DCS/CSCI 2350: Social & Economic Networks - Mohammad T. Irfan Events

4/24/18

        DCS/CSCI 2350:
 Social & Economic Networks

Privacy & Influence in Online Social Networks:
Case Study of Facebook & Cambridge Analytica

                  Mohammad T. Irfan

                          Data &
         Events         computation

                 Issues &
               Implications

                                                      1

4/24/18

            Events
    (and behind the scene)

 What happened (2014 – 2016)

                              Political
                             Campaigns

Aleksandr Kogan
Professor at Cambridge U

                                               2

4/24/18

Timeline
    2010          Launches Open Graph API
                  for app developers

    2011          Agreement with FTC:
                  Consent for sharing user data

    2013                     CA founded in the UK
                             to operate in the US

    2013             Makes a Facebook personality
                     quiz app for academic research

    2014             Forms a company (GSR) to
                     harvest Facebook data

    2014             Sells data to

2014–16              Works for

How it came to light
u   Christopher Wylie: whistleblower

u   March 17 & 18: sensational reporting by Carole
    Cadwalladr on NYTimes and the Guardian
     u     https://www.nytimes.com/2018/03/17/us/politics/cambridge-
           analytica-trump-campaign.html
     u     https://www.theguardian.com/news/2018/mar/17/data-war-
           whistleblower-christopher-wylie-faceook-nix-bannon-trump

                                                                            3

4/24/18

Behind the scene
u   As told by Carole Cadwalladr
    u   https://www.cbsnews.com/news/cambridge-
        analytica-channel-4-expose-facebook-users-
        monday-night/

Behind the scene:
Prof. Kogan & Cambridge University
u   Prof. Kogan: blamed by Facebook for violating ToS
u   University colleagues upset
    u   ❝Prof John Rust, the director at Cambridge University’s
        Psychometrics Centre, … accused Kogan of trying to make
        $1m in “personal profit in terms of asset and data” from
        the scheme, while only reimbursing his fellow
        psychologists, Dr Michal Kosinski and Dr David Stillwell,
        who had led much of the cutting-edge research, with
        $100,000.❞
    u   https://www.theguardian.com/education/2018/mar/24/cambridge-
        analytica-academics-work-upset-university-colleagues

u   Prof. Kogan’s defense
    u   http://money.cnn.com/2018/03/20/technology/
        aleksandr-kogan-interview/index.html

                                                                            4

4/24/18

Behind the scene:
Cambridge Analytica (CA)
u   Alleged business practices (denied by CA)
    u   https://www.youtube.com/watch?v=mpbeOCKZFfQ

Behind the scene:
Facebook
u   August 2016: Facebook lawyers asked all
    parties to delete the data
u   Everyone certified deletion
u   But did they actually delete it?
    u   March 2018: News channel obtains data of 136,000
        Colorado residents
    u   https://www.channel4.com/news/revealed-
        cambridge-analytica-data-on-thousands-of-
        facebook-users-still-not-deleted
u   Facebook kept denying data/trust breach till
    March 2018

                                                                5

4/24/18

Data and Computation

What does the data look like?
u   Courtesy of Prof. David Carroll
    (Parsons New School)
u   Voter data
    u   “accurate and mostly complete for me”

                                          Partial snapshot

                                                                  6

4/24/18

What does the data look like?
u   Predictive scores for Prof. Carroll
u   “Feels roughly accurate.”

But… is this his real data?

                                               7

4/24/18

What algorithms are used?
u   Leaked codes from Aggregate IQ
    u   https://www.upguard.com/breaches/aggregate-
        iq-part-one
u   Data collection
u   Algorithm to go from quiz-takers to their
    friends, friends-of-friends, etc.
    u   75 million nodes starting with just 250,000

From data to prediction
u   Computational problem: Given Facebook
    data, predict the OCEAN scores
    u   Openness, Conscientiousness, Extroversion,
        Agreeableness and Neuroticism
u   Machine learning

                                                           8

4/24/18

  Machine learning: big picture

                   Training data: Facebook activities
                        and "true" OCEAN scores

                           Machine learning
                              algorithm

    Test data:                 Model of
                                                         Prediction:
Facebook data of            classification/
                                                        OCEAN scores
  a new person                regression

                    Issues and
                   Implications

                                                                            9

4/24/18

Issues
u   Privacy in digital life
u   Psychographic modeling
u   Computational issues: local à global
u   Microtargeting and influence in online world
u   Political campaigns
u   Ethics in research
u   Legal issues
u   Public relations

Writing assignment
u   The Tipping Point
     u   Give mathematical models
     u   Extend existing models
u   Is Gladwell's thesis valid in the online
    world?
     u   Three laws and how/whether they apply
         to our digital life
          u   Law of the few– connectors, mavens,
              salesmen
          u   Stickiness factor
          u   Power of context

                                                        10