NEW DELHI: Over the past week, a viral message on WhatsApp has been creating panic amongst users about the possibility of the novel coronavirus being aerosolized. The message consists of a legitimate news link, but has been grossly misread by people thanks to the lack of scientific understanding and facts.
Despite efforts, misinformation about the virus has spread rapidly on various platforms. And while people can try to battle them one message at a time, automated systems can at the very least equalise the possibility of gaining dependable information through such platforms.
The world’s largest instant messenger tied up with the World Health Organisation (WHO) to make a chatbot service that answers users’ questions on Covid-19.
The Indian government has also set up a chatbot called MyGov Corona Helpdesk, which works through WhatsApp to provide verified information to users in India. The chatbot was created by Reliance Jio-backed Haptik Technologies.
Similarly, Goa has rolled out a chatbot--Cobot-19--in collaboration with healthcare startup Portea Medical and Bengaluru-based automation firm Verloop. The bot can respond to queries on Covid-19 in English and Konkani, with Hindi support in the works.
Chatbots like Cobot-19 have been trained using basic machine learning (ML). Gaurav Singh, founder of Verloop explained that there are two parts to the process. First is offline training that happens in the backend. Second is online real-time usage.
The first step of training is intent mapping where a chatbot is trained on what novel coronavirus is and how it spreads. Intent is further divided into categories. Due to this training, the bot tries to match the intent behind a user’s questions with “intent categories" it has been trained on before answering said question.
Unlike more sophisticated bots like the ones used for customer service on e-commerce websites, Cobot-19 chatbot is very basic as there is only one level involved. Singh added, “But ideally if a user asks a question, it is supposed to figure out what the answer is. If the information is not available, the system will collect the question it could not answer and we will take them to doctors from Portea Medical. Once we have the accurate response we train the chatbot to respond to those questions if they are asked again."
The primary reason for this is that customer service bots are created based on years of data, whereas the the new strain of coronavirus is only a few months old for the world. Adequate data sources aren’t available yet, so the focus is on answering questions from the most accurate sources available while also understanding what users want to know.
The ML enabled Q&A in Cobot-19 where users can ask specific questions on COVID-19 and get answers to them is not working as of now. Currently, the chatbot can only fetch answers to a list of pre-defined questions highlighted by a number. If the user types a number linked to a question, the chatbot will provide an answer to it with links to explanatory videos.
In case of Haptik’s chatbot, the scope of conversation is quite large but it operates within a boundary of defined questions and answers, said Aakrit Vaish, the company’s CEO.
Haptik’ decision to not use artificial intelligence (AI) or ML was intentional. “We did not want to get anything wrong in the responses. The reasoning here is that with AI, there is always a chance of getting 1-2% incorrect responses. Given the sensitive nature of the information being shared on this bot, we did not want to take that risk," Vaish said.
“People are in a situation of panic and trying everything that is possible. Messages about unproven home remedies are being shared. In the absence of any scientific evidence backing them up, they are nothing but rumours and should be avoided," warned Dr. Laxman Jessani, consultant, Infectious Diseases at Apollo Hospitals.
He said that the more scientific knowledge people have on the disease, better it is for everybody. However, information provided by chatbots should be authentic and reliable too.
In the case of WHO’s chatbot, the information would come from the organisation’s own records. Verloop’s Cobot-19 relies on sources like the WHO and the US Centres for Disease Control and Prevention (CDC), while Portea Medical will review the information adding an additional layer of verification.
The primary objective behind these chatbots is to leverage messaging platforms being used to spread most of the misinformation on the disease and increase the share of dependable information, which is available quickly and easily. WhatsApp has over 2 billion users globally, making it the perfect information delivery mechanism in this case.