Designing bots in games with a purpose

The massive amount of time that people spend in online gaming is being increasingly exploited by a particular kind of Serious Games, the Games with a Purpose (GWAP), used to solve complex problems as a byproduct of a collaborative gameplay. The required tasks are solved by exploiting game mechanics that often require the submission of thousands of players' annotations, to achieve a robust estimate of the results. Gathering a consistent playerbase able to solve computational problems at a scale is extremely difficult, due to better entertainment alternatives on the market and the necessity of pairing each player with another one due to the inherent multiplayer nature of this genre. Artificial players (bots) may be introduced when the online platform has not enough human contributors to employ, but their functional requirements and implementation is much different than the one of traditional videogames. In this work we describe the framework and the design choices that have been used to implement a bot in an existing GWAP for fashion garment image segmentation, showing how supervised methods can be applied effectively to emulate human behaviour in the resolution of computational tasks through gameplay actions.