Introduction
This study aims to explain the process of filter bubble formation in users’ encounters with political news on the Instagram platform, focusing on the dynamic interaction between the user, the algorithm, and the social network. Despite the growing role of Instagram as a news source and concerns about algorithmic personalization leading to information isolation, there is a lack of process-oriented, qualitative research that examines how filter bubbles emerge in the specific socio-political context of Iran. This research addresses that gap by developing a grounded model of filter bubble formation.
Theoretical Framework: The study integrates three theoretical perspectives: Giddens’ structuration theory (duality of structure), Postman’s media ecology theory (media as environments that shape cognition and social organization), and Katz’s uses and gratifications theory (active, goal-oriented media selection). This synthesis allows the research to move beyond either/or explanations (algorithm vs. user) and instead capture the dialectical relationship between human agency and technological structure in platform-mediated news consumption.
Materials & Methods
A qualitative approach with a grounded theory strategy (systematic version of Corbin & Strauss) was employed. Data were collected through three complementary tools: (a) semi-structured interviews with 20 active Instagram users and 6 media/platform experts (selected via theoretical sampling until saturation); (b) non-participant observation of participants’ public profiles over one month; and (c) document analysis of 200 comments on 10 high-engagement political news posts from pages with varying political orientations. Data were analyzed using open, axial, and selective coding in MAXQDA software. Trustworthiness was ensured through member checking, triangulation, and peer review.
Discussion & Result
Analysis yielded 114 initial codes, later condensed into 28 basic concepts and then into three main categories: (1) strategic user actions (selective search strategies, membership and alignment of followed networks, confirmation-oriented interactions such as likes, comments, and shares); (2) algorithmic reinforcement mechanisms (invisible personalization under content restrictions, controversy-and-emotion-based prioritization, and double filtering—legal plus automated censorship); and (3) the socialized bubble context (like-minded sharing among followers, closed homogeneous groups acting as small-scale echo chambers, and active content forwarding that feeds back into the algorithm).
From these categories, the study develops a “Tripartite Interaction Paradigmatic Model” which demonstrates that the filter bubble forms through a self-reinforcing cycle: initial psychological motivations (reducing cognitive dissonance, identity reinforcement) → strategic user actions (selective following, engagement) → algorithmic data feeding → algorithmic reinforcement (personalization, emotional prioritization, double filter) → social network facilitation (sharing, closed groups, forwarding) → consequences (information isolation, confirmation bias, political polarization) → feedback to user, intensifying homogeneous interactions.
The study introduces three novel concepts grounded in the Iranian context: “strategic self-censorship” (users actively avoid opposing views not out of fear but to preserve cognitive comfort), “double bubble” (the overlap of algorithmic personalization with external political-legal restrictions on content visibility), and “complex trust effect” (users accept news because it is shared by a like-minded peer, not necessarily because of its veracity).
Conclusion
The filter bubble on Instagram is not merely a technical glitch or an algorithmic imposition; it is an inherent, emergent property of the adaptive system comprising user agency, platform algorithms, and social networks. The key paradox is that users simultaneously complain about information homogenization while reproducing it through their own selective behaviors. Breaking this cycle requires more than algorithmic transparency; it demands “critical algorithmic literacy” that enables users to recognize the hidden feedback loops they participate in. Policy interventions must target three levels simultaneously: enhancing users’ media/algorithmic literacy, ensuring algorithmic transparency and mandatory content diversity, and addressing the social network context (e.g., reducing group polarization). The study’s limitations include reliance on self-reported data, lack of access to Instagram’s actual algorithmic code, and context-specificity to Iran, which limits generalizability. Future research should adopt mixed-methods longitudinal designs and comparative studies across platforms (Twitter, Telegram).