Christie Bahlai: Passive crowdsourcing
June 8, 2016
Christie Bahlai is a research associate in the Department of Entomology in the College of Agriculture and Natural Resources. An ecologist with research interests primarily involving the phenology and population ecology of insects, she is a member of the Landis Lab. Read her latest paper, Predicting plant attractiveness to pollinators with passive crowdsourcing.1
A while back, my colleague, Doug Landis, was searching the web for pictures of flowers for a project about native plants, and noticed that some pictures of flowers he looked at frequently captured insect visits. He got to wondering — do the bees we occasionally observe in this sort of photo have…meaning?
He asked me what I thought of his observation — and could we test it. Are flowers that are photographed more frequently with insects, indeed, more attractive to insects? This idea got me pretty excited.
If you’ve been following me for long, you know I delight in finding data and patterns in places we don’t normally think to look. It’s kind of my “thing.” And through the course of everyday activities, humans passively collect data about the world around them. It seems to reason that common leisure activities — like photographing and sharing pictures of flowers — could potentially capture ecological phenomena — in this case, the visitation rates of pollinators to flowers of different species.
Just a human engaging in a normal human leisure activity.
So we developed a method to test our hypothesis. Using a technique we termed “passive crowdsourcing,” we searched Google Images for pictures of blooms of 43 common flowering plants that are native to Michigan and identified insects that were visible visiting the flowers in the photos. We then compared these observations to visitation rates observed in controlled experimental trials using these same plants.
We found that we could predict how often a flower was visited by wild bees by the number of visits we observed in the internet images, although relationships were less clear for honey bees and bee mimicking flies. Patterns were strongest for flowers that bloom in late summer, when more bees tend to be around in our area.
We’re pretty excited to see how passive crowdsourcing can be applied in the future. This method could be used by scientists to make predictions about other ecological phenomena that may be documented by human use of the web. Essentially anything that people tend to photo-document with any frequency could be capturing data, and could potentially help unlock the scientific mysteries of the future!2
1. And it’s all completely open access, because that’s how I roll. Our raw data and code is available here.
2. If anyone asks you to put down your phone and stop taking pictures of everything, you can gently explain how you’re advancing science.
Reprinted with permission from Bahlai’s blog, “Practical Data Management for Bug Counters”
Photo by G.L. Kohuth