As a software engineer, I often need to elicit information from others: Requirements, technical specs, how something failed, etc.
Sometimes it's easy. More often it is not. Even the most careful, cooperative conversation can be misleading. The fact that the Reid Technique is used by law enforcement to collect such critical information is scary.
The best confessions are the ones that reveal "guilty information" - information that is to available to people not involved in the crime.
The Science Mag story talks about tracking down a potential problem with a computer crash. I have certainly been there. It is possible for some people to accurately reproduce exactly what they have mistakenly done (assuming they ever discovered it), but it's not common. In the best of cases, the person can, through experimentation, reproduce the problem. But failing that, the best I can do is understand and capture exactly what they are saying and know that in all probability, what they are saying is basically substantive - but from the point of view of someone that was caught off guard by the situation.
I got to see first hand a pro tester and developer duke it out in a friendly way.
The tester found a race condition when keys were hit in a certain order. The developer confidently fixed it but the tester typed a little faster and it occurred again.
The last battle before it was finally fixed.
The developer brought in his code and the tester was typing furiously think ” Flight of the Bumblebee “ and after a few minutes boom it broke again.
A day later and a new design fixed it once and for all never to reappear.
No customer was harmed that day but the tester and and developer found great reasons to really respect each others profession.
The tester moved on to become a manager of managers and the developer moved on to other software ventures.