top of page

Politus Project
Rümeysanur Erikli Doğan

Digital techniques of data collection and analysis enable us to question the efficiency of commonly used methods of social science, in terms of time, money, and scientific value. These techniques also shift our focus to study the ongoing experimental environment people present in social media rather than constructed environment of surveys and interviews (Salganik, 2018, p.3). Politus project is significant in utilizing the Twitter data to measure public opinion which is a field of study dominated by survey companies in Türkiye.

Politus is funded by the European Research Council (ERC), the proof-of-concept fund, in 2022. The project is composed of a large team of social scientists, computer scientists, mathematician, and economists and they are leaded by Assoc. Prof. Dr. Erdem Yörük, who is also directing the Computational Social Science research group at Koç University. By employing big data and using methods of NLP and AI, the project creates an alternative method for measuring public opinion in Turkey which is less costly and time-consuming than surveys and producing real-time results. To achieve this aim, initially the twitter data of 53 million users are used and the age, location, and gender information of 1.5 million users are identified. Based on the data of these users a new way to assess public opinion is created by using AI and NLP methods and thus, Twitter data is transformed to represent the patterns of voter and consumer behaviors in Türkiye.

 

They created a data platform, Politus Analytics. Although till now they have published two reports, the Panorama of Turkey, they prefer to present their results on their Twitter account, Politius_AI. As they state in their second panorama of Turkey report, based on their twitter data and AI model, they measured the level of support for the candidates of the 2023 presential election day by day (Figure 1). Their emotion and stance analysis have enabled them to present the weekly change of different emotions between Erdogan and Kılıçdaroglu supporters (Figure 2). They also show the weekly change of topics mostly discussed among electors on Twitter (Figure 3).

Figure 1. The change of weekly support for Erdogan and Kilicdaroglu.

   

 

 

 

Figure 2. Weekly change of fear among Erdogan and Kilicdaroglu supporters.

 

Figure 3. The change of hot topics among electors

They presented their analysis of voter behaviors during the 2023 presidential elections which is very courageous act in an environment of survey dominancy in measuring public opinion. Although their posts created some debates on the validity of their results on Twitter, bearing in mind that almost all survey companies failed in their predictions on the election results, the fact that their results created live debates shows us that their alternative method is welcome. But their accomplishment is facing a challenge created by Elon Musk’s roughe capitalist attacks with higher prices for more limited Twitter API services and such draw backs on the data sources is a great limitation to research for computational social science and digital humanities scholars.

 

 

References

Salganik, M. (2018). Bit by Bit: Social Research in the Digital Age. Princeton University Press.

image.png
image.png
image.png
bottom of page