The academic’s favourite question. Why am I doing this? What is the point? Who cares? I love research and learning new things every day. Keeping my mind active. But I could be doing this outside of academia without having to justify the learning.
Fast forward to the Today programme on Radio 4 at 8.50am yesterday. I am almost stopped in my tracks as I approached my office and words such as “galaxies”, “algorithm”, “patterns”, “volunteers” hit my ears. Researchers at the University of Hertfordshire have trained a computer algorithm (full paper here, pretty pictures here) to classify new galaxies at very high speeds. Karen Masters of the University of Portsmouth was fortunately on hand to point out that volunteers have been doing this successfully on Galaxy Zoo for almost eight years now. In 18 months, in fact, 150,000 volunteers identified 50 million galaxies in images from all the major astronomical surveys, exploiting their ability to identify/recognise these patterns. The algorithm groups parts of an image that have similar properties in much the same way. So are these volunteers now redundant?
As Karen articulated, not necessarily. I think it does usefully remind us, however, to be mindful of, and pay attention to, the volunteers’ role within the process of knowledge construction that underpins Citizen Science projects. The number of galaxies that future astronomical surveys contain will be, well, astronomical. There remains potential for man and machine to collaborate, whereby machines classify the easy (often boring) features and humans apply their judgment to the more interesting, weird patterns in these images. Considering the petabytes these new surveys will reach, if only 1% of a billion objects is weird then many, many images will still contain patterns no algorithm could discern. We could reasonably envisage an effective task flow consisting of an algorithm that reports to a human what it finds, who then interprets it to grow our knowledge of the astrophysics of galaxies or stars. The iMars project, however, offers an opportunity to explore the interplay between man and machine in more detail, so that the skills and expertise of man/machine are optimised throughout. 1.4 million volunteers are registered on the Zooniverse and it’s incumbent upon us to respect the time that they donate. Finally, as humans, individual task performance cannot be expected to be consistent so we have to consider how they experience and carry out tasks over extended time periods; what is best for science may not be best for the volunteer. Expect to hear more from me on this (and eventually some pictures!) as the project progresses.
Longer term, what is especially exciting is that, these algorithms do not care what they look at; they could even applied to medical, Earth observation and security imagery. In medical imagery, for example, they could diagnose certain diseases much earlier than currently possible by identifying structures in images that a human could potentially miss. I’m looking forward to ZooCon this weekend where I will meet the developers and volunteers who are driving all of these innovations.