Akin Unver, Ozyegin University and Ahmet Kurnaz, Çanakkale Onsekiz Mart University
The literature on online disinformation studies focuses disproportionately on the United States – especially on the 2016 Presidential elections – and has failed to generate an equally robust and diverse research agenda elsewhere.[1] Empirical studies have drawn on a very narrow pool of cases, with the overwhelming majority of the scientific and policy focus on what Russia is doing in the United States, or a handful of Western nations.[2] This impairs construction of a truly comparative and generalizable scientific inquiry, especially in terms of what disinformation (deliberate use of false information to deceive) or influence operations (deploying a mix of accurate, semi-accurate and false information to achieve strategic goals) mean for the broader world and international competition dynamics. To that end, the study of both fields is in need of longitudinal and comparative works: to provide perspective on how disinformation dynamics observed at one time are different than those at others; how dynamics observed in one country differ from those in other countries; and how operations conducted by different external actors vary. What’s more, availability bias afflicts the wider disinformation studies field, as very few studies deal with the question of what the existence of disinformation means in relation to the cases where information manipulation doesn’t exist. In this essay, we examine Russian information operations in Turkey as a first step towards addressing these shortcomings in the literature.
Turkey as a Case Study
Why Turkey? ‘Buffer countries’ or ‘insulators’ as defined in Regional Security Complex Theory (RSCT) are well-suited for such comparative work.[3] Although there are clear theoretical and methodological differences between how these two terms are studied, they both indicate countries that lie at the intersection of two or more large security communities. Such countries are usually not powerful enough to dominate either community, but also not weak enough to be dominated by either. To that end such countries are regularly influenced by multiple security communities, and their domestic power dynamics acutely reflect external security-related influences; in turn, these internal dynamics have significant impact on policy towards external security communities.
Turkey is one of those buffer or insulator countries.[4] Its imperial and Republican foreign policy were both heavily influenced by hedging and balancing dynamics against the Russian Empire, and then the USSR. Even as a NATO ally, Turkey competed with other NATO countries (most specifically Greece) and cooperated with the USSR (especially in building the Turkey’s heavy industries in the 1970s) as circumstances dictated.[5] Although the end of the Cold War and the next two decades enabled Turkish policymakers to build a new security identity against a weaker Russia, the rise of an emboldened and revisionist Russian foreign policy after 2010 brought back structural balancing considerations for Ankara. Especially after the annexation of Crimea, the Russian military encroaching into the Black Sea, Syria, eastern Mediterranean and the Caucasus, Turkish decision makers increasingly found themselves dealing with situations that amounted to being strategically surrounded by Russia.[6] Ankara felt that NATO continually failed to provide sufficient security commitments against Russian encroachments, resulting in Turkish hedging and then bandwagoning with Russia.
Turkish foreign policy after 2014 can thus be described as multilateral hedging at a time of significant changes in the balance of power in its immediate environment (Russia-related), and also at the global level (China-related).[7] In addition, a growing strategic divide between the US and Europe (as well as within the EU itself) placed Turkey at the intersection of multiple strategic influences originating from Washington, London, Brussels, Berlin, Moscow and Beijing. Turkey was also embroiled in regional competition with countries such as Egypt, Saudi Arabia and the United Arab Emirates which actively engaged in digital disinformation and influence operations. As a result, Turkey became a battleground for foreign influence operations, not just limited to Russia. Therefore, Turkey after 2014 is one of the most interesting case studies for the study of multilateral digital influence operations in general, and disinformation in particular.
Russian Digital Influence Events in Turkey: The Data
In a recently concluded project, our lab has focused on building a ‘Russian influence event dataset’ (RUSDAT) that collects social media data on such activities since 2014.[8] This paper updates the original 2019 publication, with new data which we continued to collect as Russian disinformation activities continued.
RUSDAT was built on several criteria. First, we focused on bilateral geopolitical events between Russia and Turkey, constructed a keyword corpus that contained terms and word combinations related to each high-profile strategic event and extracted all Twitter data that corresponded to those events. We then sorted them according to the amount of clean data we had after weeding out irrelevant posts (including tweets from brands, football clubs or Korean pop bands, which surprisingly often post local hashtags to rise into the trending topic list!). Finally, we ranked these events based on how much clean data we had on them and discarded cases that contained too few tweets (below 2 million) or had too much dirty data as percentage of the whole dataset. Ultimately, we focused on four of the most important events that also contained the highest volume and percentage of clean data to explore deeper, although initially discarded cases were retained within RUSDAT.
