TRAC: COVID - TRust And Communication: a Coronavirus Online VIsual Dashboard

Lead Research Organisation: Birmingham City University
Department Name: ADM School of English

Abstract

The COVID-19 crisis has required mass communication and public understanding on an unprecedented scale. During this time there has been a proliferation of online discussion, news sharing and emergence of 'information sources' concerning COVID-19. Such proliferation has raised concerns about the potential dangers of dis/misinformation. As yet, however, neither the extent of the issue nor the sources of information have been studied in detail. Furthermore, despite communication being at the heart of COVID-19 public health efforts, there has been a surprising lack of input from linguistic experts.

This project aims to build a large-scale dataset of Twitter posts, which will be made available via an open-access online dashboard incorporating intuitive visualisations. The dataset will be novel in capturing not just the content of tweets, but also the content of web-pages shared in the tweets. Drawing on automated corpus linguistic methods and social network analysis, the dashboard will uncover the multi-layered content of shared information (original links, tweets, replies, retweets), alongside a deeper understanding of the online networks through which (mis)information is shared.

To demonstrate the applicability of our novel approach to a wide range of stakeholders, the methodology and dashboard will be validated through two case studies, each focussing on a potentially dangerous area of miscommunication relating to COVID-19. These case studies will approach the problem from a linguistic perspective, examining the clarity and reception of official messaging and the trustworthiness of information sources.
 
Description The TRAC:COVID project was successfully completed on schedule, overcoming the challenges presented by the short time span and the need to run the project entirely online during the pandemic. As detailed in the other sections of the submission, we have built a large corpus containing over 80 million UK tweets as well as a novel open-access 'dashboard' to allow other researchers and the general public to search this corpus (https://www.traccovid.com/). We have produced two case study reports, designed as proof of concept for the dashboard and to disseminate our findings to stakeholders (available in full at https://www.traccovid.com/traccovid/reports). Key findings are as follows:

Case Study 1: Public perception of the government management of the pandemic
Conclusions in relation to public perception of government handling of the pandemic:
1. Strong support for safety measures, incl. need for stronger measures
2. Widespread criticism of non-compliance
3. Views associated with 'covid-deniers' still in minority but increasing polarisation of public opinion

Our recommendations for government messaging:
1. Follow plain language principles and avoid using bureaucratic language
2. Use short sentences and everyday vocabulary
3. Provide enough detail and ensure the details included are correct. Avoid being vague
4. Ensure there is only one communicative function per message
5. Address the audience in a clear way
6. Be inclusive of people from different backgrounds
7. Avoid posting too many messages which do not provide much information and only link to
external content
8. Develop a strategy for addressing common misconceptions and emerging conspiracy theories
9. Involve experienced applied linguists in the process of designing public communication
strategies and analysing their effectiveness

Case Study 2: Misinformation, authority and trust
1. Majority of tweets about COVID-19 vaccines either do not contain or are critical of vaccine misinformation
2. Vaccine misinformation exists within a wider web of misinformation and conspiracy theories, where attempts
are made to undermine confidence and trust in vaccines, health professionals and policymakers
3. Some vaccine misinformation contains language related to known conspiracy theories but other forms are
novel, subtle, evolving, and designed to circumvent automated moderation systems
4. Vaccine misinformation is communicated in many forms so there is no 'silver bullet' to detect and prevent it
5. The ongoing role of expert human analysts in interpreting these linguistic behaviours is therefore crucial
Exploitation Route We submitted Case Study 1 to the House of Commons Public Accounts Committee in response to a call for evidence on "Initial lessons from the government's response to the COVID-19 pandemic". Our evidence was cited in the resulting report produced by the Committee (paragraph 22 at https://committees.parliament.uk/publications/6954/documents/73046/default/). Case Study 1 was also forwarded by our project steering committee member Professor Maxine Lintern to the Deputy Chief Nursing Officer for England, Professor Mark Radford, and Deputy Chief Medical Officer for England, Professor Jonathan Van-Tam in response to a request for ore details of our project. In addition, we have presented both case studies and the dashboard to several workshops organised by other UKRI COVID-19 project teams, published an article in The Conversation, and recorded a podcast to discuss our findings (details given in other sections). Our open-access dashboard is available for use by other researchers and by the wider public.
Sectors Digital/Communication/Information Technologies (including Software),Healthcare,Government, Democracy and Justice
URL https://www.traccovid.com/
 
