In late July I led a week-long course about crowdsourcing and data visualization at the Digital Humanities Oxford Summer School. I taught the crowdsourcing part, while my friend and collaborator, Sarah, from Google, lead the data visualization part. We had six participants from fields as diverse as history, archeology, botany and literature, to museum and library curation. Everyone brought a small batch of images, and used the new Zooniverse Project Builder (“Panoptes”) to create their own projects. We asked participants what were their most pressing research questions? If the dataset were larger, why would crowdsourcing be an appropriate methodology, instead of doing the tasks themselves? What would interest the crowd most? What string of questions or tasks might render the best data to work with later in the week?
Within two days everyone had a project up and running. We experienced some teething problems along the way (Panoptes is still in active development) but we got there in the end! Everyone’s project looked swish, if you ask me.
Participants had to ‘sell’ their projects in person and on social media to attract a crowd. The rates of participation were pretty impressive for a 24-hour sprint. Several hundred classifications were contributed, which gave each project owner enough data to work with.
But of course, a good looking website and good participation rates do not equate to easy-to-use or even good data! Several of us found that overly complex marking tasks rendered very convoluted data and clearly lost people’s attention. After working at the Zooniverse for over a year I knew this by rote, but I’d never really had the experience of setting up a workflow and seeing what came out in such a tangible way.
Despite the variable data, everyone was able to do something interesting with their results. The archeologist working on pottery shards investigated whether there was a correlation between clay color and decoration. Clay is regional, but are decorative fashions regional or do they travel? He found, to his surprise, that they were widespread.
In the end, everyone agreed that they would create simpler projects next time around. Our urge to catalogue and describe everything about an object—a natural result of our training in the humanities and GLAM sectors—has to be reined in when designing a crowdsourcing project. On the other hand, our ability to tell stories, and this particular group’s willingness to get to grips with quantitative results, points to a future where humanities specialists use crowdsourcing and quantitative methods to open up their research in new and exciting ways.
-Victoria, humanities project lead