sources. Some chatbots might rely on just one source of information. In other chatbots, the back-end repository that chatbots retrieve information from might not be frequently updated.

Also, most chatbots deployed today have a relatively simple backend that looks up information by walking through a “decision tree” based on the users' input. If a chatbot’s decision tree is not updated frequently, suggestions made by the chatbot will become obsolete when expert recommendations on the pandemic change.

In the long run, when we emerge from this crisis, chatbots are likely to become digital portals for interactive healthcare, helping patients find a doctor or service, schedule appointments, facilitate symptom checking, conduct triage in emergency care, prepare for procedures and follow post-discharge instructions. They can also act as virtual assistants to physicians.

While these uses can result in better care management and customer engagement, several issues could arise, such as miscommunication between chatbots and users, customer misperceptions of chatbots, incorrect/poor guidance, wrong diagnoses, or failure to achieve timely interventions.

To be sure, broader ethical concerns could also apply. Those concerns include: chatbots pretending to be human; mental health chatbots that attempt to fully replace therapists and expert practitioners; data privacy violations; data bias that results from training on data from one population and deploying in another; and lack of continued monitoring to check the advice given.

To address these and other ethical concerns, clear and effective governance frameworks are needed to guide chatbot developers, platform providers, and healthcare systems. These frameworks must take into account a range of factors, including validation/accreditation, performance assurance, patient expectation management and access, legality, privacy, and security. Further, for the use AI in chatbots, the frameworks must address transparency, bias, fairness, and data privacy and data rights issues.

The COVID-19 pandemic is an accelerator for chatbot technology, helping people around the world get more and more comfortable with leveraging this tool for healthcare. As we move beyond the pandemic, the adoption of chatbots in broader healthcare applications will continue to grow. As they do, public and private stakeholders must come together to create governance frameworks that maximize these benefits while minimizing risks.