Based on joint work with Bernhard Rieder, UvA
Presentation at the Politics of Big Data conference at King's College London, May 8
http://www.politicsofbigdata.net/
Based on joint work with Bernhard Rieder, UvA
What counts in social media? - Politics of Big Data conference
1. What
counts
in
social
media?
Politics of Big Data – Conference & Masterclass
Kings College, May 08 2015
Dr. Carolin Gerlitz - University of Amsterdam
2. Which data matters?
• Data critique often focuses on
calculation (Callon & Muniesa 2005):
the recombination of data-points.
• Second order metrics: scores,
recommendations, rankings,
sentiment, derrivatives, dashboards.
• But what do the first order metrics
that feed such composite metric make
countable and comparable in the first
place?
• Based on joint work with Bernhard
Rieder.
3. Becoming data-point
• Empirical research: ex-post classification.
• Digital media come with specific grammars of
action (Agre 1994) which invite & capture user
action in a standardised form.
• Grammars naturalise distinct use practices into
comparable data points, making heterogeneous
qualities countable and commensurable
(Espeland & Stevens 1998).
4. • Activities can come with different
intentions (Gerlitz & Helmond
2013).
• Interpretative flexibility build into
platforms (van Dijck 2012) allows
for resignification &
transformation.
• Multiple meanings may lead to
more data.
One number, multiple
meanings
5. • Platforms are increasingly being
accessed through clients,
automators, mobile interface or
cross-syndication practices.
• Platform-interoperability (Bodle
2012) & programmability: allow
for various ways of engaging with
and producing content.
One number, many
platforms
6. Repurposing digital
methods
• What lures behind social media
metrics and what animates
them?
• How to use digital research
methods not to repurpose but to
re-embed first order metrics?
• Example: Twitter.
• Twitter Capture & Analysis Toolkit
(DMI-TCAT).
7. 1% sample
• Ongoing project on 1%
random Twitter sample with
Bernhard Rieder (2013).
• Retrieved via Twitter
Streaming API.
• 1% sample as cross-section
on Twitter practices.
Links
Hashtags
The Data Set
1% Random 1% sample 14-20. June 2014
Mentions
Retweets
Replies
16.8
15.8
58.1
32.9
18.2
Tweets
Users
31.707.162
14.313.384
8. Decomposing metrics
• Starting point: source metric.
• Proliferation of access points to
Twitter: web, mobile, clients,
automators, cross-syndication,
custom clients.
• 72.000 sources in our sample.
14. • More nuanced account of non-
human activity beyond the notion
of ‘bots’ (Wilkie et al. 2014).
• Organic & automated content:
cross-syndication, scheduled
tweets, in-game tweets, automated
action, bots accounts.
• Approach to automatisation
beyond data-cleaning.
Dealing with the non-human
15. • Sources allow for different
regimes of being on Twitter:
alternative use practices,
grammars & politics.
• Data-formats/practices of Twitter
informed by data-formats of third
parties.
• Platform-interoperability (Bodle
2012) & -programmability:
technique of commensuration.
Dealing with
platform ecologies
16. The happening of
commensuration
• Commensuration not only a media or
metric effect.
• Distributed accomplishment: use
practices, platform interoperability,
hijacking, spam, humans, bots.
• ‘Happening’ (Lury & Wakeford 2012):
relational, dynamic, distributed.
17. Lively metrics
• What a metric counts is not
predefined by comparable grammars
of action.
• Subject to distributed
accomplishment, invite users & third
parties to write themselves into them.
• Lively metrics: realised differently,
subject to change, happening.
• What counts? Non-objective, dynamic
& situated.
• What can be counted counts (Badiou
2008): need for debates on
commensuration.