List of Figures:
Figure 1 : Hashtag categorisation per year
Figure 2: Top ten Hashtags per year National Database
Figure 3: privacy mentions by countries
Figure 4: Vorratsdatenspeicherung (data preservation) in Germany
Figure 5: hashtag co-occurrence
Figure 6: hashtags by country
In order to study activist-led Twitter conversations on European digital rights, we have built two datasets, relying on key national and one EU-level advocacy organization - EDRi (European Digital Rights Initiative) - as starting points. The data was analyzed using the DMI-TCAT (Borra and Rieder 2014), specifically addressing research questions.
The first data set (“NATDIGORGS”) was composed of tweets mentioning the twitter handles of member organizations or observers of EDRI as keywords (ie: not including the “@” sign). The data set only included mentions of EDRI members/observers who are also based in Europe (member organizations or observers that operate transnationally are not included, accounts for VIBE and xnet were also exclude due to significant noise associated with their twitter handles). Keywords used to compose this data set included bitsoffreedom OR chaosupdates OR DRIalerts OR digitalcourage OR digiges OR dfri_se OR Effi_ry OR FITUGeV OR iure_cz OR netzfreiheit OR mensenrechtenBE OR fmeta OR OpenRightsGroup OR panoptykon OR privacyint OR quintessenz_AT OR StatewatchEU OR laquadrature OR AKVorrat_at OR NURPA OR ShareConference OR akzensur OR SOITsk). The resulting corpus contains 104K tweets and their metadata.
The second data set (“EDRI”) contains tweets mentioning EDRI (the keyword “EDRI). The data set included in significant noise because the term “edri” is part of a phrase in some languages, also a common proper name in several places. However, we observed a significant reduction in noise from 2012. The reasons are, however, unclear.
Clearly state your research questions and any hypotheses / expectations.
In order to analyze the discourse of the European digital rights activism we posed the following questions:
Which actors can be seen as most important or influential in the twitter conversation about digital rights?
How does this conversation differently manifest across different countries (with a specific focus on France, Germany, Great Britain, Netherlands)?
What is the role of Twitter for the enterprise of digital rights activism in Europe
for the organizations;
as methods window on the issue space; and
how is it limiting for this case?
TCAT was used to conduct multiple analyses for specific groups of hashtags and user mentions, querying the data sets across times and at 1 year and 6 month intervals. Other tools were also used, including Gephi for network analysis and visualisation (Jacomy, M. et al., 2014), RAW for data visualisation, and the DMI data triangulation tool.
Coding the hashtags (Sub-Group Hashtag Categorisation)
Twitter hashtags may represent many different rhetorical functions ( Daer, Hoffmann and Goodman, 2014). In order to analyze the data in a more qualitative approach, a set of categories was created in an iterative form, so we could (1) take a closer look at the hashtags that relate to concrete issues and problems, (2) to understand the relationships between more concrete issues and general problems and between themselves and (3) understand the different uses of the hashtags over time.
The main types of hashtags identified are illustrated in the table below:
Type | Description |
Policy Issue | Concrete problems (e.g. ACTA, VDS) |
Problem Area | General concerns (e.g. privacy, surveillance) |
Event | Conferences, workshops, and meetings (e.g. shareconference) |
Organization | Governments, NGOs, civil society associations (e.g. ccc, GCHQ) |
Location | Actual place or natural setting (e.g. UK, Berlin) |
Campaign | Work in an organized and active way toward a particular objective, like crowdfunding a NGO (e.g. 1dag1000donateurs) |
Technology | Technological tools, applications, softwares (e.g. touchid, VPN) |
Others | Things which can't be classified within the previous categories and don’t seem relevant for the analysis as a category, such as expressive hashtags or others (e.g. ff, citizenfour) |
Country-specific and cross-country analyses (Sub-Group Countries)
We used the country-specific EDRI-members as a starting point. We first categorised our initial list of names of EDRi members/observers by country (Germany, UK, Holland, and France). For each country, we put the names in TCAT’s query box and used the “user visibility” module to generate sub data sets including all mentions of those uses, as a proxy for national conversations about digital rights. Within this subset we identified users who had more than a 100 mentions so as to generate a list of influencers within each national discourse. Based on the fact that the data-set is covering six years worth of tweets, 100 mentions was chosen as a threshold.
