Well, while I'll actually agree with the objections about what "lie" means, I think I can see where Eyron is coming from with that question. It basically boils down to how much you can trust a statement, just because it passed a "lie detector" test while being said.
I think the most concise way to answer that is that a "lie detector" is not the same as a "truth detector." Even if lie detectors actually worked (and, as said before, they don't), something not being a conscious and intentional lie, doesn't make it a truth.
Probably the best example are false memories. There are a ton of people who, thanks to that fad back then, were helped by therapists to essentially create memories of stuff that never actually happened. It's not a "lie", it's an actual memory, indistinguishable from actual real memories.
If someone got convinced that way that their parents mollested them while they were little kids, you can hook them up to a polygraph and it will come out as "not lying". Because it genuinely isn't. They're telling you, honest to <insert deity>, what they're remembering. It's exactly as they're remembering it, nothing changed or distorted. But the memory itself is of something which never happened, and isn't as much "recovered" as it didn't even exist in their head until, say, last year.
And false memories can be as easy to plant as asking a loaded question. E.g., "how fast was the suspect going when he passed the stop sign?" can actually make people remember a stop sign where there was none.
But even normal memories fade and distort over time. Even people explicitly asked to remember a scene, got it increasingly wrong, the more years had passed. You start remembering a plaid shirt, when it was a t-shirt, and stuff like that.
Again, hook the guy even to a hypothetical magic-based lie detector that actually works, and it will come out as "not a lie". Because it's not. It's remembering it wrong.
People's self assessments are even more interesting, since I see that bit about deviants saying they're not deviants.
It's not even actually about morals or ethics. It's about self-image. Generally, any self-assessments people give, are not about their actual self, they're about an idealized image of what they think they should be. Some examples stuck in my head from an older anthropology book include:
- a tribe traditionally defined themselves as hunters and warriors, and their whole culture was skewed towards valuing being a hunter and a warrior. So on a poll the majority of men defined themselves as hunters and warriors. But to someone who took the time to live among them, it quickly became clear that their actual way of life had shifted towards agriculture over the years. Most of them didn't even have a weapon, nor had gone hunting in years.
- at one point, the meat prices went up, and people in a neighbourhood were polled about their meat consumption. Most basically said, "screw this, I'm eating less meat until the prices come back to normal. That'll show 'em." But checking the sales data from shop, as well as sifting through their garbage for packaging (no, seriously), showed that their meat consumption had actually gone _up_.
- a community answered in a poll that they're tightly knit, they help each other work the fields all the time, they help each other build a barn, etc. You get the idea. But taking the time to dig deeper revealed that actually helping each other was very rare, and the last time anyone helped build a barn was decades in the past.
Those people weren't consciously lying, though. Self image can be distorted like that. They viewed themselves not as who they are, but as an idealized and more acceptable version of what they _should_ be.
Again, hooking any of those to a polygraph would be futile.
