Research highlight

Network science: Noisy bots aid collaborative problem solving


May 18, 2017

Autonomous ‘bots’ programmed to generate random ‘noise’ can help groups of people achieve a common goal, a Nature paper reveals. The study suggests that adding noisy bots to strategic positions within human networks could help to address a diverse range of problems, such as solving quantum problems and cataloguing archaeological or astronomical images.

Collective action towards a common goal, even if everyone’s interests are aligned, faces a coordination problem: an individual’s attempt to reach a solution that is best for them may not be optimal for the group as a whole. Hirokazu Shirado and Nicholas Christakis modelled this situation by inviting groups of people to solve a networked colour coordination problem. Presented with a network of 20 nodes and three potential colours, the collective goal was to make every node a different colour from its neighbours, but participants could see the colour of only their node and its immediate neighbours. When bots, programmed to exhibit small levels of random noise, were introduced to central locations in the game, collective performance of the groups increased and people took less time to solve the problem.

Noise, or meaningless information in a process, is often regarded as a cause of trouble. Here, however, the effect of behavioural noise was like seeding the network with actors who already knew how to solve the problem. Intriguingly, the bots worked not only by making the task of humans to whom they were connected easier, but also by influencing the way that humans interacted with each other. Cascades of benefit were created and the effect occurred even when people knew they were interacting with bots.

doi: 10.1038/nature22332

Return to research highlights

PrivacyMark System