The hashtags that were most associated with these users, were considered as the top hashtags for the country-specific organizations, and understood to represent dominant “issues” in each national discourse. We employed the bipartite hashtag-user graph to explore the relations between users and hashtags and identify which hashtags are common and unique for each specific country. Using the digital method tools DMI-TCAT, Gephi and RAW, we visualized our results in an alluvial graph. The triangulation tool was used to determine which issues were most prominent to all national discourses, and whether hashtags prominent in national discourses differed significantly from hashtags prominent in the pan-European discourse, as represented by the EDRI data set.
A comparison of the top 100 (frequency, Figure 1) hashtags in the pan-European and National data sets suggests that “policy issues” are less prominent in the pan-European discourse than in national discourses, as represented by 28 and 15 “policy hashtags” respectively. This suggests that policy issue debates are rather based in national than in EU level discourses. However, the data sets have constraints in terms of generalizability. Thus, to make this preliminary assumption a claim, further investigation of that specific aspect would be necessary,
National | EDRi |
| |
Figure1 : Hashtag categorisation per year, National and Regional Data Sets
Some hashtags are very distinctly positioned within the time period, such as #ccc, #shareconference and #wikileaks, which appear to be organizing frames for the 2011 period, while, policy issues such as ACTA and SOPA are more prominent in 2012. It is important to notice how the #privacy concern is prominent throughout the time period, and becomes dominant around the time of the Snowden revelations (2013). Simultaneously, Figure 2 demonstrates how in national discourses, specific and topical issues such as Snowden and DRIP enjoy sudden and volatile spikes, while broad thematic issues such as privacy remain constantly represented throughout the time frame.
Generally, however, thematic hashtags appear to dominate mores consistently than other types. This is particularly clear in Figure 1, where across both national and regional tweet sets, Policy Issues and Problem Areas together demand an increasing share of the twitter conversation over time, as opposed to tweets referencing events, organizations or technology. This would seem to indicate an increasing preoccupation with advocacy themes in the twitter discourse surrounding European digital rights.
Figure 2: Top ten Hashtags per year, National Database
Moving in the direction of the research question that asks which issues connect different countries, we approached from the most frequent hashtags perspective. First, we filtered the hashtags according to type - from our own categorization - and left only those two that relate to issues more directly: Policy Issues and Problem Areas. Then we selected the top 20 most-mentioned hashtags over six years in the natdigorgs dataset, as well as users that mentioned them more than 5 times, and plotted the data as a Social Network in Gephi, using the ForceAtlas2 (Jacomy et al., 2014) distribution mode to organize the results as a user-hashtag (bipartite) network, such as the nodes would be attracted by their structural similitude.
Finally, by colouring the nodes by country we get a perspective of how some issues like privacy appear to resonate regionally, even though it’s a hashtag in English language. In the graph we visualize actors from Germany, England, France, Nederlands, Belgium, USA and others.
Figure 3: privacy mentions by countries
On the other hand, issues could be national when we use hashtags as proxies, such as German-specific #VDS (stands for Vorratsdatenspeicherung, which means data preservation, or data retention in English), can be visualized through the same process, with radically different results, as seen below:
Figure 4: Vorratsdatenspeicherung (data preservation) in Germany
The discussion can be furthered when comparing with analysis per top users in each country. The analysis according to such view - which results in a different data set - shows almost no international or transnational perspectives (see Figure 6). One possible explanation for that difference in data approach and results is that the most mentioned users in our data set are more concerned with country-specific issues and problems, while if we look at the most popular issues and problems, being more inclusive with respect to users, those kind of inter or transnational problems surface.
This is an interesting assessment to investigate further on, since diffusion of information is an important research area on Twitter and communications. If a few users with a big influence - measured as number of mentions in the network - point to different issues than a lot of users with lesser mentions on the network, what would be the results? Are the most popular hashtags still the same for both datasets? As many authors argue, are less prominent users in Twitter important actors in order to bridge places, propagate different issues than those by Twittertariat (Bastos & Mercea, 2015)?
