Crowdsourcing – the process of getting ideas, work, or funding from a crowd of people – has become central to modern business and, in the guise of online communities, to market research. It would be easy to assume that the bigger the group, the better the outcome. But new research suggests that this may not be the case.

A team of academics, led by Zoran Levnajic of the Faculty of Information Studies in Slovenia, has created a new mathematical model that says – for some tasks – crowdsourcing a problem works best when a certain subset of the population chooses not to participate.

“Human beings probably have an innate tendency to operate in groups,” the paper explains. “The emerging skill of social problem-solving has allowed our ancestors to face the challenges of gradually increasing complexity. This factor not only determined our evolutionary success, but still shapes our cultural and social behaviour.”

But the paper points out that while the existence of large sub-groups of collaborators guarantees a stable support for the community as a whole, it also creates a “socioecological niche for the free-riders”, who are described as those that benefit from the available social support while “providing little to its development”.

Levnajic and his colleagues tested this hypothesis virtually, by simulating a scenario in which groups of varying sizes within virtual populations were randomly assigned tendencies either to collaborate or to work individually on a problem. The groups were then assigned problems of varying levels of complexity to be solved, and the results compared. A score was also introduced for each player: ‘fitness’, which represents the player’s own benefit in terms of new knowledge acquired from the task.

It was found that for very hard problems, players had no preference on collaborating, regardless of the size of the group. But, for very easy tasks, players in small groups strongly tended to collaborate – in fact, tendency to collaborate increased with task simplicity, while for players in big groups, this tendency decreased.

The team reasoned that when a task is extremely difficult, it is hard to solve it collectively as well as individually, which leads to there being no preference for either problem-solving approach. But, as the tasks become gradually easier, it becomes less beneficial for players in small groups to work individually, since this reduces the size of the ‘collaborating subgroup’, making it so small that little to no fitness – knowledge benefit – is gained from it for anyone involved. So the choice to collaborate is a safer one in terms of guaranteeing at least some fitness for all players.

But in a larger group carrying out an easy task, there is more incentive to choose not to collaborate, as the collaborating subgroup will always be reasonably large and contribute significant fitness. But, if a player chooses to work individually, they will still be guaranteed some knowledge benefit, making it a relatively safe choice to free-ride in this way.

“An individual facing a new challenge in isolation has considerably smaller chances of solving it, despite relying on the available knowledge,” the paper says. “Yet, if a solution under these circumstances is ultimately found, this subject learns much more than when the solution is found collectively.”

It was found that a group size of 50 had the optimal system whereby both collaboration and individualism were equally beneficial for a task of any difficulty. A group size of 100 saw too many players working in isolation in the hope of getting a free ride, so that the group gained less overall.