Innovative solutions to tackle the coronavirus outbreak are popping up everywhere. From tracking and tracing technology to reshaping offices, once well-established social norms are being torn apart. One of the key areas that healthcare and government are investigating as a way of broadcasting public health messaging — and triaging potential coronavirus patients — is the use of chatbots.

The World Health Organisation (WHO) and US Centers for Disease Control have rolled out chatbots providing information about the coronavirus, as has the UK government through WhatsApp. The city of Buenos Aires has launched Boti, a chatbot based in WhatsApp, as a way to answer residents’ common questions about the coronavirus. Among the pre-programmed questions you can pick from is “What is coronavirus?”.

Ensuring that any information imparted through such chatbots is correct and up to date becomes ever more important when talking about a life and death situation such as the spread of coronavirus

In all, more than 1,000 coronavirus-specific chatbots have been launched on Microsoft’s chatbot creation programme alone, according to Venkataraman Sundareswaran, fellow in digital trade at the World Economic Forum (WEF), who has been studying their deployment during the Covid-19 pandemic. Most of those are used like Boti and the WHO to provide information, and as such are more dumbed down than the most cutting-edge chatbots, which traditionally try and take inputs from users, understand what they’re asking, and infer intent while having a proper conversation.

“Covid-19 chatbots are somewhat limited in their capacity, and are watered down in their capabilities — with good reason,” says Sundareswaran. “You can’t leave it to chance for the chatbot to wander off in some direction.”

That’s particularly important when considering just how troubling the medical community is finding the coronavirus. The virus itself, and the disease it causes, is still a relatively unknown quantity, and medical advice in different countries is changing by the day. Ensuring that any information imparted through such chatbots is correct and up to date becomes ever more important when talking about a life and death situation such as the spread of coronavirus.

We have to be aware of the digital divide and digital literacy: some people will not have the technology to use the chatbot

Because of that, many of the chatbots being used right now are little more than glorified web searches, says Sundareswaran. “You ask a question, it gives you five answers, and if you pick one, it gives you the next five. You’re navigating pre-defined answers.”

But those answers could provide vital information to governments and public health services looking to identify the spread of coronavirus within their communities, and helps them understand how better to target responses.

The challenges, though, are numerous. For one thing, having a decision tree-based chatbot — essentially a dumb bit of computer logic — is easy to dupe, or worse, for people to reverse engineer. As with all health advice on the internet, people will often plug in symptoms or provide answers they don’t necessarily agree with to see how the answer changes, and then take on that answer as the one they want.

There’s another issue that many governments rushing to implement chatbots as part of their coronavirus response are overlooking: not everyone feels willing or is even able to interact with them. “We have to be aware of the digital divide and digital literacy: some people will not have the technology to use the chatbot,” says Tom Nadarzynski of the University of Westminster, who has looked at chatbots’ role in e-health.

Empathy is an important issue when talking about a life and death issue

Studies show that people are willing to use chatbots as a way of being signposted to the correct health service, but are sceptical about sharing personal information with technology. And as with any tech-enabled solution in health, just because something may appear more efficient, it doesn’t mean it’s more effective.

“When the conditions people seek information for are serious, the only way to manage that anxiety is when they talk to health professionals who tell them they have nothing to worry about,” says Nadarzynski. “That doesn’t happen with a chatbot.” Empathy is an important issue when talking about a life and death issue.

The extraordinary situation we find ourselves in may not be making much difference to chatbot adoption and acceptance now, but it could have a much wider impact in the future, believes Sundareswaran’s colleague Kay Firth-Butterfield, head of artificial intelligence and machine learning at WEF. “One of the things you might see is people become more and more used to interacting with chatbots because of Covid,” she says. “What we think we’ll see is more deployment of chatbots because of that familiarity.” — Chris Stokel-Walker

(Picture Credit: Unsplash)