1 / 48

Reporting Data in the Social Media World Panel Title: Practical Applications for Social Media in Survey Research

Reporting Data in the Social Media World Panel Title: Practical Applications for Social Media in Survey Research. Casey Langer Tesfaye American Institute of Physics. It has always been difficult to control the trajectory of published data C an we leverage social media to increase

paxton
Download Presentation

Reporting Data in the Social Media World Panel Title: Practical Applications for Social Media in Survey Research

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Reporting Data in the Social Media World Panel Title: Practical Applications for Social Media in Survey Research Casey Langer Tesfaye American Institute of Physics

  2. It has always been difficult to control the trajectory of published data • Can we leverage social media to increase • Awareness of trajectory? • Accessibility of findings? • Accuracy of interpretations? Problem

  3. Twitter search stream • Women + physics • September + October 2013 Proposed solution- topical case study At the Solvay Conference on Physics in 1927, the only woman in attendance was Marie Curie (bottom row, third from left).Source:http://ontd-political.livejournal.com/10806258.html#ixzz30UyWOvOz

  4. Twitter search stream • Women + physics • September + October 2013 • Find references to survey data • Look for patterns Proposed solution- topical case study

  5. Twitter search stream • Women + physics • September + October 2013 • Find references to survey data • Look for patterns • Goal: practical, data-driven recommendations Proposed solution- topical case study

  6. Corey S. Powell ?@coreyspowell Oct 10 Men and women really do communicate differently...according to an analysis of old Enron emails. https://medium.com/the-physics-arxiv-blog/e7dbd9d6a518 … View summary Reply Retweet Favorite More Olivier DA COSTA ?@olivierdacosta Oct 10 Why Are There Still So Few Women in Science? ¦ NYT - Meg Urry, professor of physics and astronomy at Yale http://nyti.ms/18ROWK9 View summary Reply Retweet Favorite More Physics ?@Physics360 Oct 10 Women: Life Sciences v. the Lifeless Sciences: My current Taki's article reflects upon an impassioned New York... http://bit.ly/19jPruC Expand Reply Retweet Favorite More James Holloway ?@grasshapa Oct 9 “Data Mining Reveals the Emotional Differences in Emails Written by Men and Women” by @arxivblog https://medium.com/the-physics-arxiv-blog/e7dbd9d6a518 … View summary Reply Retweet Favorite More Matthew Feickert ?@HEPfeickert Oct 9 For those applying to grad school in physics, you should seriously consider a department's Female Friendliness. | http://www.aps.org/programs/women/female-friendly/index.cfm … Expand Reply Retweet Favorite More

  7. Twitter data needs a framework or structure • Generalizing conclusions is tricky territory • Completeness • Representativeness • BUT there is much to be learned Working with Twitter data

  8. First goal: transform the data into a usable form Data Structure

  9. Dataset had 1894 lines, which collapsed down to 251 tweets • 66 no url • 185 with url Some basic numbers

  10. Isolate relevant tweets • Some don’t reference survey data or relevant articles Groom the data

  11. I only want references to pieces that report survey data • but how? Grooming text data is challenging

  12. Some Tweets are easy to remove

  13. Some are harder to judge

  14. In a larger scale data set, manual investigations are impossible • Start with a sample, develop a strategy, hone and measure its effectiveness Automating is the way to go

  15. Structure the data

  16. 40% of the tweets containing links led to survey data (n=74) • Only 6 of those tweets linked to original sources Some metrics

  17. How to group Tweets or sources that belong together? Looking for a starting point

  18. Many different ways to look at links • Link in Tweet (lowest # of dupes) • Full link of referred url • Autolink from url Link duplication

  19. By Autotext: 52 of 74 Tweets were in duplicate sets Link duplication

  20. Largest set: 25 Tweets

  21. Date range of Tweets October 3-20 NY Times article

  22. The Viral Myth and Justin Bieber Effect Source: http://www.technologyreview.com/view/512271/researchers-peek-at-the-structure-of-the-viral-internet/ Original paper: http://research.microsoft.com/pubs/176494/diffusion.pdf

  23. Second largest set: 13 Tweets

  24. Let’s Talk about Autotext

  25. Text comes from: • Autotext • Title • photo caption • article content • from the user themselves Tweets come from a variety of sources

  26. Strategies include • Summarizing the story • Personalizing the story • Pulling out key points • Commentary about the article Tweets use a variety of strategies

  27. And yet only one Tweet used the autolink:

  28. None of the autotext included stats

  29. Easy to spot large scale misinterpretations • But how to catch small? • What do misinterpretations look like? • Case study? Corpus? • Comment streams or a variety of searches can provide more data Accuracy

  30. Tweeters are independent content curators • We can’t control what they say or how they say it or even necessarily be aware of all of it • But maybe we can afford the kind of talk we want People of Twitter

  31. Twitter behavior patterned, but far less streamlined than we might have assumed In summary

  32. What are the key metrics?* • Timeline of Tweets • Twitter handles (counts, networks) • Source of text In summary

  33. Consider your goals: • To be credited with analysis • To spread correct information • Make sharing easy • Link to specific article, not a dynamic page stream • Prime headlines, captions, autotext, and links for sharing • Be aware of key players Some guidelines

  34. Monitor Tweets over time • Set up API & web crawler • Generate and check custom metrics (e.g. dashboard) • Store whole articles and compare text to Tweet • Isolate & compare citations of stats? Potential areas for future work

  35. Network structure of dissemination for stories • Find key nodes in your network Potential areas for future work

  36. Email me: Casey Langer Tesfaye: clanger@aip.org Questions? Comments? Suggestions?

  37. Extra slides start here Really? There’s more? I’m not amused.

  38. Can we know our networks? What kinds of searches can or should we monitor and how? How can we set up our research results in the most shareable way? Questions

  39. In the Fall of 2013 the persistently low proportion of women in physics became a hot topic of discussion. This discussion was in part fueled by a series of reports that grew additional lives once they were published, echoing among individual blog posts, news sites such as the Huffington Post and the New York Times, and reverberating into scientific magazines. The published pieces generated countless comments and tweets among readers and interested parties. What follows is an analysis of the interpretations, misinterpretations and trajectories of published findings, anchored in the “women physics” keyword search on Twitter. This goal of this analysis is to generate helpful, data-driven guidelines for researchers who are now forced into a social media space and hoping to stem misinformation and misreports of their results. Talk abstract

  40. Research questions can evolve from the data • Timeline of dissemination • Network of dissemination • Use of autotext • Original sources? • Correct interpretations? Research Questions

  41. Sharelines

  42. Only one account tweeted links to survey data more than once Network analysis?

  43. http://computationalculture.net/article/the-algorithmization-of-the-hyperlinkhttp://computationalculture.net/article/the-algorithmization-of-the-hyperlink Read this:

More Related