Artificial intelligence and social media as new arenas of political competition: challenges for democracy
This study examined how users’ perceptions of algorithmic influence and AI manipulation are associated with key democratic indicators, including trust in political information, perceived political autonomy, and the quality of online deliberation. The study is based on a mixed research design. The quantitative part included a survey of 1,795 active social media users in the Republic of Kazakhstan. The analysis used correlation analysis, multivariate OLS regression with interaction effects, and sample weighting. The computational component included principal component analysis (PCA) to validate the indices of perception of algorithmic influence and AI manipulation (explained variance: 52 and 58%, respectively). The qualitative part was based on interviews analyzed based on thematic analysis. The results indicated that the perception of algorithmic personalization has an ambivalent democratic effect: it is moderately negatively associated with trust in political information and the quality of online discussions ( β ≈ −0.10…−0.15), but under certain conditions is associated with a higher sense of control over information choices. In contrast, the index of perception of AI manipulation had a strong and stable negative relationship with all democratic indicators ( β ≈ −0.30…−0.45). Heterogeneity of effects was revealed: the most pronounced negative consequences were recorded among Telegram and YouTube users, as well as among younger respondents (18–29 years old). Thus, the study indicated that the democratic challenges associated with AI and social media are technological and socio-interpretive in nature. The vital factor in democratic erosion is not the fact of algorithmic mediation itself, but the perception of opacity, control, and manipulability of AI. The findings contribute to the interdisciplinary debate on democracy in the digital age and identify the need for policy and regulatory responses to the challenges of AI.
Paper
Full text
Artificial intelligence and social media as new arenas of political competition: challenges for democracy
Semantic Scholar · 2026
Abstract
This study examined how users’ perceptions of algorithmic influence and AI manipulation are associated with key democratic indicators, including trust in political information, perceived political autonomy, and the quality of online deliberation. The study is based on a mixed research design. The quantitative part included a survey of 1,795 active social media users in the Republic of Kazakhstan. The analysis used correlation analysis, multivariate OLS regression with interaction effects, and sample weighting. The computational component included principal component analysis (PCA) to validate the indices of perception of algorithmic influence and AI manipulation (explained variance: 52 and 58%, respectively). The qualitative part was based on interviews analyzed based on thematic analysis. The results indicated that the perception of algorithmic personalization has an ambivalent democratic effect: it is moderately negatively associated with trust in political information and the quality of online discussions ( β ≈ −0.10…−0.15), but under certain conditions is associated with a higher sense of control over information choices. In contrast, the index of perception of AI manipulation had a strong and stable negative relationship with all democratic indicators ( β ≈ −0.30…−0.45). Heterogeneity of effects was revealed: the most pronounced negative consequences were recorded among Telegram and YouTube users, as well as among younger respondents (18–29 years old). Thus, the study indicated that the democratic challenges associated with AI and social media are technological and socio-interpretive in nature. The vital factor in democratic erosion is not the fact of algorithmic mediation itself, but the perception of opacity, control, and manipulability of AI. The findings contribute to the interdisciplinary debate on democracy in the digital age and identify the need for policy and regulatory responses to the challenges of AI.
References (33)
Scroll for more · 21 remaining