You might have heard rumors that the newsfeed algorithm at Facebook and the video recommendation algorithm at YouTube are spreading fake news, or that artificial intelligence (AI) can now rapidly generate convincing articles and make videos of people doing and saying things they never did, or that machine learning algorithms will save us from fake news by automatically detecting it and labeling assertions as true or false. But what do these claims even mean, and what should you believe?
Much of what we know, or think we know, about what is happening in the world we learn by reading the news. But nowadays “the news” means something different than it did in generations past. What we read primarily today are articles on the internet—everything ranging from casual blog posts to meticulously researched stories on national and international news sites. The transition of journalism from print to screen does not inherently mean what we read is less truthful than it used to be. However, this technological transformation has enabled a less overt but nonetheless extraordinarily influential economic transformation: the datafication of the journalism industry.
At the bottom of the internet media food chain, if you will, are small blogs and websites that cover very focused issues, interests, or regions; these can be single author or multi-author. The next tier up comprises the blogs of newspapers, magazines, and television stations. This is a confusing middle ground because many of these blogs share the name, URL, and logo of a recognizable news source yet the editorial standards are generally lower than those of the parent organization, and many of the contributors lack the journalistic training one might expect from the parent organization. Then at the top are the official news sites, which can be regional but tend to draw a large national or international readership. This hierarchy is not about quality—indeed, some very focused small blogs produce content of extremely high quality, while some big-name national news sites consistently publish articles of seriously questionable accuracy. The levels here are more about the size of both the audience and the organization and about the scope of the content.
Information flows both vertically and horizontally through this internet news hierarchy. When the Washington Post breaks a big story, it is only a matter of hours before the New York Times covers it as well, and vice versa, often simply by reporting what was reported in the other newspaper’s article. This is horizontal propagation , and it happens because even though the second newspaper cannot claim credit for breaking the story, it does not want its readership to obtain this information directly from the competitor newspaper. Vertical propagation happens in two directions. A big story broken at the top will be covered and duplicated by smaller news organizations and blogs because, similar to horizontal propagation, this is an easy way of keeping readers without doing much work; this is a downward flow of information.
While there is an obvious redundancy, hence an overall systemic inefficiency, to both horizontal propagation and downward vertical propagation, the only real harm to the truth-seeking reader is that important details might be omitted and facts distorted as the story is passed from organization to organization—though sometimes a more specialized blog will provide a valuable service by delving deeper into a particular facet of the story than would be appropriate for the top-level organization. It can be quite illuminating to find a story that was broken by one newspaper and then compare its coverage across a range of other newspapers and blogs; this is an excellent way to uncover the ideological inclinations of different organizations, since the same set of facts will be colored by the different viewpoints involved.