IC2S2 2019 round-up
The baton was passed to the University of Amsterdam for the fifth addition of IC2S2 with the core conference taking place between 18-20 July. An interesting aside from the opening ceremony was the scientific discipline migration network of the 480 participants at IC2S2, based upon the original and current disciplines delegates entered when registering. You can also check out what the IC2S2 programme looked like as a network put together by André Vermeij.
Thursday 18th July
The two opening keynotes were given by Emma Spiro and Ken Benoit. Emma Spiro explored how understanding the spread of rumor, fact and misinformation during disasters helps with improved emergency responses. If these teams can better track the impact of a disaster on different communities then it can target information and resources with more focus. However, if online becomes the primary focus of communication then which communities will miss out as a result.
Ken Benoit spoke of two fields divided by common methods—social science and computer science. He advocated for better models and less denial. ‘We need more complicated models to deal with more complicated problems’ referring specifically to text analysis not least because people talk in phrases and not in words. Ken went on to present a tongue and cheek look at how computer science and social science differ when analyzing textual data. Social scientists treat text as data and use it to test older theories and try and work something out about the social actors generating the text. To computer scientists text is material to be explored and it’s more about predicting, moving fast and breaking things. You can check out the full slides from this talk here.
What a brilliant keynote talk by @kenbenoit @IC2S2.
— Walid Magdy (@walid_magdy) July 18, 2019
I think the moral of the talk is that social scientists need to speak more with #NLP researchers to create dedicated text analysis algorithms for the purpose of more accurate computational social science studies.#IC2S2 pic.twitter.com/P81wTA9Qoc
Following lunch we heard from Facebook about their lessons learned from the 2016 US election and their subsequent efforts to protect elections. They were keen to demonstrate that they’ve used rigorous methodologies to understand and tackle misinformation on Facebook. Monica Lee ran us through everything that they’re doing to prevent false amplification of stories, elimination of inauthentic accounts, better verification of political advertisers and the extensive Facebook Ad Archive where all political adverts are kept for 7 years.
When it comes to how people decide what is false news and what isn't, Devra Moehler showcased the multi-method research that went into their Candidate Connect product, working with different methodologies across India, Brazil, Indonesia and the US. The goal of the product was to build users knowledge about candidates and the voter journey; the outcomes were an improvement in knowledge in India and Indonesia, but not in the US.
FB strategy for defending Against Election Abuse by Devra Koehler and Monica Lee
— Omena (@JannaJoceli) July 18, 2019
1. Cracking Down on Inauthenticity
2. Bringing Transparency to Political Advertising
3. Reducing Distribution of False News
4. Supporting an Informed Electorate #IC2S2
Caroline Nevejan Chief Science Officer of the city of Amsterdam
In the evening we all dodged the rain—some more successfully than others—to gather at the nearby Hotel Arena, set in the beautiful grounds of Oosterpark.
Caroline Nevejan Chief Science Officer of the city of Amsterdam welcomed guests and gave an enthusing talk on how the city of Amsterdam are dealing with our now data intensive society.
Friday 19th July
Adam Pah, Clinical Assistant Professor at the Kellogg School of Management and Organizations at Northwestern, incidentally last year’s IC2S2 location presented research into gun violence in American schools. Pah touched upon a subject that would be widely discussed later in the day, namely as social scientists we are not here to simply put random data sets through models and then just leave it there, instead we need to dig in and engage with the data, asking difficult questions about its quality and relevance, even if this takes longer, is morbid or difficult work, and doesn’t give us the answer we wanted.
Duncan Watts continued this question of delving deeper; thinking about the practical application of research to make real world impact.
Duncan Watts - "For 20 years I thought my job was as a basic scientist. Publish papers and throw them over the wall for someone else to apply. I now realise that there's no one on the other side of the wall. Just a huge pile of papers that we've all thrown over." #IC2S2
— George Richardson (@g_r_richardson) July 19, 2019
Watts discussed the past, present and future of computational social science asking whether social science should be more solution-orientated, rethinking traditional models of social science routed in the number of publications and citations. ‘Instead we should be looking to measure impact in terms of the practical knowledge gained’. He touched on the growing interest of publishers and funders in computational social science and the increased training opportunities now available to researchers. There were also areas of slow progress, researcher’s relationship with tech companies and data privacy concerns.
As the talk moved into the panel session, Helen Margetts stressed the importance of bridging the divide between tech companies and academia. Sandra González-Bailón spoke about the importance of interdisciplinary research: ‘in this field, your institution or background doesn’t matter, it’s the problems you’re interested in solving that unites us’. It felt a good time to reflect on these big issues as it marked five years since the first IC2S2 in Helsinki and ten years since David Lazer et al redefined the field in 2009.
Saturday 20th July
Petter Törnberg gave a fascinating presentation on his work into Gender advertising on Instagram, using image recognition and text analysis to explore Goffman’s ideas of gendered advertising. Solomon Messing led a panel also featuring Rebekah Tromble and Arjun Wilkins discussing Social Science One. After hearing of the slow progress made between researchers and tech companies expressed on the Friday, a chance to hear more about Social Science One was welcome. The three core challenges of the project: protecting user privacy, ensuring accurate and meaningful social scientific results, and maintaining the independence of academic researchers. Rebekah Tromble was honest in her outlook that it’s going to suck for a while whilst researchers work out what core problems they want to address with this data from Facebook. The key unresolved issues according to Solomon Messing are: replicability, differential privacy, squaring privacy with open science and what datasets would be most valuable.
Summary
IC2S2 is always a highlight in our calendar and this year’s event did not disappoint. It’s always so fascinating to get a chance to discuss all things computational social science with such a diverse range of academics and industry professionals. Thanks to all the organizing committee for putting on such a brilliant conference in a beautiful setting. We hope we’ll see you at IC2S2 next year in Boston, USA.
Resources
List of essential skills for aspiring computational social scientists to invest in by Rense Corten.