Facebook said its users were separated by 3.57 degrees, but the 2016 result meant 4.57 friendship links on average across 1.59 billion active accounts, not all humanity

In February 2016, Facebook announced that the average user was separated from every other user by 3.57 degrees. It sounded like the old “six degrees of separation” had been measured at planetary scale and cut nearly in half.

The result was real within Facebook’s analysis, but the familiar summary loses three important details. The graph contained 1.59 billion active Facebook accounts, not every person alive. The number was an estimated average, not a maximum applying to every pair. And Facebook counted “degrees” as the people between two users, making 3.57 degrees equivalent to an average shortest path of 4.57 friendship links.

A friendship network becomes a graph

Network analysis turns a social platform into a mathematical object. Each Facebook account becomes a node. Each confirmed friendship becomes an undirected edge connecting two nodes. The distance between two accounts is the smallest number of edges needed to travel from one to the other.

If Alice is friends with Ben, their graph distance is one. If Ben is friends with Carla but Alice is not, Alice and Carla have a distance of two, with Ben as the single intermediary.

The term “degree of separation” is not always counted consistently in popular writing. Some accounts use it for the number of links. Facebook followed the convention in which degrees equal graph distance minus one. Its 3.57 figure therefore described 3.57 intervening people on an average path, or 4.57 Facebook friendship edges.

A decimal does not mean a path contains part of a person. Some pairs were closer and some farther apart. Averaging those whole-number path lengths produced 4.57 links and, after subtracting one, 3.57 intermediaries.

The 2011 study established the counting convention

Facebook’s 2016 announcement built on a much more detailed 2011 analysis. In Four Degrees of Separation, Lars Backstrom, Paolo Boldi, Marco Rosa, Johan Ugander and Sebastiano Vigna analysed the complete network of roughly 721 million active Facebook users and about 69 billion friendship links.

They estimated an average graph distance of 4.74 links. The paper explicitly translated that into 3.74 intermediaries or degrees of separation. It also found that 99.91 per cent of accounts belonged to one enormous connected component, meaning a friendship path existed between almost every pair in the analysed graph.

The network was too large for a simple all-pairs calculation. The researchers used probabilistic graph algorithms to estimate how many accounts lay within each distance of every other account, then checked the accuracy of those estimates through statistical metadata. The paper was later published in the proceedings of the 2012 ACM Web Science conference.

Facebook’s 2016 number was 3.57 intermediaries

Five years later, Facebook had more than doubled the graph to 1.59 billion active accounts. In its 2016 research post, “Three and a half degrees of separation”, the company said the typical pair’s separation had fallen from 3.74 to 3.57.

Facebook again used probabilistic counting rather than checking every possible pair directly. The company described an approach based on the Flajolet-Martin algorithm, which estimates the number of distinct items in very large datasets without storing every item individually.

The reduction from 3.74 to 3.57 was plausible because a larger, denser graph can acquire new bridges between communities. A single friendship connecting two previously distant clusters can shorten many paths at once. Highly connected users and cross-border relationships can have an outsized effect on average distance.

That does not mean growth automatically makes every network closer. Adding isolated or weakly connected users could produce the opposite result. The 2016 figure described what Facebook estimated in its own graph at that moment.

It did not measure all humanity

The phrase “everyone in the world” appeared in much of the coverage, but the relevant population was Facebook’s active-user graph. In 2016, 1.59 billion accounts represented a large share of humanity, yet billions of people were outside it.

Accounts were not equivalent to unique, verified humans. Some people had no account, some had more than one, and access to Facebook varied sharply by country, age, income and internet availability. Platform friendships also differed from offline acquaintance. A confirmed Facebook edge might represent a close relationship, a former classmate, a professional contact or someone met once.

The result therefore cannot establish that any two humans on Earth were separated by 3.57 intermediaries. It established an estimated mean within the connected portion of a particular digital network.

An average is not a promise about any pair

The “six degrees” phrase is often interpreted as a guarantee that everyone can reach everyone else within six links. Facebook’s figure was different: it averaged shortest paths across reachable pairs.

An average of 4.57 edges can include many pairs separated by three or four links and others separated by seven, eight or more. People near dense hubs generally have shorter average paths than people connected through a sparse local cluster.

Nor does a short path mean that information, influence or an introduction will actually travel along it. A graph identifies possible connections. It does not show whether intermediaries know one another well, see a post, trust its source or agree to pass on a message.

This distinction matters whenever small-world statistics are used to imply easy access to investors, employers or public figures. Four friendship edges in a database are not four warm introductions.

“Six degrees” was never a fixed law

The small-world idea long predates Facebook. Frigyes Karinthy explored short acquaintance chains in a 1929 story, and Stanley Milgram’s experiments in the 1960s asked participants to route documents towards a target through people they knew personally.

In a 1969 paper by Jeffrey Travers and Milgram, completed chains reached the target through a surprisingly small number of intermediaries. But many chains never arrived, the starting samples were limited, and the experiment did not measure every relationship on Earth. The memorable number six became a cultural shorthand, not a universal constant established by one experiment.

Different networks naturally produce different average distances. Microsoft researchers later examined 30 billion instant-message conversations among 180 million users and found a median path length of seven links in that communication graph. Facebook’s reciprocal friendship network was structured differently and much larger.

The number says something real about platform scale

The most defensible reading of 3.57 is narrower and more interesting than the slogan. By 2016, Facebook had assembled a digital friendship graph so vast and densely bridged that two active accounts in its main connected component were typically linked through fewer than five friendship edges.

That architecture helps explain why online platforms can feel both enormous and intimate. Local circles overlap through school, work, migration, family and public-facing accounts. A small number of bridges can pull distant communities into the same navigable graph.

It also explains why network structure matters to recommendation systems, privacy analysis and the spread of content. Short paths create opportunities for rapid movement, though platform algorithms and human choices still determine which opportunities are used.

Facebook’s calculation did not prove that humanity had precisely 3.57 degrees of separation. It showed that 1.59 billion active accounts in one 2016 platform snapshot had an estimated average of 3.57 intermediaries, or 4.57 friendship links. Keeping those units and boundaries intact makes the result less magical, but considerably more informative.