Search for contacts, projects,
courses and publications

iTrust - Interventions against Polarisation in Society for Trustworthy Social Media. From diagnosis to therapy



Rocci A.


External participants

Budzynska Katarzyna

(Third-party responsible)


Digitalisation is rapidly transforming our societies, transforming the dynamics of our interactions, transforming the culture of our debates. Trust plays a critical role in establishing intellectual humility and interpersonal civility in argumentation and discourse: without it, credibility is doomed, reputation is endangered, cooperation is compromised. The major threats associated with digitalisation - hate speech and fake news - are violations of the basic condition for trusting and being trustworthy which are key for constructive, reasonable and responsible communication as well as for the collaborative and ethical organisation of societies. These behaviours eventually lead to polarisation, when users repeatedly attack each other in highly emotional terms, focusing on what divides people, not what unites them.

The– iTRUST project will deliver (i) the largest ever dataset of online text, annotated with features relevant for ethos, pathos and reframing; (ii) a new methodology of large-scale comparative trust analytics to detect implicit patterns and trends in hate speech and fake news; (iii) a novel empirical account of how these patterns affect polarisation in online communication and in society at large; and (iv) AI-based applications that will transfer these insights into interventions against hate speech, fake news and polarisation. The iTrust consortium is coordinated by Prof. Katarzyna Budzynska (Warsaw University of Technology), while the local team leader at USI is Prof. Andrea Rocci. The consortium members are experienced researchers with expertise in rhetoric, comparative political science, corpus linguistics, natural language processing, multi-agent systems and computational argumentation. 

Additional information

Start date
End date
36 Months
Funding sources
Swiss National Science Foundation / ERA-NET + EJP