Nature has long inspired inventors and engineers. The solutions to evolutionary pressures that natural selection has delivered have inspired some remarkable inventions. Ants’ behaviour has motivated the design of efficient computer networks, burdock burrs have influenced the invention of Velcro, and Gecko’s feet have stimulated the design of non-sticky adhesives.

We can call these inventions ‘models’, and they are designed to replicate aspects of the real world. In the cases above, researchers are interested in designing products that work, even if the design of the final product deviates from what the natural world looks or behaves like. In other cases, we are interested in creating exact replicas of nature. Take Siri for example. Apple have designed her responses to be as human as possible, in order to replicate real human behaviour. In these cases, how can we assess whether the model (i.e. Siri) has adequately captured the essence of human behaviour?

Alan Turing proposed a solution to this problem. He designed a theoretical game where two players were asked questions by an interrogator. If the interrogator could not assess which one of the two players was a machine, and the other a human, then the machine had passed the test and exhibited intelligent behaviour. In essence, this simple test asks whether models can accurately mimic the real world.

We took Turing’s ideas and applied them to collective animal behaviour. We asked whether we could design a model that could accurately capture the collective movements of a fish school. To do this, we designed an online game where players were asked to choose between the movements of dots generated by our simulation model, and the movements of real fish also represented by dots. You can play the game here:

Whilst our model could replicate some of the statistical properties of schooling fish, was it good enough to fool the human interrogators? We found that both members of the public and experts could distinguish between the model and the real fish, which highlights the need for model improvement. This simple test, therefore, is a powerful way of quickly assessing the effectiveness of model fitting. You can read the full paper here, published in Biology Letters.

Our implementation of Turing’s ideas is the first example of using tests like these to improve model fitting. Moreover, our approach also highlights that the public can, and should, play an important and engaging role in science. The public provided critical feedback on our model, which has inspired us to develop new games where they can help our research. On Monday, we launched our new game; ‘Fish In Danger’. In this new game, the player takes the role of a predator and has to eat prey to stay alive. You can play ‘Fish In Danger’ here:

We hope our approach of using online games and communication engages the public in our research, but at the same time, allows you to have fun!