Hi and thank you for your interest in participating in our revert assessment survey! (Start survey)
You will be given several pairs of revision text Diffs of Wikipedia articles to assess if one edit is a revert of another. We do this in the EU research project RENDER (together with Wikimedia Germany) to model user revert behavior. I.e. we would like to know what the Wikipedia community sees as a "revert".
What will the data be used for? The results help us to automatically identify reverts and build models and social network graphs of how editors interact. This knowledge will be used in RENDER to identify articles which are prone to show harmful editor behavior. Together with other techniques such as fact coverage detection, we hope to aid the editors of Wikipedia in finding low quality articles by developing respective tools for this task.
How to 
A short how-to is given in the survey. Start Survey: http://people.aifb.kit.edu/fde/limesurvey/index.php?sid=34839&lang=en