Wiki
Social Graphs and Interactions

Final Project
Page last edited by Sune Lehmann Jørgensen (sljo) 27/11-2014

Overview

The purpose of the final project is to measure what makes human Twitter accounts susceptible to follow social bots. Since we're currently infecting Twitter with an army of artificially intelligent social bots, we have a unique opportunity to measure which properties (e.g. age, # of followers, types of hashtags, etc) that make human accounts susceptible to follow social bots.

 

The data material will be accounts you have attempted to follow, both successfully and unsuccessfully.

 

The final project is your chance to show what you have learned during the course, so I expect you to use (some of) the skills you have learned during the course, along with basic math and statistics to figure out what characterizes susceptible Twitter users.

 

Literature

There is some literature on the topic. You may use the papers listed below (or anything else you can find on twitter) in order to get ideas on what to measure

  • C. Wagner, S. Mitter; C. Körner and M. Strohmaier. When social bots attack: Modeling susceptibility of users in online social networks. In Proceedings of the 2nd Workshop on Making Sense of Microposts (MSM'2012), held in conjunction with the 21st World Wide Web Conference (WWW'2012), Lyon, France, 2012. [ Link]
  • S. Mitter, C. Wagner, and M. Strohmaier. A categorization scheme for socialbot attacks in online social networks. In ACM Web Science 2013,  May 2-4th, Paris, France, 2013. (Extended Abstract) [ Link]
  • Emilio Ferrara, Onur Varol, Clayton Davis, Filippo Menczer, Alessandro Flammini. Rise of the social bots. [Link]
  • [Presentation]. Predicting susceptibility to social bots on twitter [ Link]

 

Final project, part A. Five minute presentations

Present your bot and your plans for understanding susceptibility. In a 3-5 minute presentation for the class. There are two options for presenting. This assignment is due Tuesday November 11th 11:59.

 

Option A

Is a 3-5 minute movie. The movie should

  • present your bot,
  • contain an outline on the data you’ll need to get to your goal,
  • explain the central methods & strategies you will use,
  • display your implementation plan.

But other than that, there are no constraints. And we do appreciate funny/inventive/beautiful movies, although the academic content is most important. Note that we'll display the movie to the entire class. Evaluation will reflect how well you have managed to achieve the requirements above.

 

Option A*

If you don't like movies, you may also do an old fashioned presentation in the Ignite Format. The talk must contain same elements as the movie described above.

  • The talk will be max 20 slides.
  • We will auto-forward each slide after 15 seconds.
  • Slides must be in pdf format. We will not accept slides in PowerPoint or Keynote formats.

It is possible to include a lot of material into 5 minutes of presentation, so you’ll need to put some effort into developing your idea. We highly recommend practicing the talk before you show up so you’re used to the autoforwarding, etc.

 

Final project, part B. Written report

Is a short report. There is no fixed length - the general rule is that it should be as short and precise as possible (we have to read a lot of pages, so good, precise writing is rewarded). The report should also be long enough to clearly explain what you've done - we cannot give you credit for work we don't know about. In practice this usually translates to between 6 - 10 pages in the ACM SIG Proceedings Template (see below), but shorter/longer is also OK if there is a good reason. 
 
The report should contain the following sections (not necessarily in the order listed below)
  • Introduction. Introducing the idea of the course and your bot's personality.
  • Implementation. Describe the workings of your final project bot (how have you used the tools from the class, how did you implement the tools in Python - think flow-charts)?
  • Interventions. Briefly describe your original tweets and timing strategy (for tweeting and retweeting) for the various interventions.
  • Susceptibility analysis. This is the central part of the report. Present the analysis related to your susceptiblity experiments. Include the following sub-headings.
  • Statistics. Report the central data - it's a good idea to use plots as well as tables.
  • Theory. Which tools from the class (network science, natural language processing, machine learning) have you used? 
  • Conclusion. What have you learned from the course? Would you do anything differently? Did your strategy work? Which advice would you offer to prospective bot designers.

Format:

  • The reports should be formatted according the standard ACM SIG Proceedings Template (http://www.acm.org/sigs/publications/proceedings-templates).  
  • THE ASSIGNMENT MUST BE IN PDF FORMAT.
  • Be precise, write in objective language (avoid: "I think ...", "In my opinon...", etc) - if you make an observation, support it using data.
Support: +45 45 25 74 43