Transcript: Interview with Ryan Scott from Episode 039

This is the interview with Ryan Scott from episode 039 of the You Are Not So Smart Podcast.


scottOur guest in this episode of the You Are Not So Smart Podcast is Ryan Scott, a cognitive psychologist who is adding to a growing body of evidence revealing that our guesses and our confidence in those guesses don’t come from the same place in our minds, and separate still is our conscious awareness of these loops of thought feeding forward and back upon each other.


David: Ryan, before we get into your research, we should probably talk a little bit about some of the things that informed it. And there’s this really great phenomenon that you’ve mentioned several times. Before you started discussing blind insight in your paper, you mentioned another phenomenon that in many ways helps make sense of what you’ve discovered. And that phenomenon is blind insight. Could you sort of describe that phenomenon to listeners who may have never heard of it before?

Ryan: Yes of course. I mean, just to point out, the reason we called it blind insight was to highlight this relationship. Because it is, in a very real sense, the opposite of this other phenomenon called blind insight. So blind insight was identified way back in the 50’s, but what it amounts to is – in certain patients, and this has also been recreated in animals. Deliberately in monkeys. But in human beings, they can have damage to their visual cortex. Which leaves them with an area of their vision which they say, “I can’t see anything there. I am blind in that region of my vision.” So they have no visual experience at all. However, if you were to present stimuli in that area then say, “Right, is this line vertical or horizontal?” I’d say well, it – what are you asking me that for? I’ve told you I can’t see anything there, I’m blind. But – and you say, “Okay well just humor me here and guess. Take a guess. Is it vertical or horizontal?” What you find is, when they’re making those guesses, they’re accurate – way above 80% in some instances. So they have this insight to the – to this stimulus. They’re able to see the stimulus in some sense. But they don’t have any conscious experience of it. And this was termed blind insight, and researches such as Lawrence Weiskrantz did a lot of studies on this. So what that really tells us that – and the reason this has been so influential is that – we can be influenced, our behavior can be influenced by stimuli that we’re not in any way conscious of. So we don’t have insight into having that knowledge, and none the less it influences our behavior and our ability to make decisions.

David: That is of course, it is super fascinating, and it’s also in some ways kind of creepy. Because I think that we – we have an intuition that we are completely aware of the sources of all of our behaviors and whatever – sort of the antecedence to our behaviors and those sources of why we feel the way we feel. What would you say, from your perspective in psychology – how would you rate that intuition?

Ryan: Proved completely incorrect in so many regards. No it’s absolutely the case that it just doesn’t stand up to the scientific evidence. We can show that we can influence people in lots of different ways. And not very strongly in the sort of ways that people worry about in the sense of, “Oh what about subliminal advertising?” and so on. I don’t…those methods can and there’s some recent studies that show that there can be some influence there. But it’s pretty small, it’s pretty trivial. But in many other regards, what we think we’re basing our judgements and decisions on clearly isn’t the case. We can– And that, that can be shown time and time again. A lot of the – a lot of our confidence in the basis of a judgement is post hoc. We seem to be justifying the decisions. There’s been some beautiful research in that area, where through sleight of hand– You present, for example, pictures of a couple of – a couple of, a couple of pictures of different women for example. And you ask someone to, “Tell me, which of these is more attractive?” And they pick which one is more attractive. And then through sleight of hand, the experimenter actually presents that one– Presents as if it were the one they chose. The other one. So then is now presented the one that they didn’t choose, as though it was the one that they had said was the most attractive. And then said, “Now, can you explain why you found this one that more attractive?” And in many, many instances, didn’t notice the switch at all. And go ahead to provide all the good rational explanation for why they found that one more attractive. Clearly, they haven’t even noticed that change, so it’s – that’s a – perhaps a slight tangent on your key point about it, not knowing what the basis of the judgment is. But it really highlights the fact that a lot of the time, this is post hoc. It’s not–

David: Yeah, yeah, yeah. I mean, that is all – some of my favorite research is that kind of stuff. And I love the– Whenever I get a chance to like give a talk, I’ll usually bring up the Nisbett and Wilson stocking study. Where they have everyone rate 4 different stockings as which one they think is the best.

