[Intro Music]
[Ryan Weber] Welcome to 10-Minute Tech Comm. This is Ryan Weber from the University of Alabama in Huntsville and I’m focusing on something a little bit different in this episode. Technical communicators often create a lot of visual material that we are hoping that people understand and can use, but rarely do we step back and ask a more basic question, which is, “How do people see the world at all?” And so to answer that question, I’ve brought in an expert.
[Marte Otten] I’m Marte Otten. I’m a Lecturer at the University of Amsterdam at the Psychology Department and I teach and I do research into social cognition.
[Weber] Otten is the lead author on a new paper in the Journal of Brain and Cognition called “The Social Bayesian Brain: How Social Knowledge Can Shape Visual Perception.” In this paper, the authors argue that even though it may feel like we’re seeing the world form a bottom-up angle, which is taking in input and then only later adding things like social context and desires and predictions, actually that knowledge is influencing or often influencing our perceptions from the very beginning. That we bring those predictions, our past expectations and our social knowledge, to the very act of perception itself, and this has radical consequences for the way that we understand the world and for the way we see things particularly when it comes to issues like race. I hope you enjoy the interview and that you find it as interesting as I did.
[Begin Interview]
[Weber] Thanks so much for joining us on the podcast today. Really excited to have something a little different and we appreciate you joining us. So, I really-I wanted to talk about your recent article that you coauthored with a few other folks and reading it as a non-specialist, what it seems to me is that you are challenging a standard model of human perception and that model assumes that perception works as kind of a bottom-up process. So, it implies that higher-level factors such as social context only have a role in the later stages after the corresponding perceptual content has been established. So, what are the problems with this standard model?
[Otten] Well I wouldn’t say that it’s a problem, so much as that we have been seeing sort of a direction happening where it seems that low-level perception is influenced by these later high-level ideas that people have. But I think, actually this different approach comes, actually much more from artificial intelligence, and sort of perceptual models there, which have shown that these kinds of models where you take two kinds of high-level ideas are actually functioning much better than the simple bottom-up models. So, I think this is much more an example where we as psychologists have looked at, I don’t know, computer models and decided that maybe this is how the brain actually you know goes about its daily job. So, it’s not so much we identify problems except if the problem maybe was that the model that we had wasn’t functioning all that well.
[Weber] You’re offering a better model in its place and one that’s sort of inspired by how AI works to perceive the world and we’re sort of remapping that back onto the brain again.
[Otten] Yes, yes, absolutely.
[Weber] Interesting. So, the model that you offer is you call it Predictive Processing, can you kind of describe that briefly?
[Otten] Yeah. First, I need to say that it’s not my model. (chuckle) I don’t want to take any credit for it. It’s-it’s been around for, I think, almost a decade and it’s becoming pretty popular in psychology. What the model is about is basically that it says that these high-level ideas that we have, so our assumptions about the world, really our predictions about the world, what we already know, that that is actually the basis for perception. So, instead of taking the physical input and working from there you take your internal assumption and map that to the physical input. So, you simply test whether your internal assumption is correct and what you take into account is the error. So, the difference between your internal assumption and the outside world and if they’re-the error is great or if the error is big, I need to say, then you propagate that back to your internal assumption, you adjust your internal assumption, and you try again until you reach sort of maximum, where your internal assumption matching the outside world. But what that does is that it makes perception much faster because you don’t have to look into all the little details and process them but instead you only have to focus on the areas where there is an error. Where your internal assumption is not matching the outside world.
[Weber] So the pointe then is that we may think that we’re just taking the world in as it is, but we’re starting with an assumption before the perception happens and then that perception is coloring the way we see the world in some way.
[Otten] Yes, it is indeed that our final precepts or our subjective experience of the world is actually a combination of the physical inputs and the assumptions that we have. And in some cases, you know physical input is just really great, very clear, and it overpowers the initial assumption, but in some cases visual input or whatever perceptual input you have is very noisy and you know you cannot rely on it. And in those cases, the idea is that your internal assumptions actually sort of start to make up a much larger part of your perception so you could see these effects in your internal assumptions, but usually when the outside world is not really providing you with good input.
[Weber] Well and it sounds too that maybe these internal assumptions are shortcuts to make perception faster, right? Because now we only have to focus on the things that don’t match and so what it allows us to process more quickly, is that right?
[Otten] Yeah, yeah, absolutely; make it easier, faster, yeah.
[Weber] Interesting. So what kind, you said this is a model that’s been around for about ten years and is gaining acceptance, what kinds of evidence are there for the model?
