How do AI recommendation algorithms create toxic feeds, and can users switch them off?

Echo chambers refer to the phenomenon in which individuals are consistently exposed to content that aligns with their existing viewpoints. Over time, this can narrow a user’s perspective and make it harder to encounter different opinions. In this systematic literature review, we looked at studies published between 2019 and early 2025 and how they have approached this issue, from understanding the cause, and examining existing detection and mitigation strategies. We went through and organized the main findings, noting patterns in the algorithms used, the role of user behavior, and the influence of the data itself. Several works also suggest ways to introduce more variety into recommendations, aiming to break repetitive exposure. Our review confirms that echo chambers and filter bubbles do exist in recommender systems and that they raise concerns for diversity and fairness. Furthermore, we end by pointing to open questions and possible directions for future work, for both researchers and practitioners.

A total of 30 studies were analyzed, using the PRISMA 2020 framework, encompassing computational audits, simulation modeling, surveys, ethnographic accounts, and mixed-methods designs across diverse platforms, including Facebook, YouTube, Twitter/X, Instagram, TikTok, and Weibo. Results reveal three consistent patterns: (i) algorithmic systems structurally amplify ideological homogeneity, reinforcing selective exposure and limiting viewpoint diversity; (ii) youth demonstrate partial awareness and adaptive strategies to navigate algorithmic feeds, though their agency is constrained by opaque recommender systems and uneven digital literacy; and (iii) echo chambers not only foster ideological polarization but also serve as spaces for identity reinforcement and cultural belonging. Despite these insights, the evidence base suffers from geographic bias toward Western contexts, limited longitudinal research, methodological fragmentation, and conceptual ambiguity in key definitions. This review highlights the need for integrative, cross-cultural, and youth-centered approaches that bridge empirical evidence with lived experiences.

This review synthesizes recent research from a social-psychological perspective, arguing that filter bubbles are not produced by algorithms alone but emerge through the recursive interaction of motivated cognitive processing, identity-based social network structures, and algorithmic amplification of behavioral and emotional cues. We identify emotion as an underrecognized yet central mechanism that operates across these layers: emotional states such as anger, threat, and defiant self-worth guide information seeking, reinforce group affiliation, and shape algorithmic recommendation patterns, thereby intensifying filtering dynamics and contributing to attitude extremization. By conceptualizing filter bubbles as systems of cognitive coherence, identity protection, and affective regulation, we propose a dynamic multi-level explanatory model and outline implications for interventions, including reflective reasoning strategies, weak-tie exposure, and approaches that address the emotional and identity foundations of information selection.

We conclude by outlining a stage-based framework for governance and future research.Public AbstractAI-driven algorithms designed to maximize engagement on social media, compounded by generative AI, can unintentionally set the stage for radicalization. It begins with Exposure, where algorithms push users toward extreme content because it captures attention. Next, during Reinforcement, algorithms feed users personalized content while AI swarms can create a synthetic consensus that reinforces emerging biases, normalizes extremity, and insulates users from alternative views. Third, during Group Integration, individuals are absorbed into extremist networks, reinforced by human peers, AI companions, and bot swarms that validate radical beliefs and deepen identity ties. By exploiting psychological needs for belonging and certainty, this stage becomes particularly pernicious, potentially opening the door for violence. We propose policy measures that can reduce radicalization at each stage.

Algorithms organised what content participants encountered, how frequently, and with what emotional intensity, within broader socioecological contexts. Three interlocking processes sustained exposure; algorithmic reinforcement of engagement, emotional feedback loops between mood and recommendations, and a persistent sense of limited control despite high awareness. Although participants developed adaptive strategies and algorithmic literacy, these were unevenly distributed and often insufficient within engagement-driven, profit-oriented systems.ConclusionsDigital mental health may be produced through interactions between individual capacities, social environments, and commercially driven platform design. Protecting population mental health therefore requires multi-level action, including developmentally appropriate social media and algorithmic literacy education alongside structural reforms to recommendation systems, transparency, and platform governance.

Echo chambers refer to the phenomenon in which individuals are consistently exposed to content that aligns with their existing viewpoints. Over time, this can narrow a user’s perspective and make it harder to encounter different opinions. In this systematic literature review, we looked at studies published between 2019 and early 2025 and how they have approached this issue, from understanding the cause, and examining existing detection and mitigation strategies. We went through and organized the main findings, noting patterns in the algorithms used, the role of user behavior, and the influence of the data itself. Several works also suggest ways to introduce more variety into recommendations, aiming to break repetitive exposure. Our review confirms that echo chambers and filter bubbles do exist in recommender systems and that they raise concerns for diversity and fairness. Furthermore, we end by pointing to open questions and possible directions for future work, for both researchers and practitioners.

This review synthesizes recent research from a social-psychological perspective, arguing that filter bubbles are not produced by algorithms alone but emerge through the recursive interaction of motivated cognitive processing, identity-based social network structures, and algorithmic amplification of behavioral and emotional cues. We identify emotion as an underrecognized yet central mechanism that operates across these layers: emotional states such as anger, threat, and defiant self-worth guide information seeking, reinforce group affiliation, and shape algorithmic recommendation patterns, thereby intensifying filtering dynamics and contributing to attitude extremization. By conceptualizing filter bubbles as systems of cognitive coherence, identity protection, and affective regulation, we propose a dynamic multi-level explanatory model and outline implications for interventions, including reflective reasoning strategies, weak-tie exposure, and approaches that address the emotional and identity foundations of information selection.

We conclude by outlining a stage-based framework for governance and future research.Public AbstractAI-driven algorithms designed to maximize engagement on social media, compounded by generative AI, can unintentionally set the stage for radicalization. It begins with Exposure, where algorithms push users toward extreme content because it captures attention. Next, during Reinforcement, algorithms feed users personalized content while AI swarms can create a synthetic consensus that reinforces emerging biases, normalizes extremity, and insulates users from alternative views. Third, during Group Integration, individuals are absorbed into extremist networks, reinforced by human peers, AI companions, and bot swarms that validate radical beliefs and deepen identity ties. By exploiting psychological needs for belonging and certainty, this stage becomes particularly pernicious, potentially opening the door for violence. We propose policy measures that can reduce radicalization at each stage.

The Australian government hopes to protect social media users from being served content they don't want to see.