The selection process is also about relationship building with a potential employee who equally is testing out the organisation for fit. I know very few desperate people who would endure a polygraph test to get a job. Otherwise the organisation will be saying "We don't trust you" in bold letters to a prospective employee. That is no way to start a relationship. Polygraph in selection represents selection gone too far. It does not matter whether they are valid or not.

Polygraphs for the most part are unreliable. They can be used to "trick" someone into giving up information, but they are not able to accurately "detect" if someone is lying. There is an excellent Pen And Teller BS episode on this that talks about how polygraphs work and how to beat them. Link to Episode

This is an intriguing question. Are polygraphs even reliable? I'm going to throw that whole can of worms aside and just assume that they are at least somewhat reliable. So then I'll throw out the question:

Why would we even need polygraphs for selection? If you take the time to validate your selection tests, what information can a polygraph give you over and about that?

Another question: What would someone lie about during an interview that you couldn't confirm some other way? Are they lying about references? Check them. Are they lying about their skills? Give them a skills-based test. Are they pretending to be a "nice guy" when they are really the office jerk? Well, that's harder to confirm, but there are still things you can do to weed out the jerks before they destroy your organization.

To me, the idea of using polygraphs for selection just seems lazy. I know that sounds a little harsh but I picture an ignorant hiring manager saying, "Cool! Let's just plug em all into a polygraph and that'll give us the answers we need!"

There are no shortcuts to designing a valid selection system.