Why do some ideas, products, or political candidates take off, even when they may not be the best option?
In new research published in Science Advances, EUI Professor Arnout van de Rijt, Head of the Department of Political and Social Sciences, and co-authors Alexandros Gelastopoulos, Pantelis P. Analytis, and Gaël Le Mens, examine how even small differences in popularity can shape collective decisions. In this interview, Prof. van de Rijt explains why crossing the line into "majority" status can matter more than we might expect, and what that means for elections, online platforms, and public opinion.
Your study finds that when something becomes just slightly more popular than its alternative – enough to be seen as the majority – people treat it differently. Why does that small shift matter so much?
A key insight from our study (what we call the marginal majority effect) is that people don’t always respond smoothly to popularity. Instead, there is often a sharp psychological or behavioural shift at the exact moment when one option becomes the majority, even if that majority is extremely small. When an option goes from being just below 50% support to just above 50%, people begin to treat it differently, because it now qualifies as "what most people are choosing."
One mechanism behind this effect is that many people rely on simple heuristics like ‘follow the majority’ These shortcuts are efficient, because they save time and effort, but they also create a tipping point. When that threshold is crossed, the fact that something is now technically the majority becomes a powerful signal in itself, regardless of how small the difference actually is. Another mechanism is more mechanical. The common practice of ranking things by popularity, visually or in textual presentation, leads options to change ranks at the threshold. This makes them more visible and displayed more prominently. Bestseller lists, election polls, music charts, and online sorting algorithms present choice alternatives by how often they have been chosen before, in many domains of industry and culture.
Importantly, our results show that this shift is often not gradual. It’s not just that influence slowly increases with popularity. Instead, there can be a sudden jump in influence when an option crosses the majority line. That jump makes the majority self-reinforcing: Once something is slightly ahead, people are more likely to choose it precisely because it is ahead.
In other words, a small numerical difference can translate into a big behavioural difference, because it changes how people interpret the social signal. That change in interpretation, more than the size of the difference itself, is what gives marginal majorities their power.
Why can that small early advantage sometimes cause a group to stick with an option that isn’t actually the better one?
The key mechanism is feedback, or reinforcement. Early in a group decision process, chance events can give one option a slight popularity advantage, even if it is objectively worse than another option. Once that happens, people become more likely to choose that option simply because it is now the majority. This further increases its popularity, which makes it even more attractive to future decision-makers, and so on.
Our theoretical results show that when the influence of being the majority is stronger than the inherent quality advantage of the better option, the system can ‘lock in’ on the inferior option. In other words, social influence can overpower quality differences.
This dynamic helps explain why qualitatively inferior technologies, ideas, or beliefs can persist despite better alternatives. The issue isn’t that people can’t recognise quality, but rather that social signals amplify early randomness. Once enough people see something as the majority choice, they treat it as the safe or appropriate option, or are swayed by its popularity into assuming it is the better option, even against their own quality judgment. This then reinforces and stabilises that outcome.
So, the group doesn’t necessarily converge on the best option. It might. But often, it converges on whichever option happens to gain a slight early advantage and is then reinforced by social influence.
Previous studies have shown mixed results – sometimes groups correct themselves, and sometimes they don’t. What was missing from earlier explanations?
The puzzle in prior research was that social influence sometimes leads to lock-in and sometimes leads to correction – cases where one option briefly gains a majority, but the group eventually shifts back toward the better alternative. Existing theories didn’t provide a clear, unified explanation of when each outcome would occur.
Our contribution was to identify a specific behavioural condition that determines the outcome: the relative strength of the marginal majority effect compared to the inherent quality difference between the options. If the majority effect is stronger, lock-in becomes possible. If quality differences are stronger, the system tends to self-correct.
Earlier models often assumed that influence increases gradually as popularity increases. But our findings show that the critical factor is not gradual influence; it’s the discontinuity at the majority threshold. Whether lock-in happens depends on what occurs at that moment when an option crosses into majority status.
We tested this idea across multiple experimental datasets, from political opinions to factual questions to aesthetic judgments. We found that it consistently explained when lock-in occurred and when it didn’t. In almost every case where the marginal majority effect exceeded the quality difference, lock-in was observed. When it didn’t, groups tended to correct themselves and arrive at the higher quality alternative.
What was missing, then, was the recognition of this specific behavioural mechanism. Once you measure how strongly people respond to marginal majorities, you can predict whether a system will stabilise around the better option or get stuck on an inferior one.
In today’s world – where we constantly see polls, ratings, and trending lists – are we more exposed to these effects, and does that change how we should think about the way information is presented?
Yes, very much so. Modern digital environments make popularity signals highly visible and constantly updated. We see which posts are trending, which products have the most reviews, which opinions are most liked, which candidates are leading in polls. These ranked positions and associated signals make even very small differences in popularity highly salient.
Our research shows that simply being labelled or perceived as the majority – for example, through language like ‘X is leading in the polls’ or ‘song Y has been #1 since xx’ – can have a disproportionate impact on people’s choices, even when the margin is low. Online platforms can amplify this effect by highlighting rankings, counts, and trends, drawing attention to who is ahead rather than how large the difference actually is.
In some cases, the effect can be even stronger than in offline settings. For example, when platforms emphasise rank, such as showing ‘#1 trending’, they create a clear majority signal even when the underlying difference is small. This can accelerate feedback loops where early random advantages quickly become entrenched.
At the same time, this doesn’t mean lock-in is inevitable. If people have strong independent evidence or convictions about quality, or if popularity signals are less emphasised, groups can still converge on better outcomes. But the increasing visibility of popularity information means that marginal majorities are more likely to influence decisions than before.
So, yes, today’s information environment makes social influence more immediate and visible, which increases the likelihood that small initial differences can lead to large collective outcomes.
Read the article ‘The marginal majority effect : when social influence produces lock-in’, published in the journal Science Advances.
Arnout van de Rijt is Professor and Head of the EUI Department of Political and Social Sciences. Van de Rijt received his PhD in Sociology from Cornell University in 2007 and worked until 2016 as Assistant and Associate Professor of Sociology at Stony Brook University, where he co-founded and -led the Center for Computational Social Science. He was Professor of Sociology at Utrecht University from 2016 through 2019. He is president of the International Network of Analytical Sociology and elected member of the European Academy of Sociology. Van de Rijt received the Lynton Freeman (2010) and Raymond Boudon (2017) early career awards. Van de Rijt is Editor-in-Chief of Sociological Science.