How Artificial Intelligence is enabling public opinion to shape the pandemic response in Scotland
Professor Amir Hussain, Edinburgh Napier University, explains how he is harnessing Artificial Intelligence (AI) to better understand the impact of COVID-19.
Our response to COVID-19 is shaped by how we track the pandemic.
We’ve learnt a lot by tracking the numbers - following the rate of infections, deaths and recoveries, and the spread of the virus across locations.
What’s missing from these numbers is an understanding of behaviours, traits, attitudes and opinions, which are harder to measure, but could have a huge influence on our response.
Tracking evolving trends in public opinion
Our project, which is co-led by BREATHE Director Professor Aziz Sheikh, is a Scottish government-funded initiative to better understand how COVID-19 public health policies (such as physical distancing, use of face coverings and travel restrictions) have affected people’s daily lives.
By examining people’s attitudes and opinions towards these interventions, we can develop a better understanding of their potential short- and longer-term impacts on public health and wellbeing, and help policymakers make decisions about how to help people during this challenging time.
Analysing public social media posts, protecting privacy
In particular, we’re interested in using social media to follow how people are feeling in real time. To do this, we are building a web-based dashboard that uses ‘expert-in-the-loop’ Artificial Intelligence (where humans are involved to improve accuracy) to analyse public social media posts, including on Twitter and Facebook.
Protecting the privacy of individuals is key. Our research will not involve any social media post that carries an expectation of privacy, such as where the user has made their account private, or where the post was made in a non-public group setting.
We only collect data which is absolutely necessary for our analysis, and apply privacy-protecting techniques to ensure the identity of individuals is protected at all stages.
BREATHE provides a secure environment to store data generated by the study, in partnership with Trusted Research Environment, SAIL Databank.
Supporting current and future decision-making
AI is ‘trained’ by being fed large amounts of data from which it can learn how to analyse and make predictions.
In this case, our AI tools have been developed using data in the form of feedback to a Scottish Government public conversation on COVID-19. By assigning topics and themes within these comments, the AI can begin to learn what might group a statement under a particular area. The involvement of human experts in this process ensures the AI is more accurate and trustworthy.
Following development, we are now working with the Scottish Government to use our AI-enabled tools to support a rapid analysis of data from a new public dialogue exercise, which completed on 11 October 2020. Here, our AI tool provides a quick route to assess the comments (much faster than the alternative of going through them manually) and has the potential to broaden the analysis to include public social media posts alongside the direct feedback received.
We’re continuing to work closely with the Scottish Government as the dashboard develops, ready to support future public engagement plans and providing further opportunities for the public to have their say.
In particular, we are keen to explore how the tool might enable access to seldom heard voices on these issues, which is difficult to do through other routes in current circumstances.
Through BREATHE, we’re in the process of recording the study metadata (which provides basic information about the data we are collecting) on the Health Data Research Innovation Gateway, and are enabling access to the statistical code for analyses to other researchers. This will allow others to replicate and build on our work.
A valuable resource
Our aim with the project is to boost understanding of the direct and indirect impacts of the pandemic on public health and wellbeing. We’re supporting decision makers by informing the development of appropriate interventions, such as public health education measures to dispel identified myths and misinformation. The project could also help inform the development of exit strategies from current and future lockdowns.
We believe this will be a valuable resource to help us make better decisions around how to react to public health crises, as individuals and also as a nation.