from Hacker News

Debunking Princeton

by friggeri on 1/23/14, 10:08 PM with 145 comments

• by hooande on 1/23/14, 10:39 PM

  I hope this becomes a bigger story than the original. I can't tell you how many times someone has thrown a junk science article in my face, thinking that the issue in question was now settled. A survey with a small sample size, outsized extrapolations and numbers that don't match the accompanying conjecture.

  There should be consequences for the people who publish these things. People have a tendency to believe anything that someone in a lab coat says, especially if it supports their point of view or anecdotal experience. In many cases the people who do the research present it with few qualifications while not standing behind assumed implications. If someone publishes sensational and link baity findings they should say, unequivocally, "I'm willing to stake my reputation on the idea this trend is real and will continue" or "These are just data and I'm not willing to say that they have any bearing on reality".

  Facebook may not have been right to dignify the initial post with a response, but I hope it works for the best. They say that some attention is better than no attention at all. It's important that this applies to self promotion and persona creation and not science. If somebody has something crazy to say, they should start a personal blog. Those who want to intentionally attract media attention should present themselves as such, instead of pretending to be doing any kind of meaningful experiment and hypothesis testing.

• by devindotcom on 1/23/14, 11:35 PM

  Seems like Facebook is responding to the media interpretation of the Princeton paper rather than the paper itself. I saw no problems with the paper: it applied an epidemiologically-inspired statistical technique to Facebook using a defunct precursor as exemplar. The limitations of this technique are obvious, but it's an interesting idea.

  People do "research" like this all the time - you throw shit at the wall and see what sticks. Most of the stuff that ends up on the floor never receives any attention at all, so you don't hear about it on first-tier news sites. But when it's about Facebook, it goes viral, and suddenly is the subject of intense scrutiny. They didn't bring this to the UN for a call to action. They didn't start a company around it. They just applied an idea to some data and wrote it up. And now the entire Internet is making fun of some exaggerated version of their idea, summarized by Huffington Post hit-mongers.

  Personally, I applaud these guys for putting in the work to test out a theory. If it's not correct, it will go in the bin with the other ten million papers with flawed theories, premises, methods, or other aspects that have been published in the last day or two.

• by crm416 on 1/24/14, 9:31 AM

  As a Princeton student, it's been pretty frustrating to watch this devolve into (drawing from comments on this page) a "sham study by Princeton", with this response being a criticism of "Princeton's methodology", and that "Princeton was deserving of a response like this".

  It's been pointed out by a few other HNers, but this type of logic does a massive injustice and disservice to all the institution's undergraduate and graduate students, as well as its professors, who work hard to produce some of the highest-quality research in the world.

  To say that this is a "Princeton study" is to present this as if it were endorsed or produced by the administration or some department or even a tenured professor. Instead, let's remind ourselves that this was a pre-peer reviewed paper posted on _arXiv_ by two PhD students (who have likely been at the university for a few years, tops). To paint this as the Princeton community getting together as a collective and putting forth their best attempt to "debunk" Facebook is just hilariously unfair.

  Look, there's a thick anti-higher education slant on HN. People love referencing the higher education bubble and the 'demise' of the current university system or whatnot. But it'd be nice if we could keep things in perspective here and at least do better than the media, who can't wait to pounce on a Princeton vs. Facebook feud.

• by btown on 1/23/14, 10:47 PM

  What drove Cannarella and Spechler, two mechanical engineering Ph.D. students, to prerelease a paper about network theory and epidemiology, without coauthoring/consulting with epidemiologists or people experienced in viral communications theories, is beyond me. But the title "Debunking Princeton" seems to suggest this should be considered representative of all of the quality of research Princeton outputs, which is certainly not the case. Many posts have been made questioning the strength of the paper in question, and the fact that one with such a link-bait title is rising on HN is unfortunate in my view.

  Once again, a reminder to everyone that there are no peer review requirements for papers posted to arXiv. There is no evidence that the original paper was ever accepted by any journal or conference, and not surprisingly given the speculative nature of the study, the advisor of the two Ph.D.-candidate coauthors declined to place his name on the paper. So as an institution, Princeton is no more responsible for this paper than Obama is responsible for that drink machine being broken down the hallway (thanks Obama), even though it happens to be a drink machine affiliated with his country. In that regard, with all due respect towards Mr. Develin, I'm going to have to "debunk" "Debunking Princeton."

  [Full disclosure: I am a Princeton alumnus.]

• by tokenadult on 1/23/14, 10:26 PM

  A suitably joking response for a "study" that didn't need to be taken too seriously (and wasn't taken very seriously by most media outlets, including the one that led a story with it). As we discussed yesterday here on Hacker News[1], the study methodology was not sufficiently validated to convince most people that Facebook will massively lose users, even if all the data were correct. More likely, the study's model was just flat wrong.

