The best and clearest explanation of what tokens ate I know of is the post by Balaji S. Srinivasan and Naval Ravikant I was replying to.

It would be great if our thought-experiment token had a way to incentivize return of results. We engineer a way for tokens to be given in exchange for results. What bugs me is that I see no easy way of evaluting the quality of the results. Need to think about it.