Ryan: Yeah.

David: But they don’t tell people that they’re all the exact same stocking. And people tend to pick the one on the right more often than not. And there’s like a lot of explanation as to why, but what’s great is that no one ever says that’s why they chose that stocking. They come up with all sorts of other explanations for their own feelings toward this stocking. And if it wasn’t for the experiment, if it wasn’t for like the debriefing, they would go the rest of their life not knowing that that is why they did what they did. And of course like the extrapolation of that is – we’re always explaining ourselves to ourselves, and sometimes we’re right, and sometimes we’re wrong and we don’t know when. That is–

Ryan: I man I have – in my own particular area of cognitive psychology, I deal more with more controlled presentations. But another example of this is where we do present subliminal stimuli, very, very briefly. And a very old task in this area is a stem completion, a priming of a stem completion. And all that means is, is you present a – at the start of a word. Say t-a-b. Now obviously that could be a number of different words. It could be table, it could be tablet, – it could be a range of different things. And you present very briefly a word that is- it’s usually a back mast stimulus. Presented for say 30 milliseconds, back mast by a pattern – and all they see is just a flash of a pattern. They can’t see the word, they can’t report on the accuracy of the word. But nonetheless, if you then present the first 3 letters of that, and you ask them to complete it – they’re more likely to complete it with the word that was flashed up.

David: Wow, so–

Ryan: No access to that information, no conscious access, but it does prime them to be completing this.

David: Oh man, that is so fascinating. We’re moving towards your research slowly here, but if – one of the things that plays into what you’re talking about. In a way, one of the ways that’s sort of demonstrated in a lab is this thing called artificial grammar learning. Could you sort of– I’m sure most people have never heard of that before. What is artificial grammar learning?

Ryan: It’s a fancy name, but it’s a very simple paradigm. It’s been extremely useful. Essentially all you do – if I take you through the way this would– A person would experience this. What you say to someone is, you say “Right, this is just a short term memory test. I’m going to present strings of letters, so like for example, X-T-T-V-T-W – something like that. And I want you to try to memorize it for 5 seconds. And then I’m going to give you 10 seconds to type it back in.” Now, you might do this for say 45 of these strings, so just go through this. What they think they’re doing is trying to memorize these strings. Now, of course, unbeknownst to them, there’s actually a systematic set of rules which dictate which letters can follow which other letters. It’s a complex set, so it’s not easy to see that’s the case – but there is some rule in there. You can refer to that rule set as a grammar, if you like, which is where the name comes from. Then what you do, is you say to that same participant, “I’m going to show you–” You tell them at that point, after they’ve done that initial training. You tell them that there were a set of rules in there, even though you didn’t know it. And you say, “I’m going to show you a new set of strings. All of these are new, and none of them are the same as the ones you’ve seen before. But exactly half of them obey the same rules as those that you saw in that memory phase.” Now, in doing that you’ve made them aware that there are rules there, but you haven’t told them what the rules are. And now they go through, and for each of these strings you present you say, “Does it obey the rules, or doesn’t it?” And they just have to say yes or no. And then you might also ask them, “How confident are you? Do you have any confidence at all in that judgement? Or do you really think you’re just picking at chance?” Now systematically in that – using that paradigm for a very long time, we show the people – show knowledge. They usually will score 60, 65% accuracy would be typical in that sort of study. Many instances they’re doing that when they feel they have no confidence at all. So it’s a – it’s what we call the confidence criterion for unconscious knowledge. So it’s if they say – it’s the guessing criteria of our unconscious knowledge. If they say they’re completely guessing, and yet they’re still scoring 60, 65% correct – then we’re saying that knowledge that that you have is unconscious.

David: So this is so insane to me. Because the idea is that – basically is that you are – you’re figuring out the grammar, you’re figuring out the puzzle, the– You’re sort of sorting out the mechanics of a problem, but you have no awareness that you are doing so, and you’re like–

Ryan: That’s right.

David: And you’re like, and you’re completing the task – you’re doing it correctly before you’re cognitively aware of how you’re completing it correctly. Is that sort of how it works?