  [1] https://news.ycombinator.com/item?id=7104904

• by md224 on 1/23/14, 10:57 PM

  So, if this all snark, does that mean there's no "strong correlation between the undergraduate enrollment of an institution and its Google Trends index"? Because that would be a pretty interesting correlation.

  I'm also curious why Princeton's search volume seems to have declined. Obviously it doesn't mean Princeton is going to disappear, but what does it mean? Could be statistically insignificant, perhaps.

  EDIT: Here are some comparisons:

  http://www.google.com/trends/explore#q=%2Fm%2F05zl0%2C%20%2F...

  Interesting how the initial spread seems to narrow.

• by joelgrus on 1/23/14, 10:33 PM

  This is pretty funny, although they ruined it with their "HEY GUYS THIS WAS A JOKE" at the end.

• by nilkn on 1/24/14, 12:36 AM

  For anyone curious about the author of this post... http://en.wikipedia.org/wiki/Mike_Develin

• by DaniFong on 1/23/14, 10:27 PM

  Big data: 0. Smart analysis: 1. Snark: 100. :-)

• by abus on 1/24/14, 12:06 AM

  I see this as Facebook's willingness to use its unfair influence to discredit anyone who dares challenge it. In other words, "if you have a page here, we won't hesitate to use it against you".

• by smoyer on 1/23/14, 10:31 PM

  Hmmmm ... I think I believe both articles! Higher education and social media are both in decline. Who needs critical thinking and the ability to gossip 24/7? I'm going to go rewatch Idiocracy (http://www.imdb.com/title/tt0387808/).

• by S4M on 1/23/14, 10:55 PM

  Call yourself "Data Scientist" and publish graphics without units. Congrats!

• by debacle on 1/23/14, 10:36 PM

  > not all research is created equal

  You mean "not all linkbait is created equal?"

• by lucb1e on 1/23/14, 11:41 PM

  > every Like for this post counts as a peer review. Start reviewing!

  Wait did Facebook just start likewhoring themselves too?

• by curiousAl on 1/23/14, 10:29 PM

  Google Trends Extrapolation: The Nostradamus of our times.

• by adharmad on 1/24/14, 1:25 AM

  Nice comeback by facebook. Instead of defending facebook against the claims in the paper (facebook is not dying! - a very hard-to-defend position) they turned it around and used the same techniques used in the paper to show that Princeton is dying. Offense is the best defense!

• by augustocallejas on 1/23/14, 10:28 PM

  One of the top comments on that page is that they loved the tone of article. Personally I hate sarcastic responses (no matter how correct you are), which sound very unprofessional. I hope to see less of these types of responses in the future.

• by yuvadam on 1/23/14, 11:31 PM

  Hey, Facebook, wanna prove Princeton wrong? How about publishing actual metrics proving that user engagement is at an all-time high?

  (Oh, you don't have metrics to prove that? Is that because users are jumping ship?)

• by dmazin on 1/23/14, 10:53 PM

  That was... unprofessional.

• by lockone on 1/25/14, 3:06 PM

  My guess is that the real research is the response to the paper and not the conclusion presented in the paper. Kind of a shoot in the air to see which way people run. The information gained by observing Facebook's response could prove useful to other big companies that take a hit in today's media driven society(internet). What works and doesn't work in their PR response. They could also be tracking other data, such as stock prices of other social media companies. It's what makes research so much fun.

• by snowwrestler on 1/23/14, 11:09 PM

  The article jokes about "the scientific principle 'correlation equals causation,'" but I feel like I see people make this mistake all the time, particularly with statistical concepts like expected value, and epidemiological risk factors.

  Statistics is descriptive--it's not predictive. It tells you about the data you have. It doesn't tell you why the systems produced that data, and whether they might produce very different data under different conditions.

• by waylandsmithers on 1/24/14, 2:45 PM

  Yikes. I found this to be shockingly childish and the fact that facebook got this defensive makes me think they believe the article has at least a kernel of truth to it.

• by jjcm on 1/23/14, 10:54 PM

  Sassy, but I feel like Princeton was deserving of a response like this. It was a pretty poor study that had a lot of lazy data and speculation behind it.

• by protez on 1/24/14, 9:15 AM

  What is causation exactly? There's no way to prove any instance of causations at all, if we delve deep into the problem of causation. Even though all data, or correlations, support F=ma, a single exceptional case can disapprove the "correlation," therefore we cannot assert force causes accelerations, and so on, since no amount of correlations is ever enough to prove causation.

• by elwell on 1/23/14, 11:37 PM

  I wonder if this article would exist if the first one had been by Harvard instead of Princeton (given that Harvard was the origin of Facebook).

