Cognitive biases are often attributed to heuristics or limited information. Yet the structure of social networks is a key, often-overlooked source of perceptual bias. When information passes through social connections, the network alone can systematically distort how individuals view society. We use a simple model in which agents have a binary attribute (e.g., atheist or believer) and show that network topology alone can cause misperceptions of peers' attributes. These misperceptions persist even after aggregation and challenge the idea of the "wisdom of the crowd." We derive an estimator that predicts the size and direction of these biases from network features. We validate our findings using three large-scale opinion surveys. Our results show that network structure is a critical factor in collective perception, with major implications for reducing segregation, polarisation, and the marginalisation of minorities.
Over the past decades, the social and behavioural sciences have shifted profoundly. Researchers have set aside the classical view of humans as fully rational agents, capable of processing all information and computing optimal choices, in favour of a more realistic account of decision-making under cognitive and informational constraints. Simon's seminal work [1] introduced bounded rationality-the concept that humans make decisions using limited cognitive resources and imperfect information. He suggested that people do not evaluate all possible options but instead stop searching once they find one that is "good enough" (a process known as satisficing).
Subsequent research by Kahneman and Tversky [2] demonstrated that humans use cognitive shortcuts, or heuristics-mental strategies or rules of thumb-that systematically deviate from rationality. People display biases such as overconfidence [3], anchoring [2,4], and confirmation bias [5,6] as consistent patterns rather than random errors. Human cognition evolved to maximise efficiency rather than accuracy [7].
However, cognitive biases alone cannot fully account for the collective patterns we observe at larger scales. Beliefs, decisions, and errors spread through social ties, as individuals both influence and are influenced by others. Network science [8,9,10] offers powerful tools to analyse how such heuristics and biases diffuse across groups, shaping opinion formation, collective behaviour, and the spread of (mis)information. Importantly, social connectivity can both result from and reinforce cognitive biases.
A prominent example is the emergence of echo chambers-social environments where individuals are primarily exposed to opinions similar to their own-in online platforms [11,12]. Users in these spaces connect preferentially with like-minded others, often driven by cognitive tendencies like confirmation bias-which favours information supporting existing beliefs [13]-or reinforcement dynamics, whereby repeated exposure to similar views increases their perceived validity [14]. Such mechanisms amplify misperceptions by filtering out dissenting perspectives and exposing individuals only to supportive content. For instance, people often overestimate the popularity of their political views [15,16,17,18], largely because their contacts share similar opinions-an effect observed across online networks [19,20]. Additionally, Lee et al. [3] described how homophily (the tendency to associate with similar others) and minority-group size can together produce perception biases: both over-and underestimation of minority-group size can emerge solely from structural properties of social networks.
In this work, we focus on these network-induced perception biases, defined as systematic distortions in users’ perceptions which emerge from the structure (i.e., the pattern of connections) in social networks, even when individuals are themselves free of cognitive bias. We argue that users extract more information from the network than merely their neighbours’ direct opinions. We extend Lee’s model to account for global network effects, clarifying which specific network attributes contribute to perception biases.
We examine how these structural biases distort the wisdom of the crowd [22], the idea that combining many independent judgments-each person offering their best answer uninfluenced by others-yields a more accurate estimate than relying on any single opinion. This principle grounds recent polling innovations, such as the “election-winner” [23,24,25] and “social-circle” [26,27] questions, which gather perceptions from respondents and their contacts. While these methods show promise, correlated responses (when answers are not independent) or biased judgments (when they systematically miss the true value) can compromise their accuracy [28,29,30,31].
Here, we show that network topology alone can generate perception biases and undermine the wisdom of the crowd. We use a simple model to generate graphs with two communities (e.g., believers and atheists), allowing precise control over community size and interconnectivity. We derive user perceptions by running a message-passing algorithm over network links, revealing how distorted perceptions emerge endogenously. Crucially, these biases persist even when perceptions are averaged across the entire network, undermining the wisdom of the crowd. We identify key structural determinants of the distortion, including community degree, size imbalance, and polarisation. Finally, we derive an analytical expression that predicts the population-level bias and validate it using survey data from [3].
Overall, our findings demonstrate that collective perception is shaped jointly by cognition and network architecture. Recognising these topological sources of bias can help explain persistent societal misperceptions-such as political belief gaps between groups [32]-and improve collective forecasting in settings where aggregation fails due to netwo
This content is AI-processed based on open access ArXiv data.