Ryan: Yeah, absolutely right. The – over time, if you gave them enough training, or even in some instances–If you give feedback on that test phase, which you don’t normally – then people will gradually start to gain more conscious access. And some people will – will some of the time say “Oh, I’m pretty sure about this one actually.” And sure enough they are more likely to be right when they do have that confidence. But nonetheless, on others they think, “I have no idea, I’m just going to pick no because I’m just guessing.” And nonetheless still show good, above chance accuracy on that. This might be a little confusing describing it this way because in actual fact, in terms of blind insight, we’re interested in those participants who weren’t able to–

David: Right, right- –

Ryan: So it’s the reverse–

David: So you took this, and you sort of twisted it up. And I’m sure we’ve – many of us have experienced this before. Often times we’ve, you can– Sometimes you go with your gut, and you’re right, and you don’t know how it could possibly be so. And you’re slowly getting to the answers before you’re aware of how you’re doing it. But your research took this sort of paradigm and fiddled with it a little bit and got something new out of it. So, let’s go ahead and get there. Take us through your research and what you discovered.

Ryan: Okay, well, really the point to start is that with this sort of typical finding and artificial grammar learning. As with blind insight, that we discussed – what you’re seeing is people are able to make accurate judgements, despite not knowing that they can do so. So they lack what we refer to as meta cognition, so knowledge about the knowledge you have. Just to put that in to everyday sentence, if I said to you, “What’s the capitol of France?” You’d say, “Well, it’s Paris.” And you’ll be pretty certain that you’re right. Some more obscure country, in Central Africa perhaps, you might take a stab at it, but have less confidence. The degree of insight that you have to whether you’re right or wrong is your meta cognition. So that’s what we’re talking about here. So in this, in this – in blind sight, what you see is they’re accurate, They can say that the line’s vertical or horizontal. But they do not have any insight at all into their ability to that. They think they’re just guessing. Exactly the same sort of thing typical in artificial grammar performance. They perform – way above chance, but they think they’re just picking at chance. So what we wanted to know, however, was whether you can get the opposite dissociation. So what you’re seeing there is the dissociation between their performance, their decision accuracy, and their meta cognitive insight, so their meta cognitive accuracy. Just to explain how we measure that accuracy is the correlation between your confidence and your accuracy. So if you’re always more confident when you’re right then when you are wrong, you’ve got a strong correlation there that your much more confident when right than wrong. Then you’ve got good, solid meta cognition. Yeah, you’ve got that insight. If you had no correlation between those things, so that when you were right your confidence was no greater than when you were wrong. Then you’d say that there’s no meta cognitive insight there. Now, what we wanted to know is whether you can get the other dissociation. So that you could get meta cognitive insight that is accurate. So you could be more confident in your right responses than your wrong responses. Despite being at chance in those original decisions. So, creating a situation where people perform at chance in the actual original decision – so they can’t get it right more often than wrong. And yet, when you ask them for their confidence about their – a response that they made that was correct, by chance. They happen to be more confident than a response that they got wrong by chance. That’s what we wanted to know. And the reason for wanting to look at that is it challenges some fundamental theories about meta cognition. So to take you through the actual experiment, what we did is – we had a very large set of data from people who had done these artificial grammar learning studies. So a large number of participants, so 450 participants had done these studies. What we wanted to do was take those participants who performed at chance on this – which of course you’re always going to find some that aren’t performing above chance. And then look to see whether their confidence was related to their accuracy – even though they were performing at chance. Now you can’t just go ahead and do that selection. Because regression to the mean is a process which would mean – that if you did that, you’ve got a biased sample. And that it’s – it’ll always give you that result. So we had to do it – something a little bit more complicated than that. But essentially what it meant was we – we selected participants who performed at chance in the first 3 quarters of their responses. And then we did the analysis on the last quarter, which overcomes those biasing issues. And so it’s a detail, but it’s quite an important one. So what we’ve got then is a sample of participants who’s responses in that last quarter – were also at chance. So we know they’re performing at chance, and now we could actually look and measure their meta cognitive insight. Look at that correlation. And sure enough, what we found was their meta cognitive insight there was – was not only significant, it wasn’t significantly different to those participants who were performing above chance on the decision. So you’ve got 2 groups of people now. You’ve got those who perform above chance on the grammaticality judgement or the – deciding whether it obeys the rules or not. And those who are performing at chance, and they were looking at the correlation between confidence and accuracy for those 2 groups. And we found that they didn’t differ. So those people who didn’t – weren’t able to correctly classify the grammar strings as being – as obeying these rules or not. So they’re at chance on that, and they may have well been just literally flicking a coin and picking. None of this. When they, by chance got it right, they had a higher confidence rating than when they at chance got it wrong. And that, really fundamentally challenges both this – these models of meta cognition and our everyday intuitions.

