Six Degrees of Separation’s New Form: The Wikipedia Speedrun (2024)

The idea of six degrees of separation is that all people on average are six or fewer social connections to each other. This means that any two people should be able to connect by making a “chain of friends” with six steps at maximum. So if I don’t know Klaus personally, but I know Beatrice, who is friends with Klaus, then there are two connections linking me and Klaus (and one connection linking me and Beatrice). Of course, six degrees of separation is only true on an aggregate average, and won’t necessarily hold on the individual level.

This concept is also the foundation behind the Bacon Number, which originates from the six degrees of Kevin Bacon, which you can learn more about at The Oracle of Bacon. An actor’s Bacon Number is the number of connections between them and Kevin Bacon, with connections being defined as two actors being in the same film/movie. This means that Michael Fassbender, who was in X-Men First Class, has a Bacon Number of 1, since Kevin Bacon starred as Sebastian Shaw in the same film. With the same method of connection, the Bacon Number of Elon Musk is 2, since Elon was in Iron Man 2, which Sam Rockwell was in, who was in Frost/Nixon, which Kevin Bacon was in.

There are plenty of applications for degrees of separation, and it’s surprising for how many types of connections the average of six is true for. Degrees of separation can be found in computer science data structures, social media/networking, and most interestingly (in my opinion), online encyclopedias.

Wikipedia is a free, multilingual, open-collaborative online encyclopedia that is entirely maintained by volunteer contributors using a wiki-based platform. If there is a topic that you’re interested in, Wikipedia is bound to have an article on it (or at least something related to it). I’m sure you’ve ran into Wikipedia before, whether it was for a school project, just a quick Google search, or linked as a source for an essay. The latter of these examples has been greatly debated for Wikipedia’s reliability, but that’s not what I’m concerned with today.

A speedrun is a playthrough of a game, done with the intention of completing it as quickly as possible. Popular speedruns involve Minecraft, platforming games, and games with story modes. At some unknown point in Wikipedia’s lifetime, someone came up with the ingenious idea to turn Wikipedia’s pages into a speedrunning platform. Using degrees of separation, the goal would be to go from one page to another as fast as possible only by clicking links (literally links of connection). However, at some point in 2020 specifically, someone came up with the humorous idea called the “Hitler speedrun,” which involved clicking on Wikipedia’s built-in “random article” feature and getting to Adolf Hitler’s page as quickly as possible by only clicking links. Just giving “Hitler speedrun” a search on YouTube will give you a plethora of videos demonstrating the popularity of this meme of a speedrun.

Inspired by this comedic implementation of degrees of separation, my friends and I decided to come up with a surprisingly entertaining game to pass some time during this global pandemic.

The premise of this game is very simple. First, visit the main page of Wikipedia. From here, click “random article” on the left-hand side of the page. This is your destination article, meaning it’s where you need to end. Study it for a bit (although it’s much more fun being less prepared), and get your timer ready. Open up a new Wikipedia tab (which you can do by middle-clicking the Wikipedia logo, or just by duplicating your current article), and click on “random article” again. The moment you do this, the timer starts! Going as fast as possible, try to make it to your destination article by clicking links on the page without leaving the Wikipedia website. It sounds simple, but it’s also surprisingly fun. Whether or not you use browser tools such as the search function (CTRL + F) and back and forth browser buttons is up to you, and you can always make it more challenging by eliminating any preparation time looking at the destination article. Time ends precisely when you get to your destination article, and that destination article only. Any related articles that are extremely close (article titles will often differ by just one word or phrase in parentheses) don’t count.

Being a professional Wikipedia speedrunner myself, I’ve come across my fair share of routes to get from one article to another. Though it was quite daunting at first, I figured out how to go from Sound of Jura to Euproctidion without hesitation. What follows are a few routes that I’ve refined over the course of navigating through a few Wikipedia speedruns.

Geography route: Though a geography route might be used to go directly to your destination article, it’s also quite the universal route for starting a run. Most of the time, the two articles that you get are completely unrelated, and it’s very difficult to establish any obvious connections between the two. That’s where the geography route (usually) kicks in. For example, if we need to go from Digital Combat Simulator (a flight simulator game) to Tile Hill Wood (a small forest in Tile Hill, in Coventry, in England), there’s a clear geography route that will end the run. Here is an example of a full route: Digital Combat Simulator>P-51D Mustang>World War II>United Kingdom>England>Coventry>Tile Hill>Tile Hill Wood. With decent browser skills, this run can be easily done in under 60 seconds. Of course, this probably isn’t even the most efficient route; there should be a route with even fewer connections. Getting a route with 7 connections is nothing to complain about, though. Since we’re just humans and not computers calculating the shortest route possible, it’s far more likely that time is dependent on input speed rather than number of connections. Even so, it’s nice to get a relatively short route like this one. In the following route types, you’ll understand why the geography route is useful even when the ending article isn’t directly related to geography.

