THERE are those who edit Wikipedia entries for accuracy – and those who use the online encyclopaedia daily without ever contributing. A new mathematical model says that’s probably as it should be: crowdsourcing a problem works best when a certain subset of the population chooses not to participate.

“In most social undertakings, there is a group that actually joins forces and works,” says Zoran Levnajic at the University of Ljubljana, Slovenia. “And there is a group of free-riders that typically benefits from work being done, without contributing much.”

Levnajic and his colleagues simulated this scenario. Digital people in a virtual population each had a randomly assigned tendency to collaborate on a problem or “freeload” – working alone and not sharing their findings. The team ran simulations to see whether there was an optimum crowdsource size for problem-solving.

It turned out there was – and surprisingly, the most effective crowd was not the largest possible. In fact, the simulated society was at its problem-solving best when just half the population worked together.

Smaller crowds contained too few willing collaborators with contrasting but complementary perspectives to effectively solve a problem. But if the researchers ran simulations with larger crowds, the freeloaders it contained naturally “defected” to working alone – knowing that they could benefit from any solutions the crowd reached, while also potentially reaping huge benefits if they could solve the problem without sharing the result (

But does that happen in reality?

“Crowdsourcing is interesting precisely because humans are not simple,” says Krzysztof Gajos of the University of Harvard.

Consequently, Levnajic’s team will soon study the behaviour of real-world crowds.

By Sarah Scoles

Magazine issue 3039 published 19 September 2015

This entry passed through the Full-Text RSS service – if this is your content and you’re reading it on someone else’s site, please read the FAQ at

Related Posts

Facebook Comments

Return to Top ▲Return to Top ▲