David: So what do you – I mean, is there something malfunctioning in this person’s mind, that they’re not sort of connecting between the intuitive sense they got, that they have in their gut – and then they’re following through with the answer. Or is there some other speculation out there that sort of explains what’s going on? What do you think’s happening there?

Ryan: I think it’s – in some ways, there is – this is speculation. I would definitely say it’s speculation. I think there is a reasonably logical explanation for it. And that is that we know that when we make responses to questions framed in different ways, we are able to draw on different information. We don’t realize we’re doing it, but that’s the case. So what I suspect is happening here is, when people are asked, “Does this string of letters obey the rules?” They adopt a very analytical process – or some of them will. And in some instances, they’ll adopt an analytical process that is flawed. So basically, their following a rule perhaps that they’ve devised – and it’s wrong. So as a result they score it “chance” in terms of that decision. However, then when you say, “How confident are you that you got that right or wrong?” They take a more relaxed, more intuitive approach to that. And say, “Oh, how do I feel?” They’re not, they’re no longer engaging that sort of analytical process. And engaging that intuition, they’re potentially drawing on information that is in addition to what they were using for that analytical process. As I’m trying to give you a sort of an everyday example of that, the best of the – able to come up with. Is, if you imagine you were trying to choose the right route on the underground. So you’re there, you’re waiting at the tube station. And you’re saying, “Right, which route should I take to get from here to there in the quickest time?” And you pick a route. You pick what you think is going to be the shortest route, just based on the number of stops it involves. So you go from station A to station B, and there’s 3 stops on the way. And then station B to station C, and there’s only 2 stops by one route – but there’s 7 stops by the other. So you’re going to go the one with the shortest number number of stops. That makes sense. It’s a – an analytical process you’ve gone through to pick which do you think’s going to be shortest. Then you hop on the train, and the moment you hop on it, you think, “Dammit. I’ve chosen the wrong one. I should’ve gone the other route.” Now how could that happen? How could that logically occur? Well it could be that there’s some forgotten knowledge that you have. Or semi forgotten knowledge that – about, having breakdowns on that line. So that line has suffered sort of more breakdowns when you’ve been travelling than the other line. And that has influenced your confidence. Your feeling of confidence suddenly says, “Hang on, this isn’t right.” But you didn’t use that information, it wasn’t made available to the original decision.

David: That is, that is – first of all – absolutely fantastic way to put that into perspective. That it really makes it click into my head better than it was before. Because that’s great, and also, wow it’s– I mean, you mentioned in the paper that we have– This is hard to grasp in some ways, because we have an intuitive sense – that when we, that we sort of draw upon the same pool of knowledge to both answer a question, and then to determine whether or not that question is correct. And the way I was making sense of it was that, if we had like – it’s like checking– If you just recently checked your bank account, and then someone asks you how much money you have in the bank. You can answer that question, and then feel confident that you’re answering the question correctly. And it feels like the confidence and knowledge both spring from the same source – the memory of the bank statement. Like the actual physical numbers that you are seeing pop in your mind. But it seems that your research suggests that our confidence, and the accuracy of our responses – and the actual accuracy itself are generated separately from different sources.

Ryan: Or potentially in some cases. I would put that caveat there. Because this was a subset of the participants of course.

David: Right, right, right.

