arXiv preprint ·

Counterfactual assessment of social media interventions by separating exposure from engagement

Tommaso Pisani, Pietro Gravino and Giulio Prevedello

Published
Source
arXiv preprint, posted before peer review
Original
Read it at arXiv
Topics
Physics & SocietyComputers & SocietySocial & Information Networks

Abstract

Social media interventions are usually evaluated in individual-level experiments, which are costly to scale and often miss collective effects. We propose the Exposure--Engagement model, whose parameters separately control what social media users are exposed to (recommendation) and how they engage (behaviour). After fitting it to the Italian Twitter debate (2018--2022), we validate it at individual and collective levels. We simulate behavioural interventions (prebunking and accuracy prompts), calibrated on effect sizes, and a recommender-system intervention (chronological feed), comparing effects on engagement, attention diversity, and segregation. All three reduce engagement. Behavioural interventions slightly reduce diversity, leaving segregation unchanged, whereas the chronological feed increases diversity and reduces segregation. For these interventions, changing recommendation does more for the debate's structure than changing behaviour. We design a recommender-system intervention that increases engagement and diversity while reducing segregation, although with about a tenth of the chronological feed's gains, showing the framework's potential for pro-social intervention design.

The authors’ abstract, as arXiv lists it. arXiv’s listings, abstracts included, are free for anyone to reuse.

Read it on ReadSearch

Each opens after you create an account, and brings you back to this paper.

Already a reader? Sign in

All new papers in Physics & Society

Get the few papers worth your time in Physics & Society every morning.

ReadSearch reads the new papers in your field each day and brings you the few worth reading, each with its key finding in a sentence.

Papers belong to their authors and publishers; every page here links to the original. All research topics