• by xtc on 1/24/14, 2:39 PM

  Rather than refute the negative prediction for Facebook we're presented with a sarcastic appeal to those who don't think Facebook could ever dissipate.

  This only speaks more towards an argument that Facebook isn't being serious enough with its own statistics. Of all organizations Facebook should be the first to spot a trend especially with esteemed data scientists like Mr. Develin.

• by logicallee on 1/24/14, 10:51 AM

  If they weren't being satirical, this response would be a logical error, this one: http://en.wikipedia.org/wiki/Tu_quoque (The logical error is that it could very well be true that Princeton is declining, and proving this would not affect the results of the Princeton study that Facebook is declining.)

  However, the fact that they did so satirically sounds like when a politician says, "that's preposterous" instead of "that's false."

  In addition, the data they use is pretty weak. The first chart shows that Princeton "crashed" or died between 2010 and 2011, since that's when that graph tanked. Since that data isn't good to make their Tu Quoque argument (that Princeton "will die") they are being satirical and ignore it. The second graph shows that more and more Non-Princeton articles are published. But this is due to more and more non-Princeton publications. Princeton has a fairly static amount of output, as the world's universities started outputing scholarly articles in English, you would expect Princeton's share to drop.

  More interesting would be if its share of what it is trying to go for, Nobel Prize Laureates, publication in Science and Nature, whatever - were on the decline. This isn't addressed, just a global proportion of all scholarly articles: not Princeton's aim.

  However, Facebook's goal is to get a majority of daily active users.

  The second to last graph actually shows a pretty good case that larger institutions (by enrolment) correlate with search relevancy. But that is not a case study of an institution whose enrolment fell, which is what it would take to make a parallel case with the Princeton paper. They would have to pick an institution whose enrollment fell with its relevancy, and then show that Princeton is on the same track.

  As it stands, it is not "longitudinal" but just a static cross-section of enrollment and mentions. Perhaps enrolment is static at all major institutions, regardless of search relevancy, and their enrollment remains full even if they become irrelevant?

  This small switcheroo is a major one, and shows why the article has to be satirical.

  Of course they did respond quickly :) It seems to indicate that Facebook did some research, but then found their results too weak to publish straight.

• by tedsumme on 1/24/14, 12:54 AM

  Not to support the princeton conclusion but to initiate your debunk by pointing at "like' trends first is silly.

• by nonconfermist on 1/23/14, 11:58 PM

  Maybe all sites on the Internet are getting less attention as the number of things to do on the Internet increases.

• by yetanotherphd on 1/24/14, 4:08 AM

  Very nice. I think the bigger issue is that there is a huge amount of academic research in which a fancy model (usually involving the latest academic fads) is considered a substitute for serious statistics.

  I think Facebook's models with linear time trends are actually much more believable than the original paper.

• by bhartzer on 1/24/14, 4:16 PM

  I think there is a big difference here between Princeton, Harvard, and Yale. Harvard and Yale clearly bought "likes" to their pages, while Princeton kept it clean and thus doesn't have any fake followers or fake page "likes".

• by raverbashing on 1/23/14, 10:54 PM

  Oh and wasn't the Facebook research based on Google trends?

  Meaning: people that can't type "facebook.com" on the address bar and type it into google then click?

  So... based on that, we can assume people are getting more familiar with the internet maybe?

• by Ind007 on 1/24/14, 12:21 PM

  Leaving all these articles, facebook has to die - this is my intuition.

• by joshvm on 1/25/14, 3:23 AM

  Good fun reading that, but.. an R^2 of 0.54? Not a convincing fit.

• by xordon on 1/24/14, 5:56 PM

  There are three kinds of lies: lies, damned lies, and statistics.

• by teaneedz on 1/24/14, 6:35 AM

  Isn't Facebook a carrier of bugs?

• by dynamic99 on 1/23/14, 11:23 PM

  This is just awesome

• by jokoon on 1/24/14, 9:22 AM

  no, I still hate facebook.

• by dapvincent on 1/23/14, 11:38 PM

  "What's beef?"

• by charleswalter on 1/24/14, 8:41 AM

  The fact that Facebook even responded reeks of fear.

  If the original article had been completely ridiculous, Facebook could've laughed it off and wouldn't have had to respond at all.

  It's like if someone tells me I'm fat. Because I'm in good shape, I wouldn't react and just think the person is weird for telling me that. But if I was anywhere close to overweight, you'd see a strong reaction of some kind from me.

  What the reaction is doesn't matter so much as the fact that there is a reaction.

• by lafar6502 on 1/24/14, 6:26 AM

  Sciencey stuff didn't work? Time to call lawyers

• by trollingineer on 1/23/14, 10:45 PM

  Couldn't take this seriously given the writer is just mad he is a Data Scientist at Facebook and not doing research at an academic powerhouse.