Ryan: For most conditions, and those decisions quite as you would expect – our confidence correlates very strongly with our accuracy. But what we’ve shown, that because that doesn’t necessarily have to be the case – the fact that it fails even in a single instance. Tells us that our brains aren’t wired such that it is a necessary arrangement. And this is, this is the point. So the – lot of the models of meta cognition, which are based around something called signal detection theory, are hierarchical. Which exactly as intuition would say is – you’re basic judgement and your confidence in it, are being driven by the same source of information. So you’d expect them to correlate. What this dissociation that we’ve discovered tells us is, at least in some instances, there’s some additional information being drawn upon for that second process. So it’s not a, not a simple hierarchy. It’s not just the same information feeding the decision of that information from the decision feeding your confidence. It doesn’t have to be that way.

David: Yeah okay. That’s incredible, that’s – it’s awesome. I love it. Yeah, I mean, I think a lot of us have experienced this meta cognition that you’re talking about. Like, you know that you know something, but you can’t actually get to the information. So like say, you know, someone asks you, “Who’s the guy that played in that movie?” And you know you know the name of the actor.

Ryan: Yeah.

David: But you can’t produce it. And then after thinking about it for a while, the name pops into your consciousness. And so the information and then the knowledge that you know the information – and the confidence that you’re correct about the fact that you know that information. It’s all – it seems as though that– Our research and other research is showing that these are all sort of separate processes in the mind. And maybe even separate physically as far as neurologically speaking. Am I on the right path there? Absolutely. I can see the connection. Sort of that feeling of knowing, as some of the research refers to. No, I can see there is a dissociation there. Where you know you know this, but you can’t tell me it right now. It’s like – and you can say, well you could confidently say, “Yes I know this, I just can’t give it to you right at this minute. Come back in half and it will have popped into my mind. Absolutely, and that’s a very real effect. It’s an interesting one, which does show similar sort of dissociation, and I would agree with that – and say it makes sense.

David: So, for someone who’s never heard of– This is all – this all sort of plugs into– Would you agree that this is adding to the research, the body of research that makes up signal detection theory? Your research is adding to that. Is that what this sort of plays into?

Ryan: Not so much. I mean, well signal detection theory has been around since the 50’s, and it’s been a hugely successful– It’s a wonderful, wonderful theory – really useful in so many domains. I mean it’s – psychology adopted it quite early on and combined it…

David: Could you sort of tell – very briefly explain to people what it is, and then sort of– And then correct me for being wrong.

Ryan: Well I can – I can try a little. It’s, basically the theory is a means to the – discern the difference between the information in some input, some signal – hence the name. And the random noise that occurs with anything. So obviously you can perhaps see that most clearly in some of the areas it was applied initially – in radar for example. Where you’ve got some signal there, but there’s a lot of static, there’s a lot of interference. And it was a theory that enables you to estimate – quite accurately estimate the extent to which you can differentiate the signal from the background noise. In everyday terms, your ability to see something that’s flashed on a screen. You’ve got noise in your visual system – all the time there’s visual noise there. And you’ve got this input, this signal coming in. And your ability to distinguish say a circle from a square, is a similar sort of contrast. So you’re saying, “Okay how much am I able to tell these things apart, and how much – how separate is– How much am I able to separate this signal, this circle that’s been presented from noise? Some sort of just noise in my visual system, if this a very, very low, low contrast stimulus for example. It’s application of the same psychology – there’s been 2 forms of it. One which is absolutely wonderful and been really effective. And that is to measure, provide us a bias free measure of meta cognition. What do we mean by bias free? Well here’s the challenge, okay? So if you give someone a task – say artificial grammar learning experiment as an example. If somebody were to say that it obeys the rules 90% of the time, okay? And let’s say only half of them do obey the rules. But they say 90% of the time. Now obviously, that’s going to affect their overall accuracy. So if you’re just looking at their accuracy, you can’t really tell how much they’re able to separate out grammatical from un-grammatical strings separately from the extent of their positive bias, if you like. They’d like to say yes more often than no. Now another participant, might tend to say “yes,” exactly 50% of the time. Now they, those 2 individuals could have the same ability to actually discern between strings, but because of that additional bias, that tendency to say “yes” more often than no. In one case and not in the other. It’s difficult to get at the – an actual sensitivity if you like. Their ability to distinguish these signals. What signal detection theory has given us, is a means by which you can establish that sensitivity independent of the bias. It’s a little complex, but essentially what you do is you– What it gives you is estimate, separate estimates for the – the thresholds that you’re putting around this – the key signal sensitivity. So that’s a little confusing way to put it, but essentially, what you’ve got – is you’ve got one point within this distribution where you’re saying, “Okay, well if it’s below this, then I think it’s – there wasn’t actually a circle there, above this I think there was.” But then outside of each of those – so further below again and further above again. You may have another threshold which says, “Above that, I have some confidence in my decision.” So your original decision might be based on whether it’s above or below an arbitrary value of 100 – just to give it a value. But you might have a confidence threshold at 90 and 110. So above and below that, you’re saying, “No I’m sure there wasn’t.” And above it, “Yes, I’m sure there was.” So it separates it out in that way. That’s quite a complex way of explaining it. It is – I mean, it is fundamentally, it is a complex mathematical theory. But it’s been, it’s been applied – as I say, in psychology, in 2 ways. First way is to give us this bias free measure of meta cognition. Which it does really well, and it’s a really nice theory. It’s not perfect, and there have been refinements of it. But it is, it is nice in that regard. Where it falls down, and where our finding with blind insight really challenges it. Is where it’s been adopted as a model of how meta cognition emerges.

