The short answer is we lie when the truth is a problem. When what we know to be the case conflicts with our communication goals, we may to deceive. Deception can be understood as problem solving. There is a whole chapter in my book Duped on the issue, or you might see:
http://dx.doi.org/10.1080/17475759.2015.1137079

That is a little out of my expertise. I could be, but it might also be something they mature of out of. In general, teens lie about twice as often as college students, and college students lie more than older adults. See https://doi.org/10.1080/08824096.2013.806254 also https://doi.org/10.1016/j.actpsy.2015.06.007

I think as most people use the words, lying is knowing and intentional, while confabulation is an "honest" memory error

Most people, especially most experts say no. Most definitions of lying require knowing what you say is false to intentionally mislead. But not everyone agrees; for example, see Robert Trivers book The Folly of Fools. My book Duped has a whole chapter on defining lying and deception and some of the difficulties and disagreements involved.

Good question, and people disagree on this. If a lie is a knowing and intentional effort to mislead someone by communicating information known to be false, then it's either a lie or not. This said, this can be done with a bit of false mixed in with lots of truth or all-false or anywhere in between. But, it's not the % false content at issue but the use of false information to mislead that makes a lie a lie.

No, but lying can be used to persuade. So can real facts and sound logical argument. In my view, the best persuasion is honest and seeks to change others minds with sound evidence and reasoning.

It surely plays some role, but may vary from person-to-person. Generally people who see human relations in terms of status and hierarchy are more likely to believe something because an authority says it. People higher on the trait openness are less susceptible.

I have an idea I call the BQ (believability quotient ) that I cover in chapter 13 of my book Duped. Likability is a big part of it, but also other things like communication fluency, composure, a consistent demeanor etc. Warmth, intelligence, and likability can all make people more credible and believable.

People can improve their ability to detect lies. I cover 5 ways in chapter 14 of my book Duped. Generally people over-estimate their lie detection ability. Generally, the more confident we are, the less like we are to recognize an error.

Good question. Hard question. People disagree about this. Here is what i think. Pathological liars lie more than others and without apparent reason and with little regard for getting caught. Pathological liars will lie even when the truth would work better for them.

The average people tells between 1 and 2 lies per day, but that is misleading because lying is not normally distributed and quite skewed. Maybe 75% of people lie less than average and most lies are told by a few prolific liars. The average is 1-2 but a majority of people tell 0-1. See https://doi.org/10.1111/j.1468-2958.2009.01366.x There is also a whole chapter on this in my book Duped.

This is a little out of my expertise, but a polygraph measures arousal, not lying directly. People who sincerely believe might not react to questions with arousal.

I'm not sure of lying evolved directly, or if was a byproduct of evolving the ability to communicate and to have theory mind.

Easy call. Believe everything. Only being able to learn from first hand experience would be awful.

We can come to believe our own lies. Lying not only erodes other's trust in us, it also erodes the liars trust in others.

Almost everyone is gullible, although probably some more than others. In my truth-default theory, the idea is that we believe unless suspicion is triggered. Some people have more sensitive triggers than others. If someone believe really implausible things and does not catch on to satire, than might be a sign.

That is currently a matter of debate and depend on how high the bar is for "reliable." If the standard is not zero validity, then vocal pitch, pupil dilatation, and the number of details might have some value on average over large numbers of people. If the standard is useful in correctly classifying some specific communication as truth or lie, then no probably not. My work shows that cues-based lie detection makes people worse lie detectors because cues vary from liar to liar and honest people do them too.