Meta's fact-checkers gone. Are facts really under threat?
TLDR: Yeah nah.
“If we don't believe in free expression for people we despise, we don't believe in it at all.”
Noam Chomsky
The future of facts on social media is (supposedly) worrying
The recent decision of Meta of axing their third-party fact checkers on both Facebook and Instagram and replacing them with a user-based model of “community notes” like X/Twitter has had rather predictable reactions. While those that have suffered at the hands of the censorship industrial complex have celebrated the move, those that have been spared — and also those who advocate for keeping social media free from so-called misinformation — have either strongly condemned the move or, at the very least, both conveyed their concern and made grim predictions about the future of social media without their desired information regulators. After all, Facebook and Instagram combined account for a substantial share of the social media market (at least amongst the platforms listed here by Statista).
Amongst the mixed sea of jubilee and grief a recent article got my attention. If you receive the daily communications from The Conversation you might have read it too. It got introduced, ominously (or should I say, click-baitily), as “facts are under threat” on the subject line. The communication read as follows:
One of the many issues [Donald Trump’s] political comeback raises is the issue of trust in media – including social media, especially after Meta announced it would stop fact-checking on its platforms.
The onus of keeping large parts of the internet civil will now fall largely to social media users themselves. But do everyday Facebook scrollers value the sorts of things that uphold democracy?
[…]
The results [of the research conducted by Mark Andrejevic and his team discussed in the article] show people who get their news on social media are among the least likely to share civic values, such as being open to opposing views. Perhaps, Andrejevic suggests, these are not the people to leave in charge of managing a healthy online debate.
To say that this summary is misleading is to be kind. Closer inspection of the (freely available) full summary of the results — a set of 49 slides that are included as a link in the article — reveals that the story is quite different. I will not rehash everything in the summary or dissect the methodology here as interested readers can parse these on their own time, but I will nonetheless take the time to explain why I believe Andrejevic’s and The Conversation’s read of these results is wrong, as it appears to be more a consequence of their political leanings rather than a more balanced and full read of the data. Kudos to them, nonetheless, for being transparent and making the results publicly available, and for conducting this research on Australians.
Mapping civic disposition, media use and affective polarisation
That is the title of the research prepared by Essential Research for Monash University referred to in the article. It is a survey of about 2,000 Australian residents aged 18+, crafted to be representative at the national, gender, location, and age levels, that ran from the 1st to the 14th of November 2024. Two critical parts of the survey involved (i) asking these people where they usually got their news from (and the results show people could list more than one source based on how many hours they used them), and (ii) compiling a “civic values scale” that “aims to measure levels of trust in media institutions and the government, as well as people’s openness to considering perspectives that challenge their own”, as Andrejevic writes. Someone with low civic disposition would be someone who does not trust mainstream media to report accurately and someone who does not trust the Federal government to do the right thing for Australians, for example, whereas someone with high civic disposition would always trust mainstream media to report accurately, and would always trust that the Federal government would do the right thing for Australians (see slides 46 onward in the research).
Andrejevic also writes that “people who rely on social media for news score significantly lower on a civic values scale than those who rely on newspapers and non-commercial broadcasters such as the ABC”, while presenting the following graph.
I got curious about the context of the data presented here, and so I tried to find the graph in the slides, at least in some other shape or form. I could not (and if I missed it please let me know in the comments section). Therefore I am left wondering what data, specifically, has been crunched to create these numbers, and how. Notwithstanding this issue, I personally find all Civic Values Scores presented here remarkably similar considering the range of the variable (it could go from 18 to 90). Additionally, and assuming the values presented here are averages, I am also not entirely sure the average score for social media is significantly lower as Andrejevic writes (there is no indication of dispersion or a p-value). Even if it were indeed statistically significant, this result appears insignificant in the larger context: the average Civic Value Scores for each main source of news are remarkably similar across the board when considering the range of the variable. Arguing otherwise would be akin to arguing that you go “significantly” slower if you went from 66 km/h to 62 km/h in your car.
Inability to find that graph on the slides aside, the results I nonetheless found point in a markedly different direction. As people were allowed to list all the sources of information and how much they used them in an average day, different types of users arose, contrasting sharply with the “main source of news” view above (see slide 16 in the pack). Therefore, Essential Research presented the results in a manner more befitting the data. The slides reveal, for example, that most people are daily social media users, including those who have high civic disposition. Therefore, The Conversation’s editor writing that “people who get their news on social media are among the least likely to share civic values” is simply wrong.
