When Data Becomes Ubiquitous: Managing the Presence in the City
of Facebook Anja Bechmann,
Head of Digital Footprints Research Group, Aarhus University, DK
PIT Summer School, August 20, IT-‐City, Aarhus, DK
Services become interwoven
Data becomes ubiquitous
RQ: How do users navigate in this interoperable service of Facebook?
Interoperability
• Technical interoperability
• Personal interoperability
Method & dataset a qualitaQve study combining open API data retrieval, screen dump analysis of Facebook apps, and semi-‐structured group interviews on Facebook. N=17 (15 Danish high school students, 2 American College students, all 18-‐20 years old) API data: enQre Qmeline (24,062 data units 116 closed, secret, and open groups (10,213 data units) newsfeed for a 14 days period (41,168 data units).
findings Average of 60 FB apps Typical quizzes, but also MyBirthday Calendar, SpoQfy and CiQes I’ve visited Youtube is the top external applicaQon used in the dataset (videolink) Consider services separate per default in terms of content uploaded (e.g. Twicer)
findings
• The parQcipants do not use privacy sedngs to control interoperability of personal data primarily but they place sensiQve content in either inbox, chat,closed or secret groups.
findings
Coding content in groups
• SensiQve data:
Images more personal than name Death, ilnesses, things they do not want to be confronted with (e.g. relaQonship status)
Conclusion • Facebook as an effecQve communicaQve tool to micro-‐coordinate and socialize with exisQng friends from different arenas.
• The use of Facebook is mostly oriented towards the closed features of inbox, chat and groups.
• ExisQng literature ogen focus on Qmeline -‐> need to focus elsewhere on the more personal features of Facebook as equally (if not most) central hubs of communicaQon.
www.digitalfootprints.dk
Research interest • More fenced-‐off and ubiquitous internet (cross-‐plaiorm/
cross-‐services through login)
• How do we get access to closed data about users on private social networks (e.g. Facebook)?
– In order to analyze user behaviors with FB across websites – User data structures – Analyze navigaQon outside FB but related to FB (checkins) – Analyze usage pacerns during the day (Qmely) – Analyze digital cross-‐plaiorm use of FB (laptop, smartphones, pdas)
– Analyze exposures to content from other website/media
ExisQng methods • In “virtual ethnography” (howard, wicel, marcus, markham, kendall, baym, boyd) – Friending:
• You are not sure to get all acQvity because of sorQng algorithms of Facebook
• You must manually export them to see pacerns over Qme
• Ethnography – Following them physically
• Time consuming • Too much intervenQon in everyday rhytms • But you will get a lot of detail on the context of the acQviQes on Facebook that is not possible to get otherwise
DIGITAL FOOTPRINTS as data retrieval tool
• Act as an external ‘company’/third party when extracQng data from Facebook
• a webbased system
• Using Facebook’s graph API • User consent that DIGITAL FOOTPRINTS draw info on users like any other applicaQon/website using facebook connect
• Users can withdraw anyQme they like • Researchers can mine data from the users and answer research quesQons in qualitaQve/quanQtaQve (?) studies
Digital Footprints
Data extracQons e.g.
• Demographics • Newsfeeds • Network and friends • Likes • Check-‐ins • Private/public groups • Pictures, status updates and uploaded material • Friends material through consent of the parQcipant etc. etc. etc….
Methodological triangulaQon (e.g.) 1. HarvesAng private data with consent, mining these data (DIGITAL
FOOTPRINTS)
2. Focus group interviews with parQcipants to understand their adtudes and strategies
-‐>Digital Footprints can help answer “what” and qualify other methods for
asking “why”
Strengths – Researchers can easily send link via email to parQcipants, asking them to sign up for the
research project – Researchers can access closed data without profiles being public – Data is saved in database which makes it possible to extract and sort different pacerns – Digital Footprints also allow researchers to study the newsfeed of the parQcipants – Researchers can study a variety of Facebook acQviQes in one system
Limitations
– Methodologically, ideally users must be chosen beforehand and asked to parQcipate (external validaQon)
– Difficult to create representaQve sampling/data – Digital Footprints relies on the graph API se3ngs which is controlled by Facebook – Therefore primarily qualitaQve virtual ethnographic tool – Cannot register user traffic pacerns (click-‐through analysis)
Future research
1. (How) can we make data retrieval through Facebook Graph APIs representaQve – how do we recruit for quanQtaQve analysis
• Problems: – RepresentaQve users or certain kind of users that uses this applicaQon – If not applicaQon – certain types of users that has public profiles – What is the Facebook populaQon from which we sample?
• Only soluQon (visible for us): – is to recruit a representaQve sample and then send out the invitaQon to
join?
• What about the ethical quesQon of retrieving friends data as well?
Law • Privacy Law:
– Comply to EU direcQve 1995, 1999, 2002, 2011 (with explicit consent, limited Qme, explicit purpose, only data needed for that specific purpose etc.)
• Danish Data ProtecQon Agency: – Apply for permission to make research project involving personal and sensiQve
user data
• Facebook’s terms of (data) use: – You can only retrieve data you need (data protecQon law) – You cannot redistribute user data to any third party stakeholder – User must be able to delete their data from the research project – Keep info up to date….?? – User’s friends data can only be used in the context of the user’s experience on
your applicaQon…?? (we do not sell it)
Ethics • Issue no. 1:
– When retrieving data friends will comment, like etc. on the parQcipant’s data and therefore be visible in the system
– Working on effecQve anonymizaQon methods before release in October
• Issue no. 2: -‐ Ethically we are interested in the best informed consent -‐ We are working with three/four step consent procedures: verbal, email, invite text, and facebook
consent text
• Issue no. 3: -‐ We are interested in the best possible storage of the data -‐ We are working with a server database model located at au where every project has its own
database structure
• Issue no. 4: -‐ We are interested in varifying researchers as such and only legal research projects -‐ We demand an university email adress from registered universiQes -‐ If working with full idenQfiable data we consider that researchers need to document legally
permission for this (??)