Tuesday, 6 October 2020

Our experience with webswing

Since 2012 we have been developing a java based application for running interactive experiments with human participants in a virtual environment (see www.viappl.org). We initially designed the system to run in a server/client configuration in a lab, with the assumption that studies would run on campus and we would be able to bring participants into a shared space to participate in groups. Of course COVID put a stop to that, and there is no certainty that we will be able to return to our in-person methods in the future either. We were faced with a choice between reconfiguring our software for remote use or stopping our programme of research altogether. 

We had written our client-side as an executable java file to run on the client PC. Although we had planned ahead to ensure that the protocols and libraries would run fine even if dispersed across the internet, we realised there would be few participants willing to run an unknown executable file on their PC and, even if we could find willing volunteers, there would be many technical issues in making sure they had the right Java runtimes installed and so on. These complications would make running adequately sized experiments prohibitively complex, time-consuming and expensive.

At first, we explored the idea of provisioning containerised desktops that participants would log in to remotely. While this would have been possible and may have worked OK, there were multiple technical and social issues that were complex and difficult to resolve. One important barrier being that most people are unfamiliar with remote desktop access and virtual desktops are tricky to deliver to web browsers via htpp.

In our search for a better solution we stumbled across Webswing (webswing.org) and to our great surprise (and relief), it only took an afternoon to get the client-side running as a web-delivered application. This will allow us to easily deploy our client applications to participants via their web-browsers, dramatically simplifying the logistics of running our experiment paradigm via crowdsourcing platforms like MTurk and Prolific Academic. So far, the main technical difficulty has been resizing the graphical user interface. More complex issues (like protocols; library compatability etc.) have been straightforward. Webswing sales & support have been helpful and responsive. We are yet to trial the solution in the field, but it’s looking very promising so far.