Gilat Lotan: It’s real­ly great to be here. I’m going to go over a bunch of data points which I think are inter­est­ing and might help us take this con­ver­sa­tion forward. 

So I’m Gilad. I’m chief sci­en­tist at the start­up called SocialFlow in New York City. And I look at a lot of data in my dai­ly rou­tine. So, we mine the full Twitter fire­hose, the full Bitly fire­hose. We build that into our prod­uct, but a lot of what I do is try to find out inter­est­ing things that we see in the data and tell sto­ries from it.

So, lo and behold human­i­ty is fair­ly con­sis­tent. We would men­tion morn­ings in the morn­ings. We get tired sort of towards the evenings. Talk about cof­fee more fre­quent­ly in the morn­ing. These are the sort of nor­mal diur­nal pat­terns that we see on Twitter, right. As expected. 

But when inter­est­ing events hap­pen and events that are out of the ordi­nary hap­pen it’s very clear that they hap­pen. So these are two very dif­fer­ent trends on Twitter. One is your typ­i­cal hash­tag that goes viral. So in this case it’s #blamethe­mus­lims. Starts very local­ly in London. It start­ed right after the events in Norway. So instant­ly, dif­fer­ent Muslim orga­ni­za­tions were blamed for what hap­pened in Norway. Actually a Muslim Twitter user around London start­ed this hash­tag and was say­ing, ​“Oh you know, your your clock is bro­ken? Why don’t you blame the Muslims? Oh, your car’s not work­ing? Just blame the Muslims.” 

And she and her friends were sort of using the hash­tag for snark. It sort of spread local­ly with­in their com­mu­ni­ty but died down at night. In the morn­ing, total loss of con­text, spreads thor­ough­ly on Twitter, becomes glob­al­ly trend­ing. And then sort of dies down. 

Alright. So we see an organ­ic trend. We see it grow. We see loss of con­text. So that’s inter­est­ing. Is that mis­in­for­ma­tion? It’s peo­ple who saw this hash­tag and are think­ing it’s some­thing very dif­fer­ent, right. There are peo­ple who got real­ly angry and said how dare Twitter have this as trend­ing? This is not okay. But it start­ed as a joke, a local joke. Spreads and then dies down. That’s total­ly organ­ic, we see this all the time. 

What we see in green is how a spam­bot net­work looks like. So you see these steps hap­pen, right. It’s as if some­one’s turn­ing a crank. Like, ​“Okay, add more tweets. Okay 50,000 more tweets. More tweets. Then take them down.” And what hap­pens here, we sus­pect it sort some­how reached some organ­ic growth. So it sort of start­ed some­how grow­ing in organ­ic traf­fic. Twitter caught it, shut it down, and then it died. So this stuff, we could clear­ly see this kind of stuff hap­pen­ing on Twitter just by look­ing at the dynam­ics of the lev­els of data. It’s so easy to tell. 

Another thing that we can tell is through­out all these social sig­nals. So we all know Justin Bieber and his scream­ing fans on Twitter. This is how much traf­fic he gar­ners, how many retweets he gets, in com­par­i­son to Pavel Globa. Who’s prob­a­bly the most retweet­ed per­son on Twitter, thou­sands and thou­sands and thou­sands of retweets. But we nev­er get to see him. Because his con­tent looks some­thing like this, right: 

We see lots of eggs, we see sort of— I mean he writes in Russian, he gets a ton of retweets. By the way, he pre­dict­ed that the Mayans were not right. So 2012, we’re safe. The world is not end­ing. He’s get­ting a ton of retreats for that. But obvi­ous­ly the social struc­ture of the net­work, and the way we sort of build our net­works in these spaces means that we don’t get to see a lot of this con­tent because we would­n’t sub­scribe to him. 

This is anoth­er thing that that we can get from look­ing at data. There’s an analy­sis that we ran on trend­ing top­ics on Twitter. If you Google ​“Occupy Wall Street trend­ing top­ics” [see fur­ther ref­er­ence links] you’ll see a sort a bet­ter expla­na­tion for what this means. But in effect, when you actu­al­ly look at the data and you see what top­ics are com­pet­ing with, so lev­els of atten­tion, Occupy Wall Street in blue is com­pet­ing with like Kim Kardashian’s wed­ding, right. Steve Jobs’ death. You see that the way we build these algo­rithms, the way these algo­rithms pro­mote cer­tain trends to trend­ing top­ics means we will nev­er see any­thing sort of slow­ly, slow­ly grows.

This is an inter­est­ing exam­ple of context-setting, right. Another infor­ma­tion flow. This is how the news about Osama bin Laden. So Keith Urban, who used to be Chief of Staff for Donald Rumsfeld wrote, he got off the phone, he’s like, ​“So I’m told by rep­utable per­son they’ve killed Osama bin Laden.” He did­n’t have a huge fol­low­ing, he did­n’t have a huge net­work. But he had peo­ple who who sort of set his post in con­text. You know, Jake Sherman for Politico, ​“Rumsfeld chief says this.” Brian Stelter, The New York Times, also point­ed to the fact that he used to be Rumsfeld’s chief. So with with­out that context-setting we sus­pect the infor­ma­tion would­n’t have spread as far. That was a case of truth­ful infor­ma­tion that spread, a rumor that spread, real­ly far. This is a case of sup­pos­ed­ly false infor­ma­tion that spread quite far. 

So at the height of Occupy Wall Street, Chopper [4] was sup­pos­ed­ly told by New York PD to move, that they’re clos­ing the air­space. So NBC post­ed this. It gar­nered a lot of respons­es. And they had to retract it because New York… It’s still unclear what exact­ly hap­pened but sup­pos­ed­ly New York Police Department said that the pilot mis­un­der­stood what they were ask­ing him, etc.

So what we see when we actu­al­ly look at the traf­fic, the green is the mis­in­for­ma­tion. So, there’s sub­stan­tial­ly more respons­es to the actu­al mis­in­for­ma­tion than the retrac­tion of it, right. But this is not always the case. This is just the case for this spe­cif­ic events, and there are actu­al­ly all these issues with these events. But the more inter­est­ing ques­tion that we should be ask­ing is first of all, what oth­er posts about the mis­in­for­ma­tion went out, right? Not the for­mal ones from NBC, but also who par­tic­i­pat­ed in the mis­in­for­ma­tion ver­sus the infor­ma­tion? And a lot of that we can get from the data. So I think I’m going to stop here because we’re run­ning out of time. And we’ll con­tin­ue the actu­al pan­el. Thank you so much.