I can think of two concepts of lie detection. In the first, the statement is compared with objective truth (i.e. Omega says, "Contrary to the Senator's assertion, this tax credit will not create jobs"). In the second, the statement is compared with the contents of the speaker's mind (i.e. Omega says, "The Senator does not believe that this tax credit will create jobs").
The first type of lie detection would be really awesome, but unlikely to be developed based on physiological study because (to paraphrase from X-Files) the truth is not in there.
The second type probably would not be useful in politics because politics is the mindkiller and I predict that most politicians believe the fundamentals of the principles they assert (based on motivated cognition, to some extent). That said, a truly reliable lie detector would be great in litigation. No more he said, she said issues. Of course, there is still the risk that the witness honestly believes some false facts.
You could ask politicians how much they've researched various questions and (in detail) how thoroughly they've considered alternative possibilities.
I have just received a survey questionnaire regarding future directions in EU (European Union) research funding, and thought it would be interesting to see how LessWrong would answer the main question:
Imagine that EU funding is available for one ambitious, visionary project extending beyond 2020.