In 2016, Sarah Parcak of the University of Alabama at Birmingham won the TED prize for her wish to build a citizen science web site where everyone can contribute to archaeology by analyzing satellite imagery.
Global Explorer is the incarnation of this idea. The initial design is focused not on discovery of new sites, but on the identification of looting. The project is also collaborating with National Geographic, with links to video articles related to the “expedition”.
“The GlobalXplorer° Project seeks to identify and quantify looting and encroachment to sites of archaeological and historical importance.”
The underlying logic is familiar. crowdsourcing science is almost always about finding visual features in imagery, utilizing the human eye and brain where it is still competitive with computer algorithms. Actually, these days this is as much a matter of money and technology, because it is likely that computer systems could be created to do this task. But archaeologists do not have money to do that, so citizen volunteers are the cheaper option.
Image classification is tedious and difficult. (I looked at their tutorial examples and, frankly, I simply cannot see well enough to do it at all. Young eyes are needed.) This means that it is important to have an exciting story to motivate the “crowd”.
GlobalXplorer focuses on detecting looting, which is a simple and compelling story. A similar strategy has been used to try to protect animals from poaching, and other DIY projects are motivated by the desire to track environmental damage or disaster relief. This is effective, since the internet makes it possible for us to do something about looting, far from our homes.
The focus on looting has its drawbacks, of course. The site is carefully designed for forensics rather than science, and it is scarcely a good example of open data. The fact is that satellite surveillance is just as useful to thieves as it is to protectors and scientists, so GlobalXporer has to be very careful, and does not provide the full dataset.
“GlobalXplorer° is designed to find and protect sites — so keeping site location data safe is our highest priority. No map tiles or images will be linked to coordinates or other data that would expose the location of a site — all tiles are assigned a random catalogue ID.”
The project has training materials and they give you a “consensus score” to accumulate how closely your agreement with the rest of the crowd. (This would be “interrater reliability”, one of the key criteria for human content analysis.) This, in turn, helps flag cases where a lot of people agree there is something going on.
The idea is that when evidence of looting is discovered by the volunteers at GlobalXplore, the operators will notify local authorities charged with protecting the material. GlobalXplore gives no examples of this process in action, and I have to wonder whether retroactive reports of past looting have any value at all in preservation.
I don’t want to discourage people from supporting global efforts to protect archaeological sites, and GlobalXplorer may turn out to be a useful educational and public relations tool. But it doesn’t seem to be doing any science, and has yet to be proved to be useful for preservation.