You are on page 1of 4

EXCITABLE TWEETS: SOCIAL COMPUTING AND ONLINE SEXISM

My name is Stamatia Portanova and I am a Research Fellow at the Department of Human and
Social Sciences, Università degli Studi di Napoli ‘L’Orientale’, where I am working on a three-year
project entitled ‘Software between critical theory and sociocultural practice’, under the supervision
of Professor Tiziana Terranova. (slide 2) I am here today as a member of the TRU (Technocultures
Research Unit, http://www.technoculture.it) of the Orientale, a research group that is part of the
larger Centre for Postcolonial and Gender Studies. The purpose of this research unit is to theorize
the connection between culture, technology and technics, a connection that is particularly evident in
digital media, computation and information networks. (Problems described so far are not only
mathematical but social and cultural, and therefore we think they need to be analyzed not only
multimodally, but in interdisciplinary collaboration)

What is the aim of my presence at this conference? (One component missing) The abstract of our
paper is entitled ‘Excitable Tweets’, and promises to deal with the issue of hate speech online,
particularly in relation to mysoginist forms of expression and violent behavior against women on
social media platforms. (slide 3) The first objective for me will therefore be to present a particular
research methodology, broadly defined as ‘Cultural Studies’, that could be quite significant in the
analysis of this online phenomenon. (slide 4) I (but it is more a ‘we’, a group, that is speaking with
me) would like to propose the adoption of such methodology, in interdisciplinary collaboration with
data science and social research. (back to slide 3) Cultural Studies was originally the name of a
group of researchers based at the Birmingham University that, from the 1950s, started to think and
write about the notion of culture (more officially, the group inaugurated a Centre for Contemporary
Cultural Studies in 1964). Their specific interest was in fact the notion of ‘popular culture’
(different from the more traditional concept of ‘high culture’). Why is all this important? Because
among the main topics of research, already from the first works that were written and published by
these theorists, were the stereotyping practices operating in various popular media of the time.
Particular attention was devoted to what we now call ‘old media’: print, radio and television, and
their representation of class, but also of gender and racial identities. It was therefore a kind of
research that still seems resonant with what we have been discussing here today, and especially with
the social analysis of data about the stereotyped representation of women on Twitter.

(slide 5) From the beginning of its research activities, the CCCS emphasized the difference between
its ‘qualitative’ research methods, and the ‘quantitative’ methodology used by sociology (a
methodology usually defined as ‘content analysis’), for the study of social phenomena. On the one
hand, content analysis assigns specific labels to the communication objects it analyzes (a newspaper
article or a TV show can be for example labelled as mysoginist and stereotypical), and then applies
statistical methods to the analysis of various labelled samples, in order to systematically identify
patterns of mysoginist communication
(https://pdfs.semanticscholar.org/4682/b9942712c00d48f819e9039023c9bae4ede6.pdf,
https://journals.sagepub.com/doi/pdf/10.1177/2158244018769755). (slide 6) On the other hand, CS
adopt a methodology mainly based on ‘close reading’, ie the reading of each media and
communication object as a text, and the decoding of its signification or meaning (a method directly
derived from the study of literature). (Ex: JP Gaultier perfume ads, corresponding to the proposed
definition of stereotyping as ‘a widely held but fixed and oversimplified image or idea of a woman;
description of women’s physical appeal and/or comparisons to narrow standards’.) (slide 7) CS, in
other words, pay attention to the structure, to the construction of a certain discourse, for example in
images, but also in verbal language, and its influence on processes of subjectivation. And yet, it is at
this point important to underline that CS cannot be defined as a precise discipline (it is not a
‘science’), but identifies itself with its different objects of study (sex and gender, race and ethnicity
studies etc). Depending on the object, different methodologies can in fact be adopted, including the
collection of data and their empirical analysis (tight relation with sociology, also ethnographic
method).

(slide 8) Despite their tight relation, CS and social research cannot be considered as one and the
same thing. How can their difference be summarized? In their book on Data Science and Social
Research, Enrica Amaturo et al. express some methodological concerns regarding the use of big
data as a starting point for doing sociological research. In this relatively new field, the analytical
work is often run by algorithms and through machine learning, using the latter for data mining to
detect, classify and segment meaningful relationships and associations, or trends between variables
(such as with natural language processing). In relation to this method, Amaturo (and social
scientists in general) suggest that more attention is needed on how data are collected, organized,
integrated and interrogated, by whom and for what purposes, as opposed to the apparent neutrality
of algorithmic automation. Natural languages in fact need particular semantics and taxonomies to
recognize patterns and extract information, and analytics are always the expression of an
epistemology. In relation to this sociological concern, CS would recognize the importance of the
epistemological question, but would also reveal a further perplexity about the possibility of
associating determinate socio-demographic labels to social media usage by groups (eg the use of
Twitter by Black and Hispanics in the US, as an argument of various sociological researches). This
further perplexity, I think, shows us one of the most important differences between the sociological
and the cultural approach. Such differences are not highlighted here as inevitable and unfillable
gaps, but more as important variations of point of view, that can indicate the importance of a
transdisciplinary collaboration.