The ‘clean’ cases picked for study were Turkey’s downing of a Russian jet in November 2015, Turkey’s failed coup attempt in July 2016, the assassination of the Russian ambassador in Ankara in December 2016, and the S-400 negotiations between Turkey and Russia, which itself was separated into six key benchmarks (declaration of interest, declaration of no-cancellation, signing of the purchase memorandum, signing of the commercial agreement and final acquisition). Social media dynamics throughout the S-400 negotiations were particularly useful as each benchmark gave a clear idea on how influence operations and media response against them changed over time. Out of these cases, we were able to identify a distinct ‘pro-Russian’ influence cluster that encompassed a large network of Turkish-language real and sock puppet accounts, occasionally supporting the narratives of pro-Russian Turkish-language outlets Sputnik News Turkish, Aydınlık newspaper and Russia Today’s Turkish-language news section.
The 2015 Downing of a Russian Jet: Distraction
Over the last few years, foreign observers of Turkey frequently asked, ‘who lost Turkey?’, meaning whose fault it was that Turkey had become so detached from the West.[9] Answers ranged from general NATO apathy towards Turkey’s changing security environment after Crimea annexation, to European analysts blaming Trump, or American analysts blaming European resistance to Turkey’s EU membership. This question can be better asked temporally: when was Turkey ‘lost’? From a digital communication point of view, our study can pinpoint a single event: the Turkish downing of a Russian jet in November 2015.
Soon after the Russian jet was shot down, we began observing the emergence of two discursive clusters (or narratives). The Turkish version argued that the decision was justified because the unidentified jet had strayed too much into Turkish territory. A second cluster of tweets asserted that the jet was shot outside the Turkish airspace, and was thus, unjustified. As internal military investigations of both sides began yielding results that supported the first claim after a week, Russian outlets adopted an organized distraction tactic which originated in accounts associated with the Ministry of Defense (based on how the initial MoD tweets were spread across both English- and Turkish-language Twitter ecosystem), which focused on Turkey’s alleged oil smuggling deal with ISIS.[10]
This distraction tactic, originally crafted in and disseminated by the Kremlin (based on its first appearance and subsequent diffusion patterns on Twitter), soon got picked up by international news and media agencies, including those of other NATO countries. Successfully distracting the discussion away from the SU-24 incident, this became one of Kremlin’s most efficient influence operations across the entirety of NATO countries, managing not only to divide and nullify NATO’s countermeasures against Russian violations of NATO airspace, but also created a very significant wedge between Turkey and its Western allies, isolating Turkey in the short- to medium-term. Although both the Pentagon and the State Department had rejected Russian allegations of Turkey’s ISIS-related oil smuggling,[11] the story was disseminated far and wide in Western capitals, ending conclusively only after Presidents Putin and Erdoğan met in August 2016. This meeting, where Turkey conceded defeat in its information war with Russia, was a major turning point in Turkey’s relations with Russia.[12] The ‘ISIS oil’ story then disappeared entirely and immediately on Russian and Turkish-language Twitter.
The 2016 Failed Coup: Amplification
During Turkey’s failed July 2016 coup attempt, Russian influence operations benefited significantly from the pre-existing and growing Turkish domestic skepticism towards the US and NATO. Some of this public skepticism was a result of growing strategic disagreements in Syria. The Turkish government’s vocal complaint that neither the US nor Europe (with the exception of the UK) condemned the coup attempt during its early hours had a major rallying effect around the narrative that the coup was instigated by NATO.[13]
Throughout the coup attempt, all of the widest-spread disinformation instances had a domestic origin. But pro-Russian accounts did try to amplify the prevalent public sentiment that the coup was planned and orchestrated by pro-NATO cells within the military. That said, compared to other major instances, the activity of pro-Russian accounts throughout the failed military coup make up only a very small fraction (less than 1%) of the total engagement clusters observed. Even months after the failed coup attempt, pro-Russian accounts continued to sustain the narrative that it was NATO-affiliated groups that were behind the coup itself.
