Algorithmic Folk Theories

• Bias: Algorithmic folk theories are informal user understandings of how platform algorithms work, formed through personal experience, pattern observation, and knowledge sharing within communities rather than from official documentation.
• What it breaks: Self‑presentation on social media, content‑creation strategies, professional decisions in data analysis, identity perception, understanding of algorithmic fairness, and interaction with digital platforms.
• Evidence level: L2 — multiple qualitative and mixed‑method studies across various platforms (TikTok, cross‑platform analyses), including a seminal work on trans‑feminine content creators (S004) that confirms influence on user behavior and identity formation.
• How to spot in 30 seconds: When you or someone says “the algorithm loves videos that are exactly 15 seconds long” or “posting at 7 p.m. boosts reach” without any technical documentation — that’s an algorithmic folk theory in action.

How users develop their own theories about how algorithms work?

Algorithmic folk theories are collective user beliefs about the mechanisms of platform algorithms that are formed not from official documentation but from personal experience, pattern observations, and knowledge exchange within communities. Users notice that certain actions — using specific hashtags, posting at particular times, a certain video length — correlate with changes in content visibility, and based on these observations they develop their own explanatory models (S004). The phenomenon attracted substantial academic attention when researchers began documenting how social‑platform users conduct collective experiments and devise shared optimization strategies.

Research shows that algorithmic folk theories are most prevalent on social platforms with personalized content feeds, especially TikTok, Instagram, and YouTube (S004). However, recent work has broadened understanding of the phenomenon, demonstrating that folk theories also affect professional decisions in data analysis and function as an organizational infrastructure within networks that manage content‑creator labor (S001). Crucially, these theories are not individual misconceptions — they are socially constructed through community interaction, where users share observations and develop common approaches.

A critically important aspect of algorithmic folk theories is their link to identity formation. Users experiment with self‑presentation, monitor algorithmic responses through reach and recommendation metrics, adjust their behavior, and develop notions of how the algorithm categorizes them. This is especially significant for marginalized groups, such as LGBTQ+ users, who craft specialized folk theories about how algorithms handle content related to their identities (S003, S004).

It is important to note that algorithmic folk theories are not necessarily inaccurate. Studies show that users can accurately predict the behavior of complex algorithms, and their folk theories contain substantial practical value (S001). This refutes the common misconception that folk theories are merely myths. Rather, they constitute a form of practical knowledge derived from experience that can be as valuable for understanding how platforms actually function as technical documentation.

Folk theories serve important social functions beyond merely filling information gaps. They provide a basis for collective action, help users navigate complex recommendation systems, and influence professional content‑creation practice. The link between illusion of control and algorithmic folk theories is especially significant: users believe they can steer the algorithm through certain actions, which motivates them to experiment and refine their strategies. Understanding this phenomenon is critical for analyzing how people interact with digital platforms and how their sense of fairness and control develops in algorithmic environments.

Key distinction from other cognitive biases:

Algorithmic folk theories are not an individual cognitive bias but a collective social process. They arise not from the reasoning errors of a single person but from interactions among users, platforms, and communities, making them a unique phenomenon in digital culture.

Cognitive Architecture of Folk Theories: How the Brain Constructs Algorithmic Reality

Searching for Causes in Data Noise

The mechanism that generates algorithmic folk theories is rooted in a fundamental cognitive process—searching for cause-and-effect relationships. The human brain is evolutionarily tuned to recognize patterns and attribute causes to them: an adaptive mechanism that helped our ancestors survive by predicting the consequences of their actions. When social‑media users receive feedback in the form of metrics (views, likes, reach), the brain automatically forms a hypothesis about a causal link between the action and the outcome (S004).

The problem is that algorithms operate under high uncertainty and many variables. A user sees a correlation between using a trending sound and an increase in views, but does not see the hundreds of other factors: content quality, posting time, audience size, random algorithmic fluctuations. The brain fills this informational vacuum with an intuitive hypothesis that appears logical and is confirmed by the first successful experiment.

Dopamine, Memory, and Social Amplification

At the neuropsychological level, the process is reinforced through the dopaminergic reward system. When a video “takes off,” it triggers a dopamine release that links the action to the result in memory and creates motivation to repeat the strategy. This explains why users not only believe in folk theories but actively spread them within communities (S004).

Social validation plays a critical role in amplifying the effect. When many people in a community share a theory and report similar experiences, it creates an illusion of consensus and validity. Users exchange “popularity‑boosting” strategies based on collective observations, turning individual hypotheses into group knowledge (S003).

Cognitive Process	Mechanism of Action	Result
Pattern Recognition	The brain looks for causes in correlations between actions and outcomes	Formation of a hypothesis about a causal link
Dopamine Reinforcement	A successful outcome activates the reward system	Consolidation of the link in memory and motivation to repeat
Confirmation Effect	The user notices confirming cases and ignores disconfirming ones	Strengthening belief in the theory despite contradictory data
Social Amplification	Group consensus and strategy sharing within communities	Transformation of an individual hypothesis into collective knowledge
Information Vacuum	Algorithmic opacity creates space for interpretation	Natural filling of the gap with folk theories

Why Theories Appear True: The Confirmation Trap

Algorithmic folk theories seem credible because they are often based on real observations. Users truly see patterns in platform behavior, even if they misinterpret the underlying causes. This creates a paradox: a theory may be inaccurate in explaining the mechanism but correct in predicting the outcome.