(slide 9) While the most recent version of content analysis corresponds to data science, ie the
automatic detection of particular social phenomena and categories and their computation (in their
article on Multimodal classification of sexist advertisements, Gasparini et al. for example, describe
the labelling and categorization of stereotyping, objectification and dominance, as different types of
sexist ads, and then their quantification, by particular algorithms), CS have developed a relatively
new field that is usually defined as Media or New Media Studies, in which the study of media
messages and their meaning is often associated to their affect (see the so called ‘affective turn’). If
we tried to visualize the two respective, quantitative and qualitative methods of analysis described
so far, we would see that the first level (content or data analysis) corresponds to a sort of remote
viewing that shows us the magnitude or the big picture of phenomena as if from above, from a
certain distance, highlighting their statistical aggregates and the patterns to be distinguished in
them; while the second level would correspond to a zooming in the singular, specific case. As an
example, let us think again of the analysis of hate speech online. In the last years, a series of racist,
sexist, homophobic and transphobic discourses and feelings has been overflowing in all social
media. In particular, while a wave of transnational feminism is continuously growing and offering
new models of social relations, social media are increasingly becoming the theatres of real attacks
against feminist and queer influencers, self-organized groups and information websites. Movements
such as Alt right, MRM at a global level, as well as their local Italian nodes, have been greatly
investing in social communication strategies, including the use of bots, sentiment analysis, fake
accounts, and data analytics. As a result, they have managed to give the impression of being
hegemonic in the social communication environment, to the point of launching a series of swarm
attacks. In some cases, these attacks have resulted in the closing of accounts and groups, such as the
Non Una Di Meno Milano page in 2017. (show) The adoption of computation as an instrument to
analyze and also to potentially prevent, or at least to start to deal with, this problem, has allowed
various counter-information groups and associations to use data mining and data analytics as useful
tools for the study of such phenomena, providing us with a huge amount of information and helping
us to find significant patterns, or types of coherent content with particular semantic characteristics,
among the collected data (see also the use of Deep Text by Instagram, for cyberbullying
prevention). And yet, from a ‘cultural’ or ‘technocultural’ point of view, other questions seem
worth being posed, for example about the effects, or affects, of these ‘flames’ as they are perceived
by women, feminists and queer activists. What kind of language processing algorithm could really
help us to throw light, or to make us perceive, what queer philosopher Judith Butler describes as the
‘wound’ that is performed, and expressed, by language? Such questions could only be addressed by
adopting a closer observation (close reading), and by considering the semiotic and affective
singularity of each attack. (slide 10) What I (we) am proposing here is therefore the weaving of a
methodological relation, to connect the two levels of analysis instead of keeping them separate (as
they have been so far considered by the sciences and the humanities alike, noise). Accordingly, as
media theorist Lev Manovich would put it, the “analysis of larger patterns” (in this case, of violent
social and linguistic behavior) can be combined with the study of individual details and of single
case interpretations, in order to take the analysis beyond the limits of pure empirical exploration.
This method would also allow us to undertake a careful study of individual attack cases in order to
arrive to wider, more general reflections about the larger social system (or network) behind them.
(slide 11) (Show Humboldt Map)

How? (slide 12) Technology, technics, and specifically the software, or in other words the analysis
of the apparatus, the medium, can indeed offer to us a good connecting link, in order to achieve a
sort of methodological triangulation between the different levels. The reason why I (we) am
suggesting such a ‘techno-oriented’ point of view is related to the ‘double face’ nature of the
medium itself: one face towards the data as they become accessible (APIs), one face towards the
user (GUI). Or, in other words, one side of the tool looking at the patterns, the other side
considering the interpretation and effect on users. Technology, in other words, could act as a
collaborative connecting agent.

The future is open and unpredictable.

Thank you.

Amaturo
Epistemic point of view.
Revolution: objectivity, induction, computation.
1= big data as windows on social reality
2= exploring data is more faithful than building models
3= prevision and application to databases

Debate guided by discussion on methodological innovation (eg native vs digitized). Consequence of


technodeterminism: technology as new, social science as old.

In fact, there is a continuity of social sciences methods.

More correct to talk about ‘epistemic evolution’ (new digital methods).

What’s new?
Intersubjectivity (platforms as methodological dispositifs, data are produced), abduction (Pierce, no
end of theory), mixed methods.
a) Materials and objects. Data assemblages are composed of technical and cultural processes
(data are constructed). Researching into these assemblages unpacks the black box. (Giddens,
double hermeneutic circle).
b) Inference of best explanation (Bayesian method, maximum likelihood). Not data-driven but
data-informed.
c) Digital as context, object and strategy. Continuum vs oppositions.

Epistemology of the digital: critical and creative (acknowledge the role does not mean that
technology guides social knowledge). Pluralist method (eg: psychological questionnaire).
Pragmatic: difficulties.

Digital data as set of confused fragments of human reality (Mills).

You might also like