Dr. Arthur Lee

Even though we created what is purportedly a lie detector, the validity of it has shortcomings, and it's actually being confounded with other things like financial interest or selfishness, those kind of things.

So this is us saying, if you just focus on improving the accuracy, these are the things you're going to miss.

So if you build a lie detector in the lab, who knows what's going to happen in the real world when there are other things that are confounded with lying.

So then we move on to the third part of the task where, which is, it's not really a task, but essentially we say, we have a way to balance out the signals in the brain such that we can make the confounding signals of selfishness cancel each other out in the brain while retaining most of the predictive power of lie versus non-lie trials.

So this is a bit of a statistical trick that we used to say, we want to build a model

that can predict lie versus no lie in the first task as much as possible while staying true to the constraint, the statistical constraint that you should not be able to distinguish selfish versus non-selfish trials at all.

That actually works pretty well.

You're able to remove this selfishness contrast from the brain while still retaining your ability to detect lie versus non-lie trials.

This is not to say that we now have a lie detector.

But this is to say, look, we created a lie detector, we identified a confounding process, and we were able to subtract the confounding process, so hopefully we now have a more purified version of lying, or a mental construct of lying.

If you have other concerns, other things that can be confounded with lying, you can now construct an experiment, subtract those out, and we can keep playing this game until either one of two things happen.

One is we remove all confounding process and somehow we're left with this pure lying signature, which means that there is a neural signature that is uniquely tied to lying.

Or we end up in the second scenario where if we remove enough confounding processes, we don't have anything left and lying is just an amalgamation of all these processes.

We don't know the answer to that, but I think the main benefit of the study was that we opened the doors to make this validity problem somewhat tractable, solvable with incremental research.

there are certain well-known activities that have a designated brain area that is very easy for you to tell.

Like motor activity when you're moving your arms and hands or when you're seeing stuff, there are like entire courtesies that is dedicated to vision and motor.

So when you see activity there, that's pretty clear that somebody is trying to move or is moving or somebody is seeing.

When you get into more higher level constructs like love, emotions, planning, learning, they inherently recruit multiple areas of the brain.

So now you need to balance multiple signals across the brain to see if you can weigh the evidence.

So there's that difficulty.