The 2024 Legal Defense Fundraiser has ended early due to the Board's decision to settle the suit. Donations will go towards the creation of a legal defense fund. Discussion on the matter can be found at the RationalMedia Foundation noticeboard. Thank you for your generous support.
Filter bubble
Someone is wrong on The Internet |
![]() |
Log in: |
A filter bubble is the phenomenon of internet users only, or almost always only, accessing content that they agree with or that confirms their biases, preconceptions, or even bigotry.
While filter bubbles could be constructed manually (e.g., homeschooling, utopian communities, cults, North Korea), many journalists and commentators have warned about the potential for the internet — and social media sites in particular — to automatically create dangerous filter bubbles for many more people than in the past, leading to consequences these consequences to people from different political backgrounds:
- Becoming more polarised in their views, since they only hear arguments from their own "side of the aisle"
- Believing more fake news and remaining ignorant of or denying the debunkings of it from neutral or opposing sources.
- Voting against their economic interests, since they only hear propaganda about the other side and very little criticism about their side's economic policies
- Failing to understand and empathise with people with different political views, since they do not encounter them very often
- Assuming that everyone they encounter will agree with them on certain beliefs, leading to cringeworthy or offensive statements
- In some cases, starting to see people with different political beliefs as "the enemy" or "traitors" — someone seeking to "do down the country" — rather than fellow human beings to have a civilised discussion with — as has started to happen with some of the more extreme Brexiters and Remainers in the UK.
How filter bubbles are formed
Recommendations
Filter bubbles can be formed in a completely unintentional way, with no intent to ignore different views on the part of either the user or the social media algorithm designers. Suppose a user joins Twitter and follows a number of people who either have liberal political views or do not tweet about politics. The recommendation algorithm will pick up on this and recommend other accounts that other users of similar tastes have followed in the past. Quite likely, no part of the algorithm has a detector for "liberal" or "conservative" sources — it just so happens that many liberals like to follow other liberals on Twitter, and many conservatives like to follow other conservatives on Twitter, and the algorithm tends to learn this and apply it at the level of simple statistical patterns. Essentially, the recommendation algorithm is blind as to what groups it is effectively classifying people into and also quite likely blind as to the "quality" (reliability, trustworthiness, etc.) of the accounts it recommends, as well. Nor did the user necessarily intend to form a political bubble — if the algorithm is poorly-designed, they may follow a handful of reasonable, moderate conservatives as well, but not enough to trigger the recommendation algorithm into recommending more of them.
The increasing use of artificial intelligence by social media companies may exacerbate this problem, by making social media and even less socially-focused sites such as YouTube, better and better at serving up clickbait and fake news that people are likely to click on because it confirms their pre-existing beliefs, rather than optimising for quality or truth.
Groups and message boards
However, filter bubbles are not always unintentional. The increasing popularity of Facebook Groups and sites like Reddit, which allow anyone to create communities of arbitrary size, has meant that many people choose to join discussion and link-sharing groups that may be highly partisan or woo-promoting, and that may be openly hostile to people with other views. Moderation policies are almost entirely at the discretion of each group's moderators on such sites, and therefore moderators are free to exclude people for having skeptical views or different political views, and sometimes do so without being open about it.
In the 1990s and 2000s, the equivalent would often have been regular political chain emails among like-minded people, often retired people who had plenty of time on their hands. Dark-money-funded conservative pressure groups deliberately made use of such tactics, pretending to be ordinary people, to spread fear about Democrats in general and Barack Obama in particular.
Both Reddit and Facebook also allow the creation of private groups, which are harder to police for site-wide rules violations, because outsiders cannot see their contents, and may find it challenging to infiltrate them. However, getting even public groups kicked off those platforms for e.g. extremism or for promoting harmful lies may be difficult, because of the US tech industry's tendency towards few rules for users, and lobbying for low regulation. Initially, Twitter only allowed users to report other users for spam, entirely ignoring the issues of harassment, bigotry and threats of violence. Reddit notoriously allowed some paedophilic and racist subreddits to exist on the site, before belatedly banning them. Today, Facebook and Reddit both maintain their support for "free speech" (i.e. pretty much anything goes).
Though Facebook has tried to use its AI algorithms for good, by automatically finding skeptical/opposing sources and placing them underneath suspected fake news posts,[1] this approach presumes the kind of people dumb enough to fall for obvious fake news are smart enough to read down to, click on, read and understand, oppositional articles. (Since some people struggle with reading English news articles for various reasons such as illiteracy, or having English as a second language, and prefer to watch videos/TV, Facebook might do better by leavening debunking videos into the mix — and encouraging fact-checking and skeptic organisations to create such videos in the first place.)
The image/message board 8chan, with its far-right-dominated /pol/ board, has been highlighted by the media and by US congressional representatives as a possible radicaliser for individuals such as the far-right mass shooter Patrick Crusius, who was a believer in the white genocide conspiracy theory.[2] 8chan was subsequently booted out by its internet access provider after the adverse publicity, and later relaunched on the dark web.
It is likely that any liberal or leftist that had attempted to enter 8chan's /pol/ to counter the racism and extremism, would have been trolled, griefed, ignored, or would have given up in disgust at the amount of extreme, offensive racism on open display. Thus, even in venues that are largely unmoderated and uncensored, extremism can effectively create its own kind of filter bubbles on message boards, by deterring all opposition (in anything other than very large numbers).
In particular, this highlights the risk that a filter bubble may lead to a person accumulating dangerously extreme and unhinged beliefs and/or growing disconnected from any moderating influences on their own violent impulses. It is reasonable to ask: if a mass shooter motivated by extremist ideologies that he had picked up online, had instead sought out a wider range of views or credible sources on the issues of concern to him, would he still have developed those extreme views, and would those mass shootings have still happened?
Geographical "filter bubbles"
In tandem with online filter bubbles, people in the United States are now becoming more politically polarised by geography. There are different theories about why this has happened. Some have argued it is due to the fact the US population is highly mobile (many Americans will relocate to take a different job) and some people have a preference for living near to people with similar political views. However, it is questionable whether very many Americans do actually factor that in when choosing a new place to live. There may be some "unconscious sorting" going on, with Democrats being more likely to prefer dense, walkable areas, and Republicans less likely. Other factors may include the two main US political parties becoming more tilted towards rural (Republican) or urban (Democrat) areas in terms of who they try to, or successfully manage to, appeal to, and people moving into heavily Democratic/Republican areas tending to become more likely to vote Democrat/Republican due to influences from people around them.[3]