The confirmation bias distortion exacerbates the problem. A content creator who believes that posting at 7 p.m. boosts reach will pay particular attention to successful posts at that time and may overlook that other factors (quality, relevance, random fluctuations) played a larger role. Each success is interpreted as confirmation of the theory, each failure as an exception or a mistake in application.

Algorithmic opacity creates an information vacuum that is naturally filled with folk theories. When platforms do not disclose the exact ranking criteria and constantly change them, users are forced to rely on their own observations and the community’s collective wisdom. This is not a user error—it is a rational response to the lack of official information.

From Individual Cognition to Organizational Infrastructure

Research by Karizat et al. (2021) showed that algorithmic folk theories significantly influence users’ self‑presentation and their engagement with the platform. Users do not merely believe in the theories—they organize their behavior and identity around them (S004).

However, the phenomenon extends beyond individual cognition. A study of the organizational infrastructure of multichannel networks (2025) showed that algorithmic folk theories function as organizational tools through which MCNs manage the labor of content creators. This reveals an institutional dimension: folk theories become not just interpretive tools but a structuring force in the social‑media ecosystem (S014).

Madamombe (2025) extended the concept to the professional context of data science, finding that folk theories influence decision‑making by specialists. This demonstrates that the phenomenon is not limited to social media but extends to broader contexts of algorithmic interaction (S001).

It is important to note that user expertise has real value. People can accurately predict the behavior of complex algorithms based on accumulated experience. This means that folk theories should not be automatically dismissed as “unscientific”—they represent a form of practical knowledge that can be as valuable as formal research.

Social media, digital platforms, decision-making

Real-world examples of algorithmic folk theories in action

Scenario 1: TikTok content creator and the “golden hour” theory

Mary, an emerging TikTok creator with about 5,000 followers, noticed that her videos posted between 6 p.m. and 8 p.m. get significantly more views than content posted at other times. She began to schedule posts in that window and indeed saw improved metrics. Mary shared her finding with the creator community, where others confirmed similar experiences, and a collective theory about a “golden hour” for publishing gradually emerged (S004).

However, reality proved more complex. The metric boost could be due to her target audience (students and young professionals) being more active at that time, rather than any special algorithmic preference. Moreover, during that period Mary also upgraded her content quality, started using trending sounds and more effective hooks in the first seconds—factors she omitted from her causal attribution. This illustrates fundamental attribution error, when external factors are credited to internal causes.

The folk theory shaped her identity as a creator: Mary began planning her day around the “golden hour,” feeling pressure to post then even when inconvenient. This illustrates how algorithmic folk theories affect not only content strategies but also self‑perception, daily life, and foster an illusion of control over unpredictable systems (S004).

Scenario 2: LGBTQ+ community and theories of algorithmic censorship

Within TikTok’s lesbian community, a persistent folk theory emerged that the platform’s algorithm systematically suppresses LGBTQ+‑related content, especially videos containing certain keywords or visual cues. Users observed that videos with explicit mentions of queer identity received fewer views and were less often featured in recommendations compared with “neutral” content. In response, the community devised resistance strategies: using coded words (e.g., “le$bian” instead of “lesbian”), visual symbols instead of direct mentions, and creating content that insiders could understand but was less obvious to the algorithm (S003).

This folk theory had profound effects on identity formation and community cohesion. On one hand, it mobilized collective action and generated a sense of solidarity through shared experiences of algorithmic oppression. On the other hand, it influenced how users presented themselves online—many felt compelled to “hide” aspects of their identity from the algorithm, creating tension between authenticity and visibility.

While some elements of the theory may reflect genuine moderation patterns, interpreting those patterns as systematic algorithmic bias was often an oversimplification that ignored the complex factors influencing content distribution. This is an example of confirmation bias, where users notice and remember cases that support their theory while disregarding contradictory examples (S003).

Scenario 3: Data specialist and folk theories in a professional context

Alex, a data specialist at a marketing agency, was building a recommendation system for an e‑commerce platform. Despite formal training in machine learning, he found himself relying on informal “rules of thumb” and folk theories circulating in the professional community: “users always prefer personalization,” “more data is always better,” “algorithms are neutral if the data are balanced.” These beliefs were formed through conference talks, online forums, and mentorship from senior colleagues rather than systematic testing within his specific project (S001).

The assumption that “more personalization is always better” led to a system that created echo chambers and reduced recommendation diversity, ultimately harming user experience and business metrics. Alex encountered the Dunning‑Kruger effect, where his formal education gave him a false sense of competence in applying generic principles to a specific context. He also displayed a bias blind spot, failing to recognize that his own convictions were rooted in folk theories rather than empirical data.

This demonstrates that algorithmic folk theories are not confined to social media or non‑professional users—they permeate professional practice and can have substantial consequences for technology design. When Alex finally conducted A/B testing, he discovered that moderate personalization with elements of diversity outperformed maximal personalization. This case underscores the importance of validating folk theories through empirical methods, especially in professional settings where decisions affect millions of users (S001).