Conversational AI to rebut COVID-19 rumours, myths, and misinformation
The Issue
The impact of the COVID-19 pandemic on all countries is unprecedented. In addition to the direct health impact, the pandemic has broader social and economic impacts that could have lasting effects on the lives of children and young people.
As the world battles the COVID-19 pandemic, an epidemic of misinformation has been circulating through different communication channels and communities, preventing people from accessing and identifying accurate information from official sources. The spread of misinformation in times of a global health crisis can mean that children and young people are unprotected and uninformed about the dangers of the disease and how to keep themselves safe.
Why Does it Matter?
Article 17 of the UN Convention on the Rights of the Child states that every child has the right to reliable access to information from a variety of sources and that children should be protected from materials that could harm them. In addition, the Sustainable Development Goals are working towards good health and well-being, quality education, reduced inequalities and peace, justice and strong institutions. Combating misinformation in the context of COVID-19 can help to make progress across these areas.
Misinformation, rumours, myths, and fake news can have debilitating and real-world effects. There is a significant concern that the repetition and amplification of rumours may refute the significance of true advice and policy. In what is already a time of fear and uncertainty, people are looking to the media for answers. If what they are finding are rumours or untrustworthy advice, there is a real worry that children could be putting their lives at risk. It is vital to ensure that this misinformation is flagged and dealt with, and that children are directed towards reliable, factual and accurate information about COVID-19.
Our Project
On 17th February 2020 U-Report launched a COVID-19 Information Chatbot to provide critical information to communities and receive feedback to inform UNICEF’s response. The platform has reached over 3.5 million people, including refugees and migrants, and has been deployed in 50 countries with over 6 million bot interactions. Through communication channels like SMS, Viber, Facebook Messenger, and WhatsApp, users are receiving life-saving information.
The objectives of our project are to increase UNICEF’s ability to identify and track rumours and myths around COVID-19 through the U-Report Chatbot, and through our partnership with Alana AI , to develop more sophisticated tools for responding to and debunking rumours submitted through the chatbot.
The key phases in the initial stages of the project are:
1. Development of a large dataset of real-world user statements involving myths and mis-information about COVID and vaccines.
2. Using topic models as well as distributional semantics for textual data to cluster user inputs into groups which are similar in meaning in order to create a taxonomy of rumours.
3. Development of the automatic rumour classifier tool based on this taxonomy. This will be able to detect different types of rumours as well as outputting a confidence score in its classification, allowing low-confidence examples to be passed to human experts for labelling.
4. Development of additional conversation flows for the UNICEF chatbot. This could be follow-up questions, ‘Do you think that might be true?’ as well as the ability to understand user answers to such questions.
The initial work will be carried out for English language inputs. Further phases of the project may look to extend the system to cover additional languages such as French, Spanish and Arabic.
You can find out more about our partners from Alana AI here.