Every month, more than 80,000 volunteers donate their time and talent to help edit Wikipedia, the free online encyclopedia of everything. Unlike professional journalists, those writers and editors work for free, are largely anonymous, and labor in an unstructured environment with no clear lines of authority. And if they receive any acknowledgment at all, it’s buried several layers deep on Wikipedia.
Yet 15 years after its first articles were published, the English language version alone (there are 282 other active versions) now includes more than 5 million articles and 28 million registered users, according to Wikipedia’s parent, the Wikimedia Foundation.
The unlikely nature of Wikipedia’s ability to harness the unpaid efforts of so many people intrigued two researchers who had studied together at the London School of Economics: Stephan Seiler, an economist and associate professor of marketing at Stanford Graduate School of Business, and Aleksi Aaltonen, an assistant professor of information systems at Warwick Business School.
“I wanted to know if we could quantify the effect of such an open production process on productivity,” Seiler says.
After analyzing 1,310 articles that were edited nearly 63,000 times over eight years, the answer they found was almost stunningly simple: Edits to articles act like magnets to attract other editors, and the edits performed by the new editors attract even more edits, and so on.
This “cumulative growth effect” accounted for nearly half of the content growth between 2002 and 2010 for a typical article, the researchers found. And the articles weren’t just longer — they were better, Seiler and Aaltonen write in a research paper published in the July 2016 issue of Management Science.
Wikipedia’s success has spawned many copycats. Major companies and nonprofits host internal wikis to create, share, and store knowledge. Because many of them use the same open source software and page layout employed by Wikipedia, Seiler and Aaltonen believe their findings are broadly applicable to similar platforms.
Adding Up the Edits
Wikipedia, Seiler says, is an excellent subject to study because so much of its data is available and in a standard format. The researchers used a dataset containing articles from the launch of the online encyclopedia in January 2001 through to January 2010, allowing them to track the evolution of content across edits for each article.
To make the study manageable, they focused on the 1,310 articles about the Roman Empire, a subject whose editorial changes were unlikely to be influenced by current events. The researchers cleaned the data by removing edits made by bots and changes they considered “vandalism,” in which deleted text was later restored in full.
They discovered that the typical edit to a Wikipedia article is shorter than many tweets — just 37 characters, the equivalent of half a sentence — though many were much longer. The researchers refined their analysis by using an algorithm that compared edits made by additions of content versus deletions. (They found that 43% of all changes were pure additions and only 8% were pure deletions; the rest were some combination of the two.)
Even small edit changes early in the life of an article can significantly boost its growth, the study shows. Ultimately, the cumulative effect of edits large and small was quite significant: “Without the cumulative effect, Roman Empire articles would have been only about half as long in 2011 — 45% shorter, to be precise,” says Seiler.
Did the growth effect occur because editors on average were working harder and contributing more to Wikipedia, or because more editors were making contributions? It was the latter, Seiler says.
Seiler and Aaltonen also found that as articles get longer because of the cumulative growth effect, they also get better. Although gauging the quality of an article can be an exercise in subjectivity, Wikipedia quantifies the concept by counting the number of references — footnotes or web links — the article contains. Articles with very few references are known as “stubs” and sit at the bottom of Wikipedia’s rating system.
By using that scale, and also measuring the ratio of references to the length of articles, the researchers found that the growth effect results in higher-quality edits.
The Wikipedia study holds a lesson for anyone hosting a similar platform and hoping to see it grow. “Simply putting it up and hoping that people will contribute won’t work,” Seiler says. “But any action that increases content can trigger further contributions.”
Seiler suggests two strategies to help jumpstart a new wiki-like platform: Prepopulate the wiki with existing content and give incentives to people who might become editors.