Assassination of Russian Ambassador: Silence
Five months later, Russian Ambassador Andrey Karlov was assassinated in Ankara, straining already fragile Turkish-Russian relations even further. Yet the pro-Russian influence ecosystem went completely dark, suggesting a centrally-planned full silence. Why? First, after the August 2016 Putin-Erdoğan meeting, the two countries had charted a common course to deconflict bilateral relations and Russia had no further interest in destabilizing Turkey. Second, the Russian government was already in close communication with Turkey to contain the damage of this incident as quickly as possible.[14] The entirety of the social media war that followed the Karlov assassination was domestic to Turkey, with arguments taking place between two pro-government clusters: one that viewed the assassination as ‘justified’ in the face of growing Russian attacks against pro-Turkish rebel groups in Syria, and the other, which advocated for calm and reconciliation in line with the mainstream government view. The assassination debate disappeared to a great extent on social media after only four days, suggesting a direct gag order by both Ankara and Moscow.
The S-400 Negotiations: Sustained Influence Operations
Finally, we explored the S-400 negotiations, as divided into six benchmarks: 10 October 2016 when Turkey and Russia declared that serious Presidential-level negotiations were underway over S-400 sales, Erdoğan’s 10 March 2017 visit to Moscow to assert Turkey’s commitment to S-400, 29 December 2017 commercial agreement between the two sides, 3 April 2018 President Erdoğan’s statement on Turkey’s ‘point of no return’ on S-400 purchase, 19 August 2018 President Putin’s statement that deliveries could be made a year earlier than planned, and July 2019 when the first shipment of S-400 ground systems arrive in Turkey. The S-400 case demonstrates the explanatory power of measuring influence operations across a longer timeframe (in this case, across almost 3 years), as the Turkish media ecosystem changed to an important degree throughout this episode, as did many other national and international variables of interest.
Figure 1 – Longitudinal sentiment scores of positive and negative sentiment clusters; October 2016 – July 2019
Table 1 – News outlets that form up the core of positive and negative sentiment clusters
The most important finding is the gradual transition of the Turkish-language sentiment scores (measured by deploying ‘BERT Sentiment Analysis Turkish’[15]) associated with S-400s, from mixed (equal measures of positive and negative), to mostly positive across the six benchmarks we observe. This means that the overall outlook of the Turkish social media ecosystem towards the S-400s started off as skeptical and divided, and gradually became very positive towards these systems. The main words associated with the skeptical topic clusters reflect worries about interoperability of the Russian systems with NATO infrastructure, Turkey’s existing NATO commitments and what the S-400 acquisitions would mean for Turkey’s other major partnership in the F35 fighter jet program. In contrast, word clusters associated with the ‘pro-S-400’ sentiments reflect the importance of strategic autonomy, NATO’s broader relevance for Turkey, and technical details that reflect the view that S-400s are ‘better’ anti-air systems than the Patriots. Over time, ‘pro-S-400’ topic clusters dominate the Turkish-language discussion with heavy involvement of pro-Russian and also pro-government accounts in Turkey.
Dynamics After 2019: COVID, Nagorno-Karabakh and Biden
Three main additional events triggered pro-Russian influence operations after 2019. The first was the emergence of COVID-19 and the onset of the global race for vaccines. The second major event was the 2020 Nagorno-Karabakh war between Azerbaijan and Armenia, in which Turkey and Russia served as major external stakeholders. The third major event was the election of Joe Biden and the proliferation of skeptical news reports from the pro-Russian ecosystem on his capacity to lead, or whether his election would really make a difference.
Predictably, the pro-Russian network in Turkish-language social media platforms (Twitter, Facebook and Instagram) sowed widespread confusion about the efficacy and side-effects of American and European vaccines, while remaining silent about the Chinese SINOVAC, and continually advertising the Russian Sputnik-V vaccine. This network published vaccine-related information which focused on vaccine skepticism, emphasizing the importance of getting vaccinated, although with a twist that always ends with a positive note about the affordability, availability and the efficacy of Sputnik-V. Further word clusters along this line focus on the positive international reception of the Sputnik vaccine and the cases of patients that were saved thanks to getting vaccinated. This information pushed back on BBC Turkish, Deutsche Welle Turkish and Fox Turkey articles that disseminated skeptical views of Sputnik V or the Sinovac/Coronavac vaccines.
