I’m following the
Computation + Journalism 2014 symposium via the hashtag and livestream. Below are some highlights I collected from the opening keynote.
2014 C+J Symposium
We live in a society that is increasingly dependent on data and computation, a dependence that often evolves invisibly, without substantial critical assessment or accountability. Far from virtual, inert quantities, data and computation exert real forces in the physical world, shaping and defining systems of power that will play larger and larger roles in people’s lives.
Highlights from the keynote (in chronological order):
Keynote by Jon Kleinberg of Cornell: metaphors of information travelling online include the library and the crowd
travels on-line via (pages, links, association) & crowd (memes, contagion) |
We can track the flow of information temporally, structurally, and in terms of content, says Jon Kleinberg
But are crowd & library metaphors dual: people trailblazing through documents or documents transmitted through networks of people?
It’s easier for algorithms to track items (quotes, photos, phrases) than stories. Q: Does that encourage pack journalism?
Tracking stories through networks reveals difficulties eg., natural language. But can track quotes to show news cycles
Kleinberg explains tracking essential elements of a story (like phrases) as they move through networks.
Half of all reshares on FB happen in large cascades (>500) |
Is virality predictable? You as poster rarely experience it w your content, but you as consumer see it often
One solution: reframe question as tracking rather than snapshot instant: what are the chances of this being shared further?
On whether something “goes viral”: “An important moment in a cascade is the moment it escapes the neighborhood of the root.”
My thoughts are on how narratives or stories in news, eg images of ‘typical’ migrants, circulate and are widely diffused
Troubling finding here seems to be that actual content has less impact on how likely something is to go viral
Kleinberg now moving from global discussion to local conversations via threads or friends. What makes them engaging, long, short?
Tracking the virality of memes: Speed is important. Pics that get the first 1k of shares fast are more likely to go viral after.
Content more likely to spread if strangers share it = good reason for journalists to make sure their networks are diverse
For a week in September 2008, Obama commandeered the news media with the line “lipstick on a pig,” says Jon Kleinberg
That would be a nice job description for a business card: Meme tracker.
Kleinberg compares memorable & unmemorable movie lines as lab setting to see what features contribute to memorable or viral text
Why do we like “these aren’t the droids you’re looking for” but not “you don’t need to see his identification”
Memorable quotes are sequences of unusual words with common part of speech patterns – application to headline writing?
Memorable quotes are less probable in their word choices but more probably in their sentence (part-of-speech) structure – Kleinberg.
Jon Kleinberg: Socially shared information – how to predict success stories? Try a sequence of unusual words.
Is there an algorithmic pattern to why a movie quote is memorable? Take “you had me at hello.” What’s so special about it?
“Memorable quotes need to have a certain portability” _Jon Kleinberg
Memorable quotes tend to be more ‘general’: more present tense, indefinite articles, fewer third-person pronouns >> ‘portability’
Slogans in are like memorable quotes. “It just keeps going & going & going.” |
Is there an analogy of genetics for text: ‘fitness’ of text for sharing, mutation of ‘junk’ parts of quotes while core parts remain
Just as genes have functional parts and junk parts, so does text – Beautiful analysis of content prolongation
“Genetic analogies for memes are becoming increasingly rich” -Jon Kleingberg
Sharing on social networks: “Can cascades be predicted?” — paper by Jon Kleinberg et al
Great question: What are the features of content that make people STOP watching/reading/commenting?
Another great question: Are there computational ways to evaluate WHO gets to be quoted in the first place?