How LLMs can be used to support and augment the fact-checking process? MISDOOM Digital Democracy Centre NORDIS Infomedia_uib Carl-Gustav Lindén Read more here: link.springer.com/chapter/10.100…

End of the first day of #MISDOOM24 with a view!

We kick off the second day of #MISDOOM24 with a keynote by W. Quattrociocchi on "Polarization: Information, Opinions and other Monsters"

Confirmed: 👌 Soft moderation interventions (e.g. fact-checked labels & blur filter) reduce engagement with misinformation about the Russia-Ukraine war among Facebook users. This finding held even when accounting for user-specific factors. #Misdoom

The study authored by Anatoliy Gruzd 🇨🇦 🇺🇦 🏳️‍🌈 Philip Mai Felipe Soares is published open access at doi.org/10.1007/978-3-…

Interestingly, the more intrusive intervention of adding the blur filter in front of a fact-checked image or video did not produce a stronger response in comparison to a simple fact-check footnote.

For this study we developed and used ‘ModSimulator’ – Open-Source Research Tool to test the effectiveness of soft content moderation. The tool is available to other researchers to customize at github.com/smlabto/modsim… (It's a branch of the Mock Social Media Website Tool)

We find that irrespective of the intervention used, there are other predictors of engagement w/false claims. For example, individuals’ preexisting beliefs in pro-Kremlin claims & trust in partisan sites for news re: the Russia-Ukraine war increases engagement w/flase claims.

On the other hand, trust in fact-checking organizations and being an active commenter on Facebook decreases engagement with false claims. More details in the paper: doi.org/10.1007/978-3-…

Engaged discussion in Panel 6: Platform regulation and moderation strategies #MISDOOM24