The Impact of the Filter Bubble - A Simulation Based Framework for Measuring Personalisation Macro Effects in Online Communities

The term filter bubble has been coined to describe the situation of online users which—due to filtering algorithms—live in a personalised information universe biased towards their own interests. In this paper we use an agent-based simulation framework to measure the actual risk and impact of filter bubble effects occurring in online communities due to content or author based personalisation algorithms. Observing the strength of filter bubble effects allows for opposing the benefits to the risks of personalisation. In our simulation we observed, that filter bubble effects occur as soon as users indicate preferences towards certain topics. We also saw, that well connected users are affected much stronger than average or poorly connected users. Finally, our experimental setting indicated that the employed personalisation algorithm based on content features seems to bear a lower risk of filter bubble effects than one performing personalisation based on authors.