Biology route: If you need to end on a species, genus, or anything related to biology for that matter, there’s a good chance that you can get there from a species’ page. Since an animal like a jaguar will have a page with its taxonomic classification, that means you also have access to the entirety of the taxonomic classification system itself. So how do we get from any random article to a species consistently? The geography route of course! Let’s use the articles that I gave earlier as an example. Euproctidion is a genus of moths, so it’s definitely a good idea to go with a biology route. But our starting article is… Sound of Jura? What even is that? It turns out that Sound of Jura is a strait in Argyll and Bute, Scotland. With a direct connection to geography, we can also gain access to ecology of geographic locations. With ecology, we can access a random species — which will then allow us to access the taxonomic classification system and finish the run. Here’s an example of a full route: Sound of Jura>Scotland>Europe>continent>North America>jaguar>Animalia>arthropods>Insecta>Lepidoptera>Erebidae>Lymantriinae>list of Lymantriinae genera>E>Euproctidion. 14 connections makes this exact route possible, which although much higher than 6, can still lead to a very fast time. Even if we didn’t start with a direct connection to geography, there is pretty much bound to be a connection to geography. Getting a person guarantees a close connection to geography (whether it’s where they were born, the country of the sports team that they play for, etc.). Getting something created by a person/people will also have guaranteed close connections to geography (an opera was written by someone, who was born somewhere). Getting a court case will give you people, who must have born somewhere or have some relation to geography. The list goes on and on, but it seems that geography has some of the strongest connections…

Sports route: For some reason, over 50% of the time when I click “random article,” I get an athlete (okay, maybe not over 50%). Going for a sports route is always accessible via a geography route (no surprise). It’s a lot more clear if I provide an example prioritizing geography but getting to sports. Here is an example route going from southern rockhopper penguin to Francisco Rueda (diver): southern rockhopper penguin>South America>continent>North America>Northern America>United States>American football>team sport>sport>diving>1908 Olympics in London>1980>Mexico>Francisco Rueda. With 13 connections, this is far from an optimal route. Even so, it demonstrates how powerful geography is. The southern rockhopper penguin, which has nothing to do with human diving as a sport (the diving that penguins do doesn’t directly link to human diving) is connected to sports by… geography! With virtually every page containing a link to geography, all that needs to happen is to find a connection from geography to sports — something I’ve already refined. One of the simplest way to get to sports in general is to pay the United States page a visit, then go to American football, which is a team sport, which links you to “sport.”

Tools route: If you still think that geography isn’t that strong of a connection category, take a look at this route going from Civita (a Norweigian liberal think tank) to adze (a primitive cutting implement): Civita>Norwegian>Northern Europe>England>England in the Middle Ages>swords>knife>tool>hand axe>adze.

After just listing a few routes, it’s quite evident that geography is somehow “stronger” than the other routes. This is in the sense that geography is used in almost every single run in which a single direction connection cannot be established. Of course this is taking into consideration that humans are performing such speedruns, since a computer program would be able to very quickly identify which pages point to the ending page and reverse engineer the simple problem to find obscure connections that wouldn’t make logical sense to humans in the first place. Taking a look at the six degrees of Wikipedia page gives a few examples of connections that were likely stumbled upon by obscure knowledge or by happenstance. Let’s consider a route that would be taken to go from Ø(the symbol) to Underoath(an American rock band that plays heavy metal and metalcore). This is what I came up with (I think it’s logically sound): Ø>Norwegian(alphabet)>Norwegian(language)>Norway>Northern Europe>Europe>England>Music of the United Kingdom>heavy metal>metalcore>List of metalcore bands>Underoath. With 11 connections, it’s not too bad. It could certainly be better, but probably not by much. I could have went from Europe and gone to the United States and probably would have only added a couple of connections at most. Now this is a route that no one in their right mind would come up with: Ø>Underoath. That’s it. There’s a direct connection between these seemingly unrelated pages! Of course, with just a little bit of digging, you can find out that Underoath published an album with this character in its title, but how would you know that? I’m not even sure how I know that.

Regardless of how horribly inefficient the path that I *logically* came up with, I’m still convinced that there is a connection (no pun intended) to be made between Wikipedia pages and neural networks. Here’s a brief explanation:

Neural networks are computing models that take inspiration from biological computing units, or more precisely, the human brain. A neural network consists of nodes, individual processing entities that all have “weights” between them. Each weight determines the strength of connection between multiple nodes. These weights will ultimately determine the flow of activation throughout the network when given an input (Bruckner & Garson 2).

The structure of a neural network can be described in threefold: The input units, output units, and the hidden units in between. All input units have an activation value that will be sent to the hidden units of the system. Whether or not the system “understands” these activation values, they all represent something extraneous to the network itself. For example, the input unit activation values of a network may stand for processing an image and detecting if it contains a specific color. The output of such a task, however, does not occur until the activation values of the input units are sent to the many layers of hidden units, which subsequently have their activation values sent to connecting nodes. The initial input given will spread via nodes passing on their activation values to other nodes, and eventually, finishing with the activation values of the output nodes.

Neural networks are designed to accomplish a task while learning. A neural network “learns” due to the dynamic property of weights. The connections between nodes can be increased or decreased, based on the model of learning and the input given. A proper neural network will adjust the weights to succeed more frequently and more accurately at the task as it is given more and more trials.

Do you see the connections between Wikipedia and a neural network? It might seem terribly obvious, but at the same time, somewhat unclear. The physical representation of a neural network matches that of Wikipedia link connections, but does that constitute an application of a neural network?

That might not be the case, but it’s certainly a possibility. If we think about how a computer would solve a Wikipedia speedrun however, it’s evident that with all of the possible connections stored in a database, there would be no need for a “learning” model. Rather, any calculations performed to go from start article to destination article as quickly as possible would match a simpler model of connected nodes. Perhaps the most important part of optimizing a program that solves for the most efficient route from start to end would be to implement the most efficient data structure. And that’s what I’ll be looking into in an upcoming blog post.

Have fun speedrunning Wikipedia!

the entirety of Wikipedia

https://en.wikipedia.org/wiki/Wikipedia:Six_degrees_of_Wikipedia#Six_Degrees_to_Kevin_Bacon