The Language, Data and Society (LANDS) Research Centre is at the forefront of various major, funded research projects relating to its core research themes.
Being Alone Together: Developing Fake News Immunity
This project is framed in the area of “crisis informatics”, the study of how (mis)information about COVID-19 is generated and flows over media platforms. The main goal is that of reverse-engineering the manipulation of information providing citizens with the means to act as fact checkers. We believe that fostering global digital activism constitutes a necessary means to fight the current info-pandemic. The majority of fact-checking and myth-busting sites (e.g. EUvsDisinfo, https://www.who.int/emergencies/diseases/novel-coronavirus-2019/advice-for-public/myth-busters) counter false narratives and news that have already become viral, unable to prevent their spread. Furthermore, AI techniques (http://www.fakenewschallenge.org) are currently not accurate enough to replace humans in generalised fact-checking. This is especially the case when the news does not contain fabricated information (disinformation), but it is framed in such a way that true evidence is used to draw false generalizations and evaluations (Wardle 2019), resulting in semi-fake news.
Leveraging NLP techniques for topic modelling and frame analysis (Das et al. 2010) we trace the topics and frames which characterize semi-fake COVID-19 news using FullFact (https://fullfact.org/) and the Coronavirus debunking archive built by First Draft (https://firstdraftnews.org/long-form-article/coronavirus-resources-for-reporters/) as benchmarks. We identify the fallacious reasonings in the sample and use the results to compile a set of guidelines about how to detect semi-fake COVID-19 news. These principles are operationalised in two chatbots to train citizens to spot misinformation. See https://fakenewsimmunity.liverpool.ac.uk for more information.
Elena Musi (Liverpool University)
Simeon Yates (Liverpool University)
Kay O’Halloran(Liverpool University)
Elinor Carmi (City University London)
Chris Reed (University of Dundee)
Other Information: This project was funded by UKRI ESRC (May 2020 – September 2021)
Speak, think, eat: leveraging conversational bots to foster healthy eating
The so-called “obesity time bomb” has revealed that in the era of the Internet of Things there is- still generalized ignorance about individual behavioural and metabolic fit. This situation calls for the development of effective solutions and policies aimed at over-weight prevention in the young population. Chatbots embedded in applications for food coaching can be of great use to cope with this issue since they combine a handy access with fast-paced information transmission. However, their effectiveness largely depends on the suitability and persuasiveness of their UI/UX design for targeted users.
This pilot research project aims at improving the design of food-assistant bots to increase users’ engagement, combining insights from Argumentation and Rhetoric with Psychology and UX (user experience). The project benefits from the collaboration with Libraway, a UK based socially-oriented startup that is developing a digital ecosystem of products/services for personalized healthy nutrition with chatbot capabilities. The company leverages an innovative nutrition logic and algorithm and is designing the chatbot to empower users’ eating choices to facilitate behavioural change. The primary target market of the company is UK but, as part of the prototyping phase, the company is conducting some early tests in Italy with the stated goal of transferring and adapting the knowledge to the UK market. Currently the human skills sought after health assistant chatbots are “affect” and “cognition” (Pereira, J. and Díaz, Ó., 2019). These components definitely increase the trust in food coaches necessary to attain change in food behaviors (Følstad, A et al. 2018). However, we believe that users need to be persuaded about the reasonableness of the provided recommendation to follow them. On the basis of existing studies about persuasive argumentation for health promotion (Grasso, F., Cawsey, A. and Jones, R., 2000) we will design different dialogue templates for the chatbot-user interaction and compare and contrast their effectiveness through a focus group composed by a sample of 20 testers. The results will be analysed according to state of the art knowledge about behavioural determinants of appetite and the templates revised accordingly. Both formal, conceptual and psychological aspects will be taken into account as part of the scientific design of this research project. Particular focus will be devoted to use of emojis as emotional clues and to types of reasonings used by the bot to back up recommendations.
Elena Musi (Liverpool University)
Rudi Palmieri (Liverpool University)
Alessandro Giudici (Cass Business School)
Rita Borgo (King’s College)
Charlotte Hardman (Liverpool University)
Neil Maiden (Cass Business School)
Other Information: This project was funded by the Wellcome Trust Institutional Strategic Support Fund, Interdisciplinary and Industry fund (May 2020 – April 2021)