In this paper we propose data augmented ethnography as a novel mixed methods approach to combine ethnographic, qualitative, observations with social media data collection and computational analysis. Using two brief studies on online interaction as examples we discuss the benefits and challenges of the combination of these two perspectives. We posit that the observations made in the qualitative phase can be quantified and hypothesized together with the data collected later during the analysis stage. Through our case studies we aim to shed light to the differences apparent on the party level and seek to understand how candidates, based on their parties political standing, differ in terms of interactivity. We ask, what insights does a mixed-method approach combining ethnographic observations to computational social science offer to the study of interactivity and its many pregnant forms? To answer this question, we use a large data set collected from different social media platforms before and during the 2015 Parliament Election in Finland. This data consists of both textual data including all candidate updates and the conversations they elicited, as well as field notes written and collected during ethnographic field work period before the elections.
Functional group interconversions(oxidation reduction)
Data augmented ethnography: using big data and ethnography to explore candidates' digital interactions
1. Data augmented ethnography:
using big data and ethnography
to explore candidates' digital interactions
Salla-Maaria Laaksonen, Matti Nelimarkka,
Mari Tuokko, Mari Marttila, Arto Kekkonen, Mikko Villi
#digivaalit2015
@jahapaula
2. Context
Finnish Parliamentary Elections 2015
• Computational data collection from Twitter, Facebook and 19
national news media outlets
• Online ethnography (netnography / media ethnography) field work
for 1 month before the Election date
Original research focus
• How candidates and other actors influence media agenda through
social media?
Extended focuses
• Interaction in social media
• Hashtags in election campaigns
• Interest groups during the election campaigning
3. Computational social science
Practices of utilizing computational, algorithmic
methods to study society and social questions.
(e.g., Lazer et al., 2009)
• data collection
• data preprocessing
• data analysis
4. Burning questions
for social science and big data
!
Do you know the context where data was collected?
(e.g. boyd & Crawford, 2012, ; Gillespie 2014)
!
Are you sure your algorithms compute the right thing?
(e.g. Grimmer & Stewart, 2013)
!
What about data selection and cleaning of your data?
(e.g. Ekbia et al., 2015)
5. • Aims to create
understanding and
make sense of
human life and
social communities.
• Usually conducted in
the natural
environments of
human action.
• Field work
• Participatory
observation
• Documents
Ethnography
Picture: Malinowski/
WikimediaCommons
6. • Several different sub-
approaches: webnography,
network ethnography,
netnography, media
ethnography, trace
ethnography…
• Challenges and questions
(e.g. Wittel, 2000; Markham,
2013):
• What counts as participatory
observation online?
• Displacement between
ethnographer and the field,
lack of physical context
• Relationship between the
online and “the real”?
Online Ethnography
Picture: DaveDen /
Uncyclopedia
7. Illustrative example studies: Candidates
& interaction in Facebook
Common theme for political research (e.g.
Stromer-Galley 2000, 2004; Grahamn et al. 2013; Goldbeck et
al. 2010, Larsson 2015, Grant et al. 2010)
RQ1: What kind of differences we observe between
the parties on interactivity in social media services?
RQ2: To what degree we observe negative
campaigning in candidate-candidate interaction?
8. Data and Method
Ethnographic field work
• Observation conducted by three researchers 19.3. – 19.4.2015
• Facebook and Twitter researcher accounts created, Tweetdeck used
for Twitter
• Focus on the forming of the online agenda around the election,
candidate communication styles and interaction with other actors
Social media data sets
• All public Facebook pages of the candidates (n = 1111)
• Collected using a social media monitoring tool 99analytics and FQL
language
• In total, 61 790 updates and 67 956 comments to those updates.
• Extracting the authors of posts and comparison with Kruskal-Wallis
Χ2-test and sentiment analysis using SentiStrenght (Thelwall et al.
2010)
9. RQ1: Party differences
Small parties interact the most: Quantified interaction
levels between large parties, small parties and non-
parliamentary parties are significantly different (p≈0.02).
10. RQ2: Candidate interaction
Engaging other candidates from other parties
with harsh conversation styles (both Twitter and
FB).
The world's highest tax rate is not an competitive advantage. To which country are your
referring here @Calle_Haglund? I would rather concentrate on questions considering Finland
(tweet from the chairman of the Left Alliance, April 7th 2015)
@paavoarhinmaki By saying this I am trying to tell where Finland would end up to if for
example the Left Alliance got to realize their election promises
(tweet from the chairman of the Swedish People's Party, April 7th 2015)
Sentiment analysis: comments made to other
party candidates’ FB pages are more negative in
tone (p≈0.002)
11. Burning questions
for social science and big data
!
Do you know the context where data was collected?
(e.g. boyd & Crawford, 2012)
!
Are you sure your algorithms compute the right thing?
(e.g. Grimmer & Stewart, 2013)
!
What about data selection and cleaning of your data?
(e.g. Ekbia et al., 2015)
Ethnography
enhances
contextual
framing
Human observations
supports the
interpretation
(Methodological
triangulation)
Ethnography
supports data
collection
+ big data sets
allow
generalization
of the results
+ computational
methods allow
bigger data
sets
13. Discussion…
1. Data and observations always remain
incomplete: the data that is visible for an
observing researcher but also data sets,
collected handles or hashtags,; the intention
and nuances of communication
2. Researchers need to be ready to follow
phenomena as they unfold: but elections
have a schedule, riots do not
3. Making some selections needed during the
analysis process
14. #vaalit2015
Thank You!
!
Project home page:
https://www.hiit.fi/digivaalit-2015
Rajapinta blog: http://www.rajapinta.co
We thank Helsingin Sanomat Foundation for providing funding through the project “Digivaalit 2015”.
We also thank Kone Foundation for providing funding through the project “Digital Humanities of Public Policy-making”.