David: Okay, yeah.

Ryan: So there, if you’ve followed a signal detection model, you would assume that there has to be – it’s relentlessly hierarchal. You can’t have more information being used at the meta cognitive level, and your confidence. Then what’s available to your first order judgement. And that’s, that’s where we can say – well that – it clearly isn’t the case. Otherwise you wouldn’t be able to have this relationship between confidence and accuracy that we see, when your decisions are a chance. And that’s – simple as that. Yeah, so and – I know that seems very complex. And as a caveat or as an aside, I love when psychologists demonstrate that psychology is actually quite complicated. And it’s not just going, “I wonder if people do this.” And then you’re like, you put people in a lab, and go “I guess they don’t.” No, there’s a lot–There’s a whole range of psychology to be fair. And I mean I, we all typically sort of have envy for those people slightly further onto the harder end of the science. I touch a little bit on neuroscience, but I’m nowhere, I’m not really a neuroscientist, I’m a cognitive scientist, but cognitive psychologist. So–

David: Yeah, yeah, yeah.

Ryan: I’m a little bit, I’m much more hard end than say a social psychologist. Which might do some of the more – more accessible studies as you say.

David: Right, right.

Ryan: Or have envy for the…of those people doing it, a little bit – the harder science as it were.

David: But it’s – it is fantastic, because this a – this a nut that psychology is trying to crack. And neuro science at the same time. And they’re both having to deal with this black box and trying to figure out what goes in and what comes out and how it’s all related. And the – this notion that it’s all low level – sensory inputs that slowly get– In a granular way, built up to the higher level cognition. And it sets up this problem that you’re talking about, which is that – if that’s always the case. It makes it so much simpler than most likely is. Seems actually, the research seems to be shaking out that it’s very complicated. Some of the speculation that you include in your paper, was that there could be all sorts of supportive, parallel nerual networks and other things that all play into both – not just the confidence, but also the answer. And it’s not just a straight up relationship between–

Ryan: Absolutely.

David: Between – just, there’s just one system that gives you confidence or lack of. And then there’s another system that gives you the answer or not. There’s a lot more going…involved… correct, is that what you’re saying?

Ryan: That’s right. And it’s not totally one process completes and feeds onto the next. Which we would be very surprised if it were, in fairness. Because we know from the neurology that the brain has an enormous number of these re-entrant loops, if you like. So there’s a lot of feedback in these processes. It’s not simply a feed forward system where one process feeds onto the next. These processes feedback to the ones before them, so they’re constantly influencing themselves. So in a very real sense, we’d be surprised if it was purely hierarchical, because it doesn’t fit with what we know from the neuroscience of systems with feedback.

David: Right.

Ryan: I just feel – that’s, it’s wonderful that this is the true way that it works. Because I would rather us be more complex than I thought we were when I was younger.