In order to understand the co-occurrence of hashtags organized by type, we plotted them as a social network visualization (see graph below) that is composed by (i) hashtags as nodes, coloured by its type (Policy Issues, Problem Area, Events etc.), (ii) co-occurrence (occurrence in the same tweet) as the edges, and (iii) weight of the edges defined by frequency of co-occurrences of both hashtags in the data set. The result (graph below) shows Problem Areas that are correlated with other Problem Areas and some clusters of Policy Issues that co-occur. It shows little mix of both and some very strong co-occurrences, like Surveillance and Privacy (strongest tie). This could indicate that the majority of the conversations on Twitter, driven by and with the main organizations that compose in first place our ‘natdigorgs’ data set in first place, revolve more around areas of problems than around specific issues. Why that happens could be an interesting path to move forward: Are discussions in Twitter less profound, due, for instance, to the shortness of messages? Are discussions around Problem Areas a strategy to keep the conversation alive in between two more focused campaigns when an issue arises?
Figure 5: hashtag co-occurrence
We put the list of influencers for each country (four countries) in the “from users” search box in TCAT and used the user/hashtag module to generate four user and hashtag networks. These four networks depict the relationship between influencers and the most often used hashtags by them for each country. We then combined these four networks, coded and grouped the users by country, as seen in the figure below.
Figure 6: hashtags by country
This figure suggests that many hashtags are country specific. However, there are a few hashtags such as net neutrality that connects countries. The hashtag net neutrality becomes more centred if we combine the German, Dutch, and French word of net neutrality. This suggests that language is a potential factor that separates hashtags by country.
There is only one Dutch organization member of EDRi. This organization, Bits of Freedom, is the core-actor who is structuring the topical space. In contrast to for instance Germany, the other most visible users within the national debate are not linked to this specific organization. While these most visible actors are invested in the cause of digital rights, it is not their sole cause. The issue space is occupied by national issues, and when transnational issues gain in importance, the national aspect of these transnational issues is highlighted.
In the UK, the main conversation is driven by a few actors (mainly NGOs organisations or people related to them). Transnational issues such as privacy, surveillance and net neutrality are very central in the debate, however UK users show also a very strong connection to national policy issues. For instance specific policy-issues act as a bridge between the most visible actors such as the #IPbill, #Snooper Charter. While more European transversal issues (e.g. migrant crisis…) lay outside the main conversation (connected to Statewatch.eu). Most strikingly, the British conversation is dominated by specialized issues, such as #blocking or #censorship. In contrast to the Netherlands, where general terms such as #privacy are used as the only mobilization message, in the UK actors mobilize their followers using general and more specific issues. This suggest that the British digital rights conversation might be dominated by experts but a general audience is following the debate.
For Germany the main actors in the issue space of digital rights activism are the twitter accounts of individuals or organisations. In comparison to the other nations it is striking how individuals are actively driving the discourse on. The two most used Hashtags are based on a national specific level. The Hashtags, most used by the german actors, are mostly based on political issues like “vds”, meaning “Vorratsdatenspeicherung” which was a big political issue in Germany in 2010 and 2015. Transnational issues do also have a certain importance, but mostly if they refer to political issues as well.
Certain organisations, such as the ccc (chaos computer club) are using several twitter-accounts to broadcast their message. In our current data-set, we did not combine these twitter-accounts due to time-constraints. In further research, combining twitter-accounts of one organization will allow us to get a clearer picture of the impact specific organizations have on the issue space, which is especially relevant in the German case.
The conversation of digital rights in France is dominated by La Quadrature, which is the only French EDRi organization. There are also individuals and other organizations who are influential in the debate, but they are all linked to La quadrature. Some examples are @ungarage, a Twitter account managed by volunteers for La Quadrature, and @jerezim, co-founder of La Quadrature.
Apart from some transnational policy and problem area issues, such as ACTA (ACTA-Anti-Counterfeiting Trade Agreement) -the most used hashtag-, surveillance, privacy, and net neutrality, the debate is focused on a national level discourse. For instance, specific policy-issues such as HADOPI law (#hadopi), which promotes the distribution and protection of creative works on the internet, and the French Intelligence Bill (#PJLRenseignement), that aims to regulate intelligence practices to prevent terrorism, are part of the main discussion.
The large nature of our group and data set resulted in a number of very preliminary insights which merit further study. The broad variety of research questions and frames of analysis suggest a number of compelling insights, which are grouped and discussed in the sections on findings.
