New models in the news feed algorithm?

Facebook announced changes to the News Feed yesterday, Remove, Reduce, Inform: New Steps to Manage Problematic Content, to enhance their “Remove, Reduce, Inform” strategy. The Facebook strategy appears to be adding more buttons with information to posts and images and tackling groups.

I find the former part interesting as the platform appears to be struggling with fact checking and the spread of misinformation which I am assuming underlies the current changes. The changes to the context button are about trust but how might one metricise trust? As Wired mentions in their write up, “Facebook is Changing News Feed (Again) to Stop Fake News” , the company has mapped a large part of the web, using similar algorithms to PageRank, to create a set of references for their model.

Yet how might this cope with the known universe of the alt-right who have successfully created their own web using links and media literacy? Are these models to be penalised or mapped into the News Feed? From a research perspective, I would be interested in seeing how we might surface and question the models that underlie the trust settings. This does come from an algorithmic accountability perspective but also from a concern about how Facebook might be trying to set itself up as a central part of the trust model.

I see that Messenger is getting some changes as well and assume that this is part of the great move to Messenger / WhatsApp. How does this centralise similar concerns but in messaging?

A longer term comes to mind. How does this continue the marge of corporate terms and conditions about what we might study? How can we study these new models and ways that a reality that is decided by machines and learned human models affects our perception of the world and cognitive practice? I am sure that a Stieglerian response, following The Automatic Society Volume 1 (Polity, 2016), might be appropriate for now but will it always? It seems that this reflects conversations about trust in the Semantic Web that I overheard years ago and it comes back again.

Trust is not an easy thing to cope with in the web, let alone the social web.

No Comments

Leave a Reply

Your email is never shared.Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.