Computational Analysis of Digital Communication (CADC)
Summary of all articles
1. Kramer et al. (2014): Experimental evidence of massive-scale emotional contagion
through social networks
2. Van Atteveldt & Peng (2018): When communication meets computation: Opportunities,
challenges, and pitfalls in computational communication science
3. Welbers et al. (2017): Text Analysis in R
4. Heidenreich et al. (2019): Political migration discourses on social media: A comparative
perspective on visibility and sentiment across political Facebook accounts in Europe
5. Mellado et al. (2021): Sourcing pandemic news: A cross-national computational
analysis of mainstream media coverage of COVID-19 on Facebook, Twitter, and Instagram
6. Van Atteveldt et al. (2021): The validity of sentiment analysis: Comparing manual
annotation, crowd-coding, dictionary approaches, and machine learning algorithms
7. Week 3: Xenos et al. (2018): Uncivil and personal? Comparing patterns of incivility in
comments on the
Facebook pages of news outlets
8. Kroon et al. (2023): Advancing automated content analysis for a new era of media
effects research:
The key role of transfer learning
Week 1 – Kramer et al. (2014)
Experimental evidence of massive-scale emotional contagion through social
networks
Introduction
Emotions can be transferred between people without their awareness; research suggests
that longer-lasting moods (happiness/depression) can be transferred through networks.
In this article, the authors test whether emotional contagion occurs outside of in-person
interaction between individuals. Method: reducing the amount of emotional content in the
news feed. Results: when positive expressions were reduced, people produced fewer
positive posts and more negative posts. Conclusion: emotions expressed by others on
Facebook influence our own emotions, which might mean massive-scale contagion via
social networks does exist. This also means that in-person interaction and non-verbal
cues are not strictly necessary for emotional contagion, and observing others’ positive
experience brings people a positive experience.
Emotional contagion might be a result from the interaction, and not the exposure to the
emotion (interacting with a happy person is pleasant, and with an unhappy person,
unpleasant). Exposure to the happiness of other people online may actually be
depressing for us (‘alone together’ social comparison effect). Not clear if nonverbal cues
are necessary for contagion, or that verbal cues alone are enough.
So we need experimental evidence that emotional contagion exists outside of irl
interactions.
Earlier research
Earlier research demonstrated that:
- emotional contagion occurs via text-based computer-mediated communication
- contagion of psychological and physiological qualities has been suggest based on
correlational data for social networks in general
- people’s emotional expressions on Facebook predict friends’ emotional expressions,
,even days later.
BUT there’s no proof that emotions or moods are contagious when experiencer and
target don’t interact directly.
Method
The experiment manipulated the extent to which people (N=689003) were exposed to
emotional expressions by omitting positive/negative posts on their Facebook news feed
(tijdlijn). It tested whether exposure to emotional content lead people to post similar
emotional content (positive/negative). Posts were positive or negative if they contained at
least one positive or negative word (according to word counting system software). No text
was seen by the researchers, only by the software, so consistent with Facebook’s Policy,
so informed consent. Two control groups where random posts were omitted (negative or
positive). Also controlled for the posts of the participants a week before the experiment.
Emotional expression was measured as the percentage of positive/negative words that
were posted by each person. Negativity reduced should mean more positive and less
negative, and positivity reduced should mean more negative and less positive.
After the publishing of this article there was a lot of critical feedback due to the ethical
side of manipulating the test subjects’ data without their (explicit) informed consent.
Results and conclusion
The results show emotional contagion. People with positivity reduced in their feed were
more negative and less positive in their own posts. People with negativity reduced in their
feed were more positive and less negative in their own posts.
This shows:
- emotional contagion doesn’t have to be from interaction, it can also happen through
exposure (because posts on a Facebook feed are not directed towards anyone in
particular, it’s more like ‘overhearing’ a conversation).
- emotional contagion does not require nonverbal behaviour, text alone seems to be
sufficient.
- the lack of difference in effect size proved that the participants’ response was to the
friend’s emotion expression, and not a response to the news itself that the friend shared.
- exposure to fewer emotional posts lead to participants being less expressive overall;
they were more emotionally positive to positive posts (this shows that social comparison
is not a thing).
The effect sizes were small but this doesn’t mean it’s unimportant, since that still would
have corresponded to hundreds of thousands of status updates per day. The manipulation
of the independent variable (feed) was minimal, the dependent variable was difficult to
influence (emotional expression/ mood).
Week 1 – Van Atteveldt & Peng (2018)
When communication meets computation: opportunities, challenges, and
pitfalls in computational communication science
Introduction
More computational methods used for communication science because of three
developments:
1. More digitally available data (social media, web archives, historical archives, digital
newspapers).
2. Improved tools to analyse this data.
3. Emergence of powerful and cheap processing power, and easy to use computing
, infrastructure (computers with sharing platforms are accessible to most).
These three developments can boost progress in communication science, if we overcome
the technical, social and ethical challenges involved.
Computational science studies generally involve:
- large and complex data sets,
- consisting of digital traces and other ‘naturally occurring’ data,
- requiring algorithmic solutions to analyse,
- allowing the study of human communication by applying and testing communication
theory.
Computational methods complement existing methods. They allow us to analyse social
behaviour and communication in new ways and have the potential to radically change our
discipline at least in four ways:
1. From self-report to real behaviour
2. From lab experiments to studies of the actual social environment
3. From small-N to large-N
4. From solitary to collaborative research.
1. From self-report to real behaviour
Digital data allows us to measure actual behaviour in an unobtrusive way rather than self-
reported attitudes or intentions. This can help overcome social desirability problems, and
does not rely on people’s imperfect estimate of their own desire and intentions. It can
also help overcome the problems of linking content data to survey data (combineren van
digitale data met vragenlijsten).
2. From lab experiments to studies of the actual social environment
We can observe the reaction of persons to stimuli in their actual environment rather than
in an artificial lab setting. Crowdsourcing platforms make it easy to source very large
amounts of (often free/cheap) participants. Dangers of social media research: ethical
problems (privacy, informed consent), time-consuming to coordinate, social media
companies nowadays are more reluctant to share data because of ethical issues.
3. From small-N to large-N
Increasing the N can enable us to study more subtle relations or effects in smaller
subpopulations. But to leverage the more complex models afforded by larger data sets,
we need to change the way we build and test our models (like using techniques
developed in machine learning research for model selection and model shrinkage).
4. From solitary to collaborative research
Computational methods make it easier to share and use resources like (clean) data, tools
and scripts – they have no choice, because the process has become too complicated to
start from scratch for every study. This will also force us to be more rigorous in defining
the research process (because we’re sharing it with others so it has to be the same for
everyone), furthering transparency and reproducibility of research.
Computational methods also foster the interdisciplinary collaboration needed to deal with
larger data sets and more complex computational techniques (sharing methods among
research fields, and across quantitative and qualitative research).
Challenges and pitfalls in computational methods
Keeping research datasets accessible: Sometimes datasets are only accessible to a
small group of researchers. Not only is this unfair, these datasets are also usually based