Description We published two reports on our website at https://www.traccovid.com/traccovid/reports. Our report on the public response to government messaging was cited by the House of Commons Public Accounts Committee in their report on 'Initial lessons from the government's response to the COVID-19 pandemic', available at https://committees.parliament.uk/publications/6954/documents/73046/default/. This report said (p.17) "Written evidence from Birmingham City University, which analysed a large body of government's and public health bodies' Twitter messages relating to COVID-19, identified a wide range of language-related issues in communications. These included: messages without specific content; a lack of clarity about who messages were directed to; the use of long sentences with complex vocabulary, grammar and syntax; and issues which could raise ambiguity and confusion and can make government's messages less likely to be understood, less likely to engage a wide audience, and more likely to elicit negative reactions and to exclude some intended recipients (for instance, through the use of the term 'house', which designates a specific type of abode, rather than the more general term 'home')".
First Year Of Impact 2021
Sector Government, Democracy and Justice
Impact Types Policy & public services
 
Description Citation in House of Commons Public Accounts Committee report
Geographic Reach National 
Policy Influence Type Contribution to a national consultation/review
URL https://committees.parliament.uk/publications/6954/documents/73046/default/
 
Title Project dashboard 
Description Public dashboard on the project website for reviewing trends in Twitter data relating to COVID-19 in a user-friendly manner, requiring no specialist knowledge. The system shows word/hashtag frequency over time, word co-occurrence and clustering, and frequently shared external links. The dashboard was developed by Co-I Matt Gee using the ElasticSearch database system and Plotly Dash visualisation library. The system analyses 84,138,394 UK tweets covering the period January 2020 to April 2021, downloaded using the Twitter Historical Powertrack API. 
Type Of Technology Webtool/Application 
Year Produced 2021 
Open Source License? Yes  
Impact A key impact is that findings generated by the project dashboard were cited by the House of Commons Public Accounts Committee in their report on 'Initial lessons from the government's response to the COVID-19 pandemic' (July 2021), available at https://committees.parliament.uk/publications/6954/documents/73046/default/. This report said (p.17) "Written evidence from Birmingham City University, which analysed a large body of government's and public health bodies' Twitter messages relating to COVID-19, identified a wide range of language-related issues in communications. These included: messages without specific content; a lack of clarity about who messages were directed to; the use of long sentences with complex vocabulary, grammar and syntax; and issues which could raise ambiguity and confusion and can make government's messages less likely to be understood, less likely to engage a wide audience, and more likely to elicit negative reactions and to exclude some intended recipients (for instance, through the use of the term 'house', which designates a specific type of abode, rather than the more general term 'home')". 
URL https://www.traccovid.com/
 
Description Pandemic & Beyond Podcast 
Form Of Engagement Activity A broadcast e.g. TV/radio/film/podcast (other than news/press)
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Media (as a channel to the public)
Results and Impact Interview with Andrew Kehoe (PI) and Robert Lawson (Co-I) on the Pandemic and Beyond podcast (Episode 4, 23 June 2021)
Year(s) Of Engagement Activity 2021
URL https://pandemicandbeyond.exeter.ac.uk/media/podcasts/
 
Description The Conversation article 
Form Of Engagement Activity A magazine, newsletter or online publication
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Media (as a channel to the public)
Results and Impact Article by the project team reporting on process and findings: "We archived 84 million tweets to learn about the pandemic - each one is a tiny historical document". 7160 views as of writing.
Year(s) Of Engagement Activity 2021
URL https://theconversation.com/we-archived-84-million-tweets-to-learn-about-the-pandemic-each-one-is-a-...