Nonetheless, Essential Research also shows that “those with high civic disposition are more likely to primarily use online newspapers and non-commercial media, whereas those with low civic disposition are more likely to use commercial and social media”, as shown below.
The differences between those with high / very high civic disposition and those with low / very low civic disposition are, however, not abysmally different, especially considering (i) that people with high civic disposition can have scores not so diametrically different to those with high civic disposition based on both the distribution of the scores (slide 10) and how they were labelled (slide 9), and (ii) that the difference between being in the high civic disposition vs the low civic disposition group could be due to strong opinions on a handful of questions in the rubric (see the minimum and maximum scores for each group on slide 13, and consider that each “wrong and strong” answer costs you 4 points in this rubric).
The most critical part of the research, at least in my opinion, is somewhat predictable and is presented in slide 21 (below). Who would have thought that reading from multiple sources might increase your civic disposition, or that, alternatively, people with high civic disposition read from multiple sources? Rather than fact-checking, those interested in higher civic disposition should encourage people to break and disrupt their echo chambers with new ideas, rather than curating them even more via fact-checking.
Considering that there is an important contingent of people of high and very high civic disposition on social media sharing the space with those of low and very low civic disposition — and that the former group also read from a diverse range of sources and (hopefully) points of view — the ability of people to write “community notes” on diverse topics should be viewed as a welcoming move that brings more information to everyone’s eyes, not just to those of low civic disposition as the writing might imply. After all, having read from a wide number of sources and having a high civic disposition does not necessarily transform someone into the holder of ultimate truth. The absence of fact-checkers, therefore, is not so much about letting lies flourish, but about getting rid of powerful, centralised censorship teams with the ability to control what can be said on social media about a given topic, often with more interest in political outcomes than in truth itself.
On the back of Meta’s decision to axe their fact-checkers, the research presented by Essential Research and Andrejevic’s team therefore suggests that Australians are well equipped to take on social media and create a rich environment for discourse where people with all levels of civic dispositions can coexist without feeling they will be censored for posting the “wrong” information. If we cannot trust people and the facts, information, and points of view they bring to the table, why should we trust so-called “fact-checkers” that at times are equally or more biased and wrong than the people at large and that are much more easily controllable and corruptible by virtue of the size of the team?
So, what about the future of facts on social media, then?
Mark Andrejevic writes that “the decision [of dismissing fact-checkers] is widely viewed as placating an incoming president”, that “social media platforms are not designed to foster democracy”, that “platforms want to wash their hands of the fact-checking process, because it is politically fraught”, and that “the algorithms they craft play a central role in deciding which forms of expression make it into our feeds and which do not.” I find myself agreeing with him in many regards.
Social media companies (and probably media companies in general) tend to ingratiate themselves with the incumbent (or with the political side that reflects the dominant internal culture) and meddle (or stop meddling) with the flow of information to realise their political visions. This is something beyond Trump’s election — the Twitter Files exposé and RMIT’s FactLab debacle come to mind, for example. Moreover, as Andrejevic writes, these algorithms and companies are designed to maximise engagement (and thus profit), not democracy (even when free speech is a cornerstone of democracy ), and fact-checking is indeed a politically (and dare I say methodologically) charged endeavour. Yet, all of the above reasons are precisely why fact-checking should not be left in the hands of these companies. After all, they are interested in engagement and profit, not in truth, facts, or democracy, and whatever “facts” will survive the fact-checking exercise will depend for the most part on the flavour of the political wind blowing in a given period. Fact-checking in this latter sense is, in essence, another form of power centralisation for the sake or whatever agenda those doing the fact-checking (and those who designated them) might have in store.
In a sense, facts are not under threat as The Conversation might suggest, as this would imply that they might get buried or disappear altogether from social media just because users are able to write caveats and annotate them. On the contrary, facts get enriched when diverse considerations are brought to the table. But if we are to take the view that facts are indeed under threat as they might be buried under a never-ending parade of community annotations that might mislead the reader, they are even more under threat under a system that allows unelected bureaucrats and activists to delete them, hide them, and wave their badges of honor in front of them to discredit them, all while working for companies that do not care about them.
Probably the only time I would personally think that facts would be under threat is when those who read far and wide leave the platforms altogether, but Andrejevic’s and Essential Research’s findings strongly suggests that we are nowhere near this point and that Australians, despite their differing views, tend to cluster together around a mean point in the civic disposition scale.
Nothing could give me more hope about the future of facts on social media.