Viewed together, we see that discussion about policy issues on Twitter mainly takes place on the national level. Conversation around general problems such as privacy and surveillance cuts across nations in Europe. In addition, a small number of organizations drive the conversation, but that might be due to the nature of the issue of digital rights. Our findings also suggest that while digital rights discourses in individual countries tend to use hashtags and address issues that are distinct to individual countries, some hashtags are used to structure conversation across countries, and that these tend to be hashtags representing abstract issues of broad appeal, in English. Similarly, our analysis of hashtag use over time suggests that hashtags representing abstract issues tend to trend more consistently over time, as compared to more specific issues which peak and fade in direct response to contextual events.
Lastly, we identified a number of distinct characteristics of the digital rights debate on twitter in each of the four countries studied directly. These characterizations are likely of interest to scholars and digital rights activists working in these countries.
For further research concerning national discourses, a few interesting suggestions can be made. For translingual interpretation of the results, we propose to establish a thesaurus of the transnational issues such as net neutrality. This would provide an opportunity to investigate how this topic is treated transnationality. This would also help clarify what issues cut across different countries. While this current study also includes a comparison of the national and transnational level, including a translingual aspect in the data, would allow for a more profound analysis.
Another way to focus our research more would be to reduce noise even more. In our current data-set, we did not combine the multiple twitter-accounts that some organizations use to broadcast their message, due to time-constraints. In further research, combining twitter-accounts of one organization will allow us to get a clearer picture of the impact specific organizations have on the issue space.
While the temporal aspect of our data was analysed, this could be done more comprehensively. This online analysis could be complemented with an offline analysis. This would provide a framework from which we can investigate to what extent policy debates on twitter, stimulate policy debates in the offline world.
Further research could also include other media outlets the organizations use, for instance Facebook. As an exploratory research on the Facebook pages of the organization already show, the networks of the organizations of Facebook differ from the networks on Twitter. Comparing or merging these two network could give more insight on the online presence and the online debate around these organizations.
In looking at the characteristics of participation within internet-based collective action on digital rights campaigning between 2007-2009, Yana Breindle (2012) concluded that, because of the level of expertise that is required, and contrary to claims of inclusiveness and openness,
“digital rights campaigning is in fact dominated by a small group of highly specialised movement entrepreneurs who mobilise occasionally to demonstrate broader support to policy-makers. The emergence of internet-based campaigning does not necessarily equal to more inclusive forms of participation. However, it allows for the engagement of resource-poor actors in traditional policy settings such as the EU.”
Considering this conclusion, our project proposes to move attention from internet-based collective action, as supported by websites, mailing lists, wikis and instant messaging channel, to the study of “connective action” dynamics (Bennett, W.L. & Segerberg, A., 2012). We do so by using Twitter as an interesting entry point to study new forms of campaigning and activism around highly complex sociotechnical issues (Weller, K. et al. éd., 2013; Earl, J. & Kimport, K., 2016). Specifically, we propose an exploratory cross-national and european comparative study of how the key activist organisations that are engaged in internet regulation affecting digital rights in Europe, animate and frame the Twitter attention dynamics on this issue between 2010 and 2016.
Bastos, M. T., & Mercea, D. (2015). Serial activists: Political Twitter beyond influentials and the twittertariat. New Media & Society, 1461444815584764.
Bennett, W.L. & Segerberg, A., 2012. The logic of connective action: Digital media and the personalization of contentious politics. Information, Communication & Society, 15(5), p.739–768.
Borra, E. & Rieder, B., 2014. Programmed method: developing a toolset for capturing and analyzing tweets. Aslib Journal of Information Management, 66(3), p.262‑278.
Breindl, Y., 2012. The Dynamics of Participation and Organisation in European Digital Rights Campaigning. JeDEM - eJournal of eDemocracy and Open Government, 4(1), p.24‑44.
Daer, A. R., Hoffman, R., & Goodman, S. (2014, September). Rhetorical functions of hashtag forms across social media applications. In Proceedings of the 32nd ACM International Conference on The Design of Communication CD-ROM (p. 16). ACM.
Earl, J. & Kimport, K., 2016. Current Research on Information Technologies and Society: Papers from the 2013 Meetings of the American Sociological Association, Routledge.
Jacomy, M. et al., 2014. ForceAtlas2, a Continuous Graph Layout Algorithm for Handy Network Visualization Designed for the Gephi Software. PLoS ONE, 9(6), p.e98679.
Weller, K. et al. (ed)., 2013. Twitter and Society. New York: Peter Lang Publishing Inc.