 In this episode we've taken a lot of the cognitive mechanisms that we've learned about and applied them to cases where lives and livelihoods are at risk. So we looked at facilitated communication and saw that expectancy effects are operating. We can use 2x2 contingency tables to help disentangle hits from false alarms and weigh the costs versus the benefits. Now Bill Thompson alluded to something that we're going to talk about next which is expertise in fingerprint identification. Now you and I spend a lot of our time working on this stuff. Contrary to what you see on CSI, it's not computers that match prints and interpret DNA profiles, it's actually a human. That's right and because we're dealing with humans here, that's when things get interesting, right? It's not up to a computer. It's not simply a matter of going to a crime scene dusting for prints, getting the bloody print, putting it into a computer and up pops. Drivers license of the person who committed the crime? Exactly. Yeah, it doesn't work like that. It doesn't work like it does on CSI. It's a lot more noisy than that, right? And yeah, it might work like that in a sense with a computer if it's really high quality fingerprints, right? That's what happens with your iPhone or at the airport often. It can be done without a human. But when you're dealing with bloody crime scene prints, that's when human judgment is involved. And we know that there's, as Bill was saying, there's an enormous amount of ambiguity in those sort of conditions. With the bloody print, often it's this partial print as well so you only see just a little bit of it and you have to match it to this fully rolled 10 print. And we can see in this figure actually, the print on the left is from person A, the print on the far right is also from person A. And they look pretty similar, right? But the print in the middle is from person B, right? And it looks really close to person A. I mean, it doesn't seem like all fingerprints are unique when you're looking at these kind of examples. It's not clear cut. There's a lot of ambiguity. And when you have ambiguity creeping in like that, that's when we're vulnerable to these sort of expectancy effects that we've been talking about. I tell you, the person confessed to the crime. Now analyze these prints and tell me whether they're from the same person or not. You can't look at them with new eyes. You can't look at them as though you don't know that information. It's going to creep in. Or DNA is also vulnerable. Bill Thompson talked about how much ambiguity creeps in here. So when you're looking at these DNA samples, it's not clear that the sample that you have is from this person or not. There's a lot of information that when you're dealing with mixed samples and so on that points in one direction or another, right? And depending on the information that you have in front of you, you can go one way or the other. They're multifaceted in that sense. Or CCTV footage, same thing. You're filming this sort of fuzzy crime that happened and you have to judge whether the person in front of you is in fact that person in this grainy footage. Sitting in the dock there. That's right. So what do you do? Yeah, that's when this sort of stuff really, we see what we expect to see. And that's a perfect example of that. Yeah, so there's also hearing what we expect to hear as well. So there's a case out of New Zealand where a guy called David Bain was accused of murdering his whole family. Now there's a call to the police or the ambulance which is made by David. And he's panting. It's obviously, you know, scared. And the quality of the audio recording across the phone line is pretty terrible. But sound engineers later on said that they heard David say, I shot the prick. Even in this noise of this panting and everything that's going on. It's really hard to listen to. Yeah, but if you are told I shot the prick and then you play that recording just like it's fun to smoke marijuana, you will hear it. I think because there's enough ambiguity there. Now think about what that means for the case. If that information was presented to a jury, what do you do about that? We know that they're likely to be influenced by that information. So again, it might be best to sort of blind them to that and just let them listen to it in the first place. We saw in episode two that Beth Loftus is able to plant entire memories of events that never actually happened. That might not work if she's trying to influence you for what happened yesterday. You were lost in a shopping mall yesterday. That might not work. But think about the ambiguity that just comes from time. When you were eight years old, you were lost in a shopping mall. That's the space in which that she's able to convince you of something that never happened. Now back to forensic expertise. Fingerprint examiners have claimed that they won't be influenced by extraneous information about a case. So if they know that the person committed later confessed to the crime or if they know that their colleague has already said that these two fingerprints match, they've claimed that they won't be biased or influenced by this information. But just as we saw in episode three in Know Thyself, people have very little insight into what's going on inside their heads. We know that they will be influenced by this information in one way or another. In episode four, Intuition and Rationality. This is essentially what I did my PhD in. I was looking at the influence of system one and system two on the judgments of fingerprint examiners. How much do they do really quickly and how much do they have to spend more time and analyzing. In episode five, Learning to Learn, we're trying to use distributed practice and interleaving to turn novice fingerprint examiners into expert fingerprint examiners more quickly. Episode six, the experiment. Now we did kind of a trivial experiment on testing whether people can discriminate wines. But we're using exactly the same tools when we do our research to find out, I think, more important things about fingerprint experts can discriminate between matching and non-matching prints. The list goes on. That's right. Again, we're trying to apply all of the lessons learned in these previous episodes to these more applied topics. The next topic that we're going to deal with is the idea of conspiracy theories. Now, I talked to Steve Lewandowski about exactly this. He's done an enormous amount of work on conspiracy theorists, applying this stuff to climate change deniers. And we can see all of these topics happening here as well. Availability heuristic, confirmation bias. And I talked to Steve about these.