Tallie Armstrong - AI’s Impact on Forensic Evaluations

I have a question that's even more basic in some ways, which is like, why would an examinee want to use AI when preparing for or participating in a forensic evaluation? Like what's in it for them? How could it help them?

That is really going to probably depend on the examinee and on the nature of the question.

As we know, not all examinees are created equal in terms of what their motivations are and how high the stakes might be for their particular case.

And not all forensic questions carry the same level of gravity.

So if you have somebody out in the community, you know, for example, who is up for a child custody or parental fitness evaluation, like I would consider that a high stakes evaluation process because they don't want their kids taken away.

And they certainly don't want to be perceived as a bad parent.

So that is one area where I think it would be very easy for examinees to start querying AI about how to look like a good parent.

And certainly the test runs that I did when I fiddled with Anthropic and ChatGPT suggested that AI was ready to answer those questions.

Not saying that they did a great job, but it definitely gave me enough information that as an evaluator, I was like, I got to start asking these questions.

I can't ignore that.

And I need to be really explicit when I do any kind of virtual interview about what can or cannot be a part of that process for those parents.

Any kind of like sexual violence risk, that's going to be a high stakes scenario.

And the SVP world is really interesting because it's not just about how they might use AI to kind of push the evaluation in a certain direction, but also like, were they using AI as part of the offense related conduct? Right.

AI is very well known at this point for creating pornographic like or just straight up pornographic images, depending on what kind of platform that you use.

And so the legal system is now starting to cotton on to how can they kind of prosecute when there's AI generated pornographic content.

Right.

Because, of course, anyone who is caught with that kind of material is going to say, well, I didn't actually hurt anyone because that's all fake.

But that there are a few documented cases where people have created AI images that look very much like someone in real life.

And and so then that person is targeted.

And then also there have been just the proliferation of AI defakes, right, where someone takes someone's actual face and puts it on a body.

And that's pornographic content.

And that has been extremely damaging for people who have been caught in the crosshairs.

So as evaluators, right, those are two areas right away that I think we could really see some need to be more thoughtful around that.

You're probably going to see it less when it comes to like competency, unless it's a high stakes competency case because the charges are particularly severe and the person is for whatever reason out on pretrial bond or they're in a low level jail setting where they may still have access to the internet in some capacity.

We know that staff can sometimes pass their phones on to people or their internet connections or security systems are not particularly strong.