It's, like, Kim: Kim Kardashian's tweets attact an audience.
The cherished idea of the Twitter universe as a gloriously turbulent and fluid place for debate has taken a major hit, thanks to new research from China.
At the same time, findings from the United States have demolished another plank of common wisdom about digital communications. There is, it turns out, no relationship at all between the number of times an online article is shared and the number of times it is actually read.
In a paper published in March, two Chinese social scientists, Fei Xiong and Yun Liu, of Jiaotong University in Beijing, revealed unexpected results from an in-depth study into how opinions form on social media.
The pair analysed 6 million posts from almost 2.5 million Twitter users during a six-month period. In looking at how Twitter users are influenced by the thoughts of other micro-bloggers, the researchers came to what they termed a ''non-trivial'' conclusion, meaning, pretty much, they aren't.
''Users seldom change their opinions,'' Fei and Yun wrote in Chaos, an academic publication devoted to ''non-linear science''.
Remarkably, the study was one of the first times actual online data had been used to test a theory in their field, known as opinion dynamics.
The Beijing findings challenge the much-vaunted utility of Twitter as a forum for debate. Fei and Yun found that in tweeted discussions around any given topic, a consensus formed very quickly, with the majority of users adopting one common position, and just a small minority taking an opposing stand.
The number of users who changed their opinion was very small, and that number continued to decrease, along a predicted frequency, as hours elapsed.
The interesting thing is that these developments - a big majority with one position, and opinion changing on an ever-decreasing scale - held, regardless of the topic under discussion.
One could be forgiven, therefore, for imagining Twitter discourses as less like ancient Greek philosophers talking in an olive grove, and more like an army of chimpanzees banging pots with wooden spoons.
However, a couple of additional observations made by Fei and Yun moderate this picture, although perhaps not comfortingly.
The largest number of opinion changes, they observed, correlated with topics that didn't get the blood boiling. In other words, Twitter users are more likely to be swayed by argument if they don't give much of a hoot in the first place.
Second, there is no obligation for Twitter posters to admit that they have changed their positions.
''In fact,'' wrote the researchers, ''even if agents can update their internal opinions, they do not always publish posts to share their ideas with the public and may drop out of interaction.''
What people really think, therefore, might be hidden. This, it turns out, is certainly the case in the matter of ''social sharing'', the posting on social networks of links to news articles.
It is often assumed that a huge number of social-media shares indicates lots of people are reading a particular piece. According to findings published by New York data science company Chartbeat, that is pretty much a crock.
The company advises media groups on the latest metric for measuring online readership - attention analytics. Previous methods measured only the number of visitors a web page received - the ''click-through'' - but this method records the amount of time a reader spends actually looking at the thing.
Again, the results are uncomfortable. According to Chartbeat, 55 per cent of people spend less than 15 seconds looking at any page they visit, and the relationship between social sharing and reading, it turns out, is at best minuscule, and at worst non-existent.
The company tracked 20 billion visits to 580,000 online articles on 2000 websites and concluded that, on average, pages generated only one tweet and eight Facebook ''likes'' for every 100 visitors
Drilling deeper, the relationship between what people like and what they read becomes even stranger.
''Some types of content are barely shared at all, although they're heavily read,'' Chartbeat strategist Lauryn Bennett says. ''How many people tweet links to stock ticker pages? Other types are heavily shared but less read. Any story that Kim Kardashian tweets will instantly have hundreds of other tweets, regardless of whether it's being read,'' she says.
It is difficult to know which is more depressing here - the idea that lots of people care what Kardashian thinks is a good yarn, or that they don't, but still feel compelled to pretend they do.
Bennett says the relationship between social shares and actual reading is complex. People do sometimes share stories because they are under discussion on Facebook or Twitter, she says.
Similarly, the appearance of any story on a media outlet's ''most popular'' or ''most emailed'' table often results in a rise in shares.
''But it's not to say that those share numbers relate to depth of scrolling or reading the page at all,'' Bennett says. ''In fact, in past research we've done, we've seen no relationship between how much of the page people have engaged with and social sharing.''
In other words - and again, Ms Kardashian, perhaps unfairly, comes to mind - it is quite possible to suggest that someone should read something even though they personally have not done so.
There is one point at which Chartbeat and the Beijing researchers find common ground: the relationship between an opinion expressed and an opinion held is fundamentally unknowable.
''So do those social-sharing numbers on the top of the page affect if people will read the full article or share it?'' Bennett asks.
''Maybe. Observing reader behaviour on an article helps us understand what people do, but doesn't allow us to understand their intent or what directly influenced them to take those actions.''
(If you liked this article, please share it now. Yes, Kim, that means you.)