What is Bridging?

A design principle that ranks highest the things supported by people who usually disagree. Instead of ordering content by likes or views, it values support that crosses divides, so that what rises to the top is what both sides accept. Community Notes and Polis are the best-known examples.

Open social media and you may see nothing but posts that stoke anger or mock the other side. Critics point out that this is no accident: many platforms are built to push up whatever gets the most reactions. So what would happen if the ranking rule were changed? Bridging is an idea for doing exactly that.

In short

Bridging is a principle for ranking and design that highlights what is supported by people who usually disagree [1][2].

Rather than pushing up "whatever gets the most reactions", as many social media platforms do, it pushes up what is supported across divides [1][2].

For example

Community Notes (X). On X (formerly Twitter), users can attach explanatory notes to misleading posts. A note is shown only when people whose past ratings have tended to differ agree that it is "helpful" [3]. It is not a majority vote. The 2022 paper on this system used the phrase "Bridging Algorithms" in its title to describe this way of selecting notes [4].

Polis (used in Taiwan's vTaiwan and elsewhere). Polis, a tool for gathering opinions, finds opinion groups from participants' agree/disagree votes and shows the statements that every group agrees with [5]. The book Plurality says that "Polis shows the clusters of opinion that exist and highlights statements that bridge them" [6].

Applying it to posts themselves. X has started a small test that looks not only at notes but at posts, to find "posts liked by people with different perspectives" [7]. It states that, for now, this does not affect how widely a post is shown [7].

How it works

What users do

Bridging systems ask nothing special of users [1][3]. Users rate things "helpful" or "not helpful", or click "agree" or "disagree", as usual.

What changes is how the ratings are counted. What rises to the top is not "what is popular with my side" but "what the other side also accepts" [1][3].

Why it is needed

In a 2022 paper, the researcher Aviv Ovadya argued the following [1]. Recommendation systems on Facebook, YouTube, TikTok and similar platforms, through what they choose to show, act as a system that tells all of society what gets rewarded. When ranking by engagement combines with human psychology, divisive behavior is what earns the reward of attention [1].

According to Ovadya, this shapes who succeeds — politicians, journalists, entertainers and others [1]. And it goes on to affect society's ability to make good decisions, its ability to cooperate, and how likely violent conflict becomes [1].

Bridging-based ranking was proposed as a force that pulls people together, to counter this force that pushes them apart [1]. Ovadya argues that other proposed reforms, such as showing posts in chronological order, will keep rewarding those who aim to divide unless they include bridging, and that bridging should be the default [1].

Defining the term

Ovadya and Luke Thorburn define bridging not by method but by outcome [8]: "an increase in mutual understanding and trust across divides, creating space for productive conflict, deliberation, or cooperation." Note the words "productive conflict". The aim is not to eliminate conflict, but to make fruitful conflict possible.

How it is calculated: two main types

Matrix factorization (Community Notes). Ratings are explained by a calculation of the form "overall generosity + the rater's generosity + the note's own quality + (the rater's position × the note's position)" [4][9]. The multiplied term soaks up the part that comes from "liked only by one's own side", so the remaining "note's own quality" represents support across sides [4][9].

Multiplying across groups (Polis). Participants are sorted into opinion groups by their voting patterns, and the probabilities that members of each group agree (estimated from the share of agree votes within each group) are multiplied together [10]. If agreement is low in even one group, the value drops, so statements supported only by the majority don't rise to the top [10]. A worked example with numbers is in the Polis article.

Vitalik Buterin, the founder of Ethereum, calls these a new kind of mechanism design: one that deliberately seeks out divisions, and then favors what bridges them over what entrenches them [9].

Effects on misinformation

In a 2024 paper, Peng and Grimmelmann argue that if you promote the responses to misinformation that are liked both by people likely to agree with the misinformation and by people likely to disagree with it, you select the responses best able to persuade the people most likely to believe it [2]. Because this takes judgment out of the company's hands and gives it back to users' votes, they say, it is also consistent with the tradition of free speech [2].

What is debated

It misses the most divisive topics, and fake consensus can be manufactured. For Community Notes, one study argues that precisely because the algorithm captures each country's main axis of conflict well, it under-corrects the most polarizing content [11]. Another showed that a small number of coordinated votes can produce a manufactured consensus [12]. The details are in the Community Notes article.
Who decides what counts as a divide, and what counts as a bridge? bridging.systems, the website of bridging researchers, acknowledges that which goals get built into a system is a value judgment and a contested question [13]. It then suggests deciding through systems that gather many people's views, or through meetings of citizens chosen by lottery [13]. Ovadya too proposes "platform democracy", in which the people affected are asked what recommendation systems should reward [1].

A fuller collection of criticisms is in Criticisms of Plurality.

