How it works is that we receive what is called a "dataset" from the Allen Institute for Brain Science. It is akin to an MRI of a single neuron. Players then trace along all parts of the neuron (very carefully following the lines). And what we end up with is a 3D reconstruction of that neuron that we pass along to neuroscientists to study. The human visual system is far better at evaluating these types of images than a computer and it is too time-consuming for neuroscientists to make much headway on alone. And so enters Mozak! We are working hard to get the same level of accuracy as expert reconstructions but much, much quicker. Novice citizen neuroscientists are already out-performing computers. With minimal oversight, citizens generate reconstructions that are 70-90% complete. The most effective, computer-generated reconstructions are frequently only 10-20% complete. The ultimate goal is to help better understand the brain and the diseases that do it harm. It's very exciting stuff. Feel free to check out the game at www.mozak.science if you want to see it in action. We also keep in touch with our community through social media outlets like Facebook and Twitter.

Those wanting to play and contribute can head to www.mozak.science and sign up for an account. We are also active on Facebook and Twitter. Tracing a neuron basically looks like very fine, careful tracing the lines of images of a neuron (these are called datasets - they are akin to an MRI of a neuron). The human visual system is uniquely equipped to take on this task, more so than a computerized system. Players' traces are then checked against other players' traces, in a process called "consensus". Running everyone's trace through this consensus process gives us 3D reconstructions of these neurons which we forward on to neuroscientists to classify and study. Mozak has allowed scientists at Allen Institute for Brain Science to switch away from their state-of-the-art tools to novice-oriented tools and observe a significant speedup in their work. With the switch to Mozak, there was a 3.6 fold increase in the number of full neuron reconstructions completed over the course of one year. It's all very exciting and we are thrilled with the contributions of our community!

The project emerged our of specific need by neuroscientists to reconstruct neuron structures, and not having enough human expertise.  The computational methods also didn't help because on these datasets they can get to only 20% of completeness.  Our lab has worked on Foldit for last 10 years, and it was clear to us that we can create an even better rapid expertise development environment in neuroscience by leveraging what we have learned about getting people engaged in science, acquiring skills quickly.  The biggest important novelty in my mind, is the near-real-time interaction between scientists and citizens that is contextualized within the actual neuron, so that know-how can be transferred rapidly while increasing engagement.   The second key insight relates to use of continuously computed consensus to promote both creative exploration, and agreement on correct reconstructions by people.  This enabled the community to advance in productive ways that provide both creative exploration and robust verification, even though the problem does not have a well-defined metric or score for a really good reconstruction.  This approach can be applied to many other scientific endeavors where measures of success are unknown.