David: And so here’s a question, and this is the question that you can answer as an expert that a lot of people want to know. And this is something that we keep getting more and more information from the world of psychology that is sort of introducing the – laypeople to the idea of this 2 model concept.. This intuition versus reason. Or this sort of like – the thinking fast and slow model that is slowly working it’s way out there. And the question I think that comes up when you start to get a good grasp of that is – when should we be trusting from your research? From your perspective especially with this blind insight thing that you’re slowly uncovering here. When should we be trusting our gut, and when should we be trusting our pure, rational logic? And of course, one of the problems is that we don’t always know which one is which, whenever we’re coming to a decision or making a judgement.

Ryan: Yeah.

David: But like, do you have any sort of tips or advice for being a good user of a human brain to–

Ryan: Yeah, no it’s a fair question. I mean, most of my early research used implicit learning. And that, that very explicitly looks at implicit versus explicit learning. Which is another way of looking at this – these 2 systems. So implicit learning, learning that happens without your awareness essentially. And then explicit learning, where you’re very conscious of what you’re learning. And there, you see certain tasks where explicit learning works really well. And other tasks where it really doesn’t. And artificial grammar learning, it’s one of those where actually it doesn’t work that well for most people. But if they just relax and just look at the strings and just take in the information – that implicit process, that process without too much analytical action going on, does work very well to improve performance. What you find is where the system – where the context is – there’s a lot of complex information there that’s not in a particularly systematic format. But there’s a lot of information in the incoming information. Then actually relying on these implicit processes is almost just necessary and inevitable. Because you couldn’t break it down and quickly think through all the logic. So if you try to apply an analytical process to something which is just too complex to really be – be handled in that way. What you end up with is a pretty poor estimate, based on very small amount of information. Whereas under those circumstances, you may be better off just sitting back and going with your intuitive judgement. Certainly in my experiments with artificial grammar learning, I had one mathematician came and did my study. And it was quite unfortunate, ’cause it was a brain imaging study, and we had her there. And I said afterwards, “So how did it go?” And she said, “Oh it’s fine. But whenever I didn’t have any confidence, I just said ‘yes.'” And I thought, “Oh, you just ruined the study, because that’s what we wanted.”

David: I just marked C on all the answers I didn’t know.

Ryan: Exactly, and of course, she was thinking – logically she was thinking, “Well if I don’t know, then it won’t make any difference whether I just say ‘yes’ all the time. Because half the time I’d be right, half the time I’d be wrong.” But of course, she lacked the insight that we have, that knows that – when you don’t think you know, actually you still do a bit – a lot of the time. So if you let yourself guess, in those cases where you really don’t think you know anything. You are more likely to perform above chance than not. So that’s, that’s the thing. Is – a lot of the time, just realizing that, in some instances, you actually know more than you know – that you know.

David: That is fantastic. Oh my God. So the – a big takeaway, we can send people away with is that– When you’re doing the multiple choice test, don’t just fill in– Don’t just fill in C when you don’t know – actually guess.

Ryan: Yes, yes. Do use that guess. Just relax, go with your intuition. If it’s not negatively marked, you’ve got nothing to lose, and you will perform better.

David: What are you working on now? Where is your research headed?

Ryan: Well there’s – I have a number of different research interests. And there’s some other work I’ve show that the – our ability to acquire associations unconsciously. So to learn that 2 things are associated without conscious perception of either of them – is possible. Which presents a challenge for some theories of consciousness. So, that work we’re definitely following up. This particular phenomenon of blind insight. I’m also following up to see if we can recreate it on the other conditions, and in other contexts such as visual perception concepts and so on. So I’m doing that. And I’m also looking at the unconscious influences of self-control. So if you’re in a depleted state of self-control, how does this influence unconscious processing? So for example, are you more likely to be influenced by an unconscious prime within that depleted state? And those sort of questions.

David: I like that last part. That’s going to be really interesting stuff. Look, thank you so much for coming on, it was a pleasure. This is a – this really is a really new area of psychology and is very median. I think it’s awesome that you’re doing this work, and I really appreciate you coming on.

Ryan: You’re very welcome, it’s been fun.