A relevant quantitative prediction would be of this form: When participants are randomly divided into two experimental conditions, "donation" and "no-donation"

A relevant quantitative prediction would be of this form: When participants are randomly divided into two experimental conditions, "donation" and "no-donation", participants in the donation condition donate significantly more than no-donation participants. The amount of donation is the dependent variable - it's the way you measure, quantitatively, that the experimental manipulation caused the participants to behave differently. EDIT: I forgot the prediction part. So, I predict that this "donation effect" generalizes. If I ran a charity and I wanted to increase donations, then I might predict that if I gave something to potential donors, they would donate more when they were given the chance. I'd measure the outcome in terms of dollars donated.