Throughout the three panels and the keynote, a key challenge that came up time and again was data sets. Where do you find data sets likely to be of interest to students? What kinds work best? What do you need to do to them to get them in a form with which students can productively work? How much choice should students have about which data sets with which they will work?