Figure 2 – Frequency hierarchy of the most popular features (terms) on Coronavirus
On Nagorno-Karabakh, Russian influence operations were extra careful, as the long-frozen conflict had been one of the core national interests of Turkey and had multi-partisan support among the Turkish voters. Since there was no domestic audience to which Russia could play on this matter, most pro-Russian accounts focused instead on the need to stabilize the Karabakh region and posted themes related to the ‘possibility’ of an Armenian-Azerbaijan reconciliation with the joint oversight of Turkey and Russia. During the conflict itself, however, these accounts pursued a distinct pro-Armenian line, occasionally sharing low-diffusion disinformation content about the course of the conflict, including false accounts of attacks, casualties and clashes.
Finally, the entirety of the pro-Russian information ecosystem turned uncharacteristically over-active after the election of Biden, regularly disseminating fake news about his physical and mental fitness and questioning his ability to lead. Additionally, this ecosystem had been using key events, such as Biden’s recognition of the Armenian genocide, to draw a wedge between him and President Erdoğan. In one such instance, a large bot campaign pushed the argument that the Turkish government must file an international lawsuit against Biden on the grounds of ‘hate speech’, in exchange for Biden’s recognition of the genocide. Another major campaign focused on pinpointing Biden as the main culprit behind the CAATSA sanctions issued against Turkey by the US Congress under the Trump administration.
Figure 3 – Feature frequency network of Coronavirus-related terms
Conclusion and Implications
To sum up, this project has so far yielded nuanced results that show a more cautious, more context-specific and more ‘under-the-radar’ digital influence strategy on the part of Russia. We hypothesize that Russia’s relative caution in influencing Turkish digital media ecosystem owes to the fact that Turkey, as an insulator country, is indeed divided between multiple foreign influence strategies and possible Russian interpretation that further destabilization of this ecosystem would trigger a backlash. This is interesting because the majority of the US-centric Russian disinformation studies report explicit, often aggressive and blatantly ‘in your face’ tactics by accounts that can rather easily be traced back to a particular Russia-origin network. In our 6 year ongoing study, we observe a subtler, ‘smoke and mirrors’ tactic by accounts that are (with one specific exception, which is the ‘ISIS oil’ campaign) several degrees separated from the usual suspect clusters that have been plaguing Western information ecosystems for quite some time. This demonstrates the value of the comparative and longitudinal studies for which we call.
[1] Edda Humprecht, Frank Esser, and Peter Van Aelst, “Resilience to Online Disinformation: A Framework for Cross-National Comparative Research,” The International Journal of Press/Politics 25, no. 3 (July 1, 2020): 493–516, https://doi.org/10.1177/1940161219900126; W Lance Bennett and Steven Livingston, “The Disinformation Order: Disruptive Communication and the Decline of Democratic Institutions,” European Journal of Communication 33, no. 2 (April 1, 2018): 122–39, https://doi.org/10.1177/0267323118760317.
[2] Eleni Kapantai et al., “A Systematic Literature Review on Disinformation: Toward a Unified Taxonomical Framework,” New Media & Society 23, no. 5 (May 1, 2021): 1301–26, https://doi.org/10.1177/1461444820959296.
[3] Barry Buzan and Ole Wæver, “Macrosecuritisation and Security Constellations: Reconsidering Scale in Securitisation Theory,” Review of International Studies 35, no. 2 (2009): 253–76; Barry Buzan, “Regional Security Complex Theory in the Post-Cold War World,” in Theories of New Regionalism: A Palgrave Reader, ed. Fredrik Söderbaum and Timothy M. Shaw, International Political Economy Series (London: Palgrave Macmillan UK, 2003), 140–59, https://doi.org/10.1057/9781403938794_8.
[4] André Barrinha, “The Ambitious Insulator: Revisiting Turkey’s Position in Regional Security Complex Theory,” Mediterranean Politics 19, no. 2 (May 4, 2014): 165–82, https://doi.org/10.1080/13629395.2013.799353.
