Experiment 2 conclusions
•Invariants
–effectively summarize value data
–support programmer’s own inferences
–lead programmers to think in terms of invariants
–provide serendipitous information
•Additional useful components of Daikon
–trace database (supports queries)
–invariant differencer
[Do not drop into “we”; keep me separate from the programmers, who made these judgments on their own.]

Make it clear that this is all good.  “Invariants had a number of positive uses” or some such.

Leading programmers to think in terms of invariants:  this is a “guerilla approach” to introducing invariants to programmers.

The useful components were used by the programmers in the study.
The useful components are beyond invariant detection.


Use the word “serendipitous”.  That’s the point:  if you know exactly what you are looking for, or you want to verify a specific property, you are unlikely to get as much from Daikon, which merely automates your task rather than providing entirely new information and functionality.  We want to raise the programmer’s suspicions and draw his/her attention to code or facts that might have otherwise been overlooked.  There are things you expect but didn’t write down (and didn’t want to be bothered to write down).