Common misunderstandings

It is not "picking the middle". What Community Notes looks at is not users' self-reported political positions, but positions inferred from their past rating patterns [3]. And Plurality values not only finding agreement but also giving rise to new conflicts (see the next section) [14].

Where it fits in Plurality

In chapter 5-4, "Augmented Deliberation", Plurality responds to the worry that engagement-ranked social media entrenches divisions by presenting Community Notes as an attempt to redesign these systems "with the opposite intention of 'bridging' the crowd" [14]. Polis is also presented as a prime example of what Ovadya and Thorburn call "bridging systems" [14].

The same chapter stresses that "while most discussion of bridging systems focuses on building consensus, another powerful role is to support the regeneration of diversity and productive conflict" [14]. By finding opinion groups without being bound by old assumptions and affiliations, new lines of conflict can emerge that do not simply trace existing divisions [14].

Chapter 5-0 of the book also points out a danger: "Bridging differences for collaboration often erodes them" — everyone may become more alike [15]. But it argues that "homogenization is not an inevitable outgrowth of bridging", because bridging also gives rise to new fields: for example, "the encounter between psychology and economics has created a new 'behavioral economics' field" [15].

In a Chinese-language column, Audrey Tang describes bridging-based ranking as something that connects two opposing groups and helps them find views both can agree with [16].

Related articles: - Community Notes — the leading example of the matrix factorization type - Polis — the leading example of multiplying across groups - Augmented Deliberation — the chapter of the book that discusses bridging - Prosocial media — a vision of extending bridging to the design of social media as a whole - Collaboration across social difference — the heart of Plurality's definition

Sources

Aviv Ovadya, "Bridging-Based Ranking: How Platform Recommendation Systems Might Reduce Division and Strengthen Democracy" (Belfer Center, 2022) — https://www.belfercenter.org/publication/bridging-based-ranking ［CC BY-NC 4.0］
Peng and Grimmelmann, "Rescuing Counterspeech: A Bridging-Based Approach to Combating Misinformation" (2024) — https://arxiv.org/abs/2410.12699 ［CC BY 4.0］
Community Notes Guide, "Diversity of perspectives" — https://communitynotes.x.com/guide/en/contributing/diversity-of-perspectives ［Apache-2.0］
Wojcik and others, "Birdwatch: Crowd Wisdom and Bridging Algorithms can Inform Understanding and Reduce the Spread of Misinformation" (2022) — https://arxiv.org/abs/2210.15723 ［CC BY 4.0］
Small and others, "Polis: Scaling Deliberation by Mapping High Dimensional Opinion Spaces" (2021) — https://doi.org/10.6035/recerca.5516 ［CC BY-SA 4.0］
Plurality, 2-2 "The Life of a Digital Democracy" — https://www.plurality.net/v/chapters/2-2/eng/ ［CC0］
Community Notes Guide, "Beyond notes: posts liked by people with different perspectives" — https://communitynotes.x.com/guide/en/beyond-notes/got-likes ［Apache-2.0］
bridging.systems, "Definitions" — https://bridging.systems/definitions/ ［MIT］
Vitalik Buterin, "What do I think about Community Notes?" (2023-08-16) — https://vitalik.eth.limo/general/2023/08/16/communitynotes.html
Small and others (2021), as above: the calculation of cross-group consensus — https://doi.org/10.6035/recerca.5516 ［CC BY-SA 4.0］
Bouchaud and Ramaciotti, "Community Notes undermoderate polarizing content by design creating risks in electoral processes" (2025) — https://arxiv.org/abs/2506.15168 ［CC BY-NC-ND 4.0］
Selvam and others, "Gaming Consensus: Coordinated Manipulation in Crowdsourced Fact-Checking" (2026) — https://arxiv.org/abs/2607.01824 ［CC BY 4.0］
bridging.systems, "Who decides?" — https://bridging.systems/who-decides/ ［MIT］
Plurality, 5-4 "Augmented Deliberation" — https://www.plurality.net/v/chapters/5-4/eng/ ［CC0］
Plurality, 5-0 "Collaborative Technology and Democracy" — https://www.plurality.net/v/chapters/5-0/eng/ ［CC0］
Audrey Tang, "讓演算法變得善良" (Making algorithms kind; Business Weekly column, 2024-07-11, in Chinese) — https://sayit.archive.tw/2024-07-11-%E5%95%86%E5%91%A8%E5%B0%88%E6%AC%84%E8%AE%93%E6%BC%94%E7%AE%97%E6%B3%95%E8%AE%8A%E5%BE%97%E5%96%84%E8%89%AF ［CC0］

We welcome error reports and questions ("What is…?"). Submissions become public issues.

Report an error in this article ・ Ask a question