[5] Ziya Öniş and Şuhnaz Yılmaz, “Turkey and Russia in a Shifting Global Order: Cooperation, Conflict and Asymmetric Interdependence in a Turbulent Region,” Third World Quarterly 37, no. 1 (January 2, 2016): 71–95, https://doi.org/10.1080/01436597.2015.1086638.
[6] Hamid Akin Unver, “The Fog of Leadership: How Turkish and Russian Presidents Manage Information Constraints and Uncertainty in Crisis Decision-Making,” Southeast European and Black Sea Studies 18, no. 3 (July 3, 2018): 325–44, https://doi.org/10.1080/14683857.2018.1510207.
[7] H. Tarık Oğuzlu, “Turkish Foreign Policy at the Nexus of Changing International and Regional Dynamics,” Turkish Studies 17, no. 1 (January 2, 2016): 58–67, https://doi.org/10.1080/14683849.2015.1136088.
[8] Hamid Akin Unver, “Russian Disinformation Ecosystem in Turkey” (Istanbul: Center for Economic and Foreign Policy Research (EDAM), March 2019).
[9] Keith Johnson Gramer Robbie, “Who Lost Turkey?,” Foreign Policy (blog), July 19, 2019, https://foreignpolicy.com/2019/07/19/who-lost-turkey-middle-east-s-400-missile-deal-russia-syria-iraq-kurdish-united-states-nato-alliance-partners-allies-adversaries/.
[10] Maria Tsvetkova Kelly Lidia, “Russia Says It Has Proof Turkey Involved in Islamic State Oil Trade,” Reuters, December 2, 2015, https://www.reuters.com/article/us-mideast-crisis-russia-turkey-idUSKBN0TL19S20151202.
[11] State Department and Pentagon arguments need dissecting in detail. Their argument is that ‘Turkey buys oil from ISIS’ and ‘Turkish oil companies buy oil from ISIS-controlled regions’ are two different narratives. Turkish companies predictably buy oil from the same Syrian oilfields they have been trading with since 1970s. When those areas were overtaken by ISIS, oil trade continued, and Turkish tankers continued to carry crude from the same oilfields (since oilfields don’t move) they have been operating from for decades. Although this looks like a tiny detail, it is specifically this kind of nuances that feed more sophisticated and successful Russian influence operations. “Pentagon Rejects ‘Preposterous’ Idea That Turkey Is Aiding ISIS Oil Trade,” NBC News, December 2, 2015, https://www.nbcnews.com/video/pentagon-rejects-preposterous-idea-that-turkey-is-aiding-isis-oil-trade-577939523771; Lucas Tomlinson, “State Dept. ‘Rejects’ Russia’s Claims That Turkey Smuggling ISIS Oil,” Fox News, December 4, 2015, https://www.foxnews.com/politics/state-dept-rejects-russias-claims-that-turkey-smuggling-isis-oil.
[12] Shaun Walker, “Erdoğan and Putin Discuss Closer Ties in First Meeting since Jet Downing,” The Guardian, August 9, 2016, sec. World news, http://www.theguardian.com/world/2016/aug/09/erdogan-meets-putin-leaders-seek-mend-ties-jet-downing-russia-turkey.
[13] Tim Arango and Ceylan Yeginsu, “Turks Can Agree on One Thing: U.S. Was Behind Failed Coup,” The New York Times, August 2, 2016, sec. World, https://www.nytimes.com/2016/08/03/world/europe/turkey-coup-erdogan-fethullah-gulen-united-states.html.
[14] Andrew Finkel, “Turkey and Russia Have United over the Karlov Killing. But Deep Tensions Remain | Andrew Finkel,” The Guardian, December 20, 2016, sec. Opinion, http://www.theguardian.com/commentisfree/2016/dec/20/andrei-karlov-assassination-turkey-russia-putin-erdogan.
[15] Atıf Emre Yüksel, Yaşar Alim Türkmen, Arzucan Özgür, and Berna Altınel. “Turkish tweet classification with transformer encoder.” in Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2019), pp. 1380-1387. 2019.