Friendship paradox biases perceptions in directed networks

Abstract

How popular a topic or an opinion appears to be in a network can be very different from its actual popularity. For example, in an online network of a social media platform, the number of people who mention a topic in their posts-i.e., its global popularity-can be dramatically different from how people see it in their social feeds-i.e., its perceived popularity-where the feeds aggregate their friends' posts. We trace the origin of this discrepancy to the friendship paradox in directed networks, which states that people are less popular than their friends (or followers) are, on average. We identify conditions on network structure that give rise to this perception bias, and validate the findings empirically using data from Twitter. Within messages posted by Twitter users in our sample, we identify topics that appear more frequently within the users' social feeds, than they do globally, i.e., among all posts. In addition, we present a polling algorithm that leverages the friendship paradox to obtain a statistically efficient estimate of a topic's global prevalence from biased perceptions of individuals. We characterize the bias of the polling estimate, provide an upper bound for its variance, and validate the algorithm's efficiency through synthetic polling experiments on our Twitter data. Our paper elucidates the non-intuitive ways in which the structure of directed networks can distort social perceptions and resulting behaviors. * N. Alipourfard and B. Nettasinghe contributed equally to this work.

How behavioural sciences can promote truth, autonomy and democratic discourse online

“…The interactive nature of social media could be harnessed to promote diverse democratic dialogue and foster collective intelligence. To achieve this goal, social media needs to offer more meaningful, higher-dimensional cues that carry information about the broader state of the network rather than just the user's direct neighbourhood, which can mitigate biased perceptions caused by the network structure 99 . For instance, social media platforms could provide a transparent crowd-sourced voting system 100 or display informative metrics about the behaviour and reactions of others (for example, including passive behaviour, like the total number of people who scrolled over a post), which might counter false-consensus effects.…”

Biased perceptions explain collective action deadlocks and suggest new mechanisms to prompt cooperation

“…Similarly, Galesic et al show that homophily and a sampling process whereby individuals derive their judgments from local information based on their social environment (e.g., family, friends, and acquaintances) can explain when false consensus or false uniqueness is expected to occur . Alipourfard et al further show that individuals' perceptions can be biased as a result of local correlations in a directed social network (Alipourfard et al, 2020), and Lerman et al show that social network effects can lead individuals to overestimate states that are globally rare, if those are overrepresented in their local neighborhoods-a phenomenon named majority illusion (Lerman et al, 2016). If perception biases result from social network effects rather than cognitive flaws, interventions based on reshaping information flows about global behaviors are possible and can be very impactful.…”

“…Article to be and change over time is indispensable for a mechanistic understanding of the feedbacks between interventions and the biases themselves. As mentioned above, the existence of perception bias can be a by-product of individuals' psychological states, as well as the influence of local assortment (Cooney et al, 2016;, specific network topologies (Alipourfard et al, 2020), or information filtering. False consensus, particularly, is likely to emerge if individuals' opinions assort them.…”

Dynamics of behavior change in the COVID world

“…As a result, the original posts represent 69% of total social media engagement in a recently analyzed sample of coronavirus misinformation (Brennen, Simon, Howard, & Kleis Nielsen, 2020). When disinformation spreads more easily through common channels—such as friends who are generally trustworthy or prestigious partisan demagogues on social media—than does accurate information broadcast by some other source (eg, the CDC), it can produce clusters of people who learn and reinforce disinformation through their social connections (Alipourfard, Nettasinghe, Abeliuk, Krishnamurthy, & Lerman, 2020; Lerman, Yan, & Wu, 2016). Social media may increase this risk drastically by enabling a few people to broadcast their opinions to millions of others (Brennen et al, 2020; Krause, Freiling, Beets, & Brossard, 2020), and by facilitating assortment according to shared opinions.…”