[Otten] I think the evidence is sort of at low-level perceptual findings that show that we have these error back propagations, so that when the internal assumption does not match the outside world, what you see in the brain is that there is this extra activation, which really relates to the error and not to the percept itself. So, I think that was one of the more important findings that showed that, well these, these AI models, might be feasible for-from a brain perspective as well. I think from a more social perspective, from sort of my angle of interest, is that we do see that what people were part seeing does not always match what’s actually present in the outside world, right? So, from my point of view this model really matches with what we social psychologists have observed in people in our experiments. And these findings they sort of challenge this model of really bottom-up perception, where you just see what’s present in the world because we see in very many cases that people report seeing something quite different.
[Weber] Interesting. Can you give a few examples of how that might happen? Where someone would, in these experiments, where people see thing that you-that aren’t there or see things that are colored by their social perceptions?
[Otten] Yes. So, it’s often that the things are colored. There are very few examples of things that are really not there, hallucinations happen more in sort of these very basic like visual illusions, but in the social domain we see much more of sort of slight shifting of what people report seeing. So, for example there’s a very interesting experiment by Kurt Hugenberg, and what he did was he had people judge faces that morphed from happy to angry or angry to happy. Just, you know slightly more so, it’s very hard to say where the happy starts and where the angry face stops, but what his main manipulation was, was that racial background of the face; either it was a Black face or it was a White face. From what he observed, was that people saw the Black face and saw more anger in that face and they were much faster to say, “Hey this is where the face turns angry,” and they were much slower to say, “Hey this is where the face turns happy.” That’s sort of an example of how perception, or at least what people report, you know this is a physician, so this is hard to say that it’s purely perception but at least it seems to have to do something with what we perceive in the world. So that’s one example. And then there’s quite a large range of experiments that deal with for example, a perception of guns on people, so there’s these experiments where they have people identifying slightly noisy pictures that are hard to discern. And you know a subset of these pictures are guns and a subset of these pictures are tools, like drills or something. What you see is when you’re primed with a Black face, so when you’ve just seen a short presentation of Black face and then you see this hard to perceive image that you’re much more likely to say incorrectly that a tool is a gun. When you see the White face, you’re much more likely to incorrectly identify a gun as a tool and also you’re faster to sort of identify the related category although you know it’s not really that guns are related to Black people, but apparently in our minds they are.
[Weber] Sure, well that’s part of this social knowledge that you’re talking about, right? Is that in for whatever reason, society those two things have become related and so that filters the perception of the subjects.
[Otten] Or at least the decisions that people take based on the evidence, right? In this case it’s really hard to make the distinction between what people see and when people decide that they see something. So that’s also it can be a criterion shift where you need, because you’ve just seen the Black face, you think you need less information to make your decision and then you’re much more likely to make these errors and also respond faster. So, as a scientist I cannot say this is definite evidence for changes of perception, but that does seem to be something going on there and it might be related to this idea that we have that you make these predictions based on what you know and it could be that you’ve just sort of pre-excavated the gun precept in your mind and therefore are more likely to identify objects as guns because you sort of integrated with your internal assumption.
[Weber] That’s a very interesting way to look at perception. Obviously, and we like to think that we see just what’s there and not kind of what’s being-but you’re saying that that is often not the case. That we’re bringing a lot of baggage and often these predictions are good and they work for us and sometimes they are not effective. What are some of the implications of this theory on sort of understanding how humans see the world?
[Otten] Yes. I mean I wouldn’t say, “Oh you cannot trust what you see,” because usually we’re actually pretty adequate perceivers, right? But that’s mainly because in most cases we have a lot of visual evidence to go on. So, you know we can, if we’re not sure we can look again. If things are vague, you know we take a minute to I don’t know process it, but I think it has bigger implications for situations where people have to make very quick decisions on very little information and I think in those cases, this is important to realize that your internal world might color what’s happening in the outside world. In our normal daily life, usually that’s only very, very, very small subset of what we’ll be doing, right? Usually we’re just out and about, driving our car, or are walking around, I wouldn’t say like, “Oh my God, reality is an illusion,” that’s not really the case.
[Weber] Sure, you’re not saying that everything is made up and we’re just you know brains in jars or whatever it is. Yeah, I understand.
[Otten] Some philosophers have said you know that this is the implication, I’m like, “Well, yes perhaps,” but I’m pretty sure we’re actually pretty good at perceiving our environments. Not like we’re bumping into things because we don’t expect them to be there right? So, it’s not that sort of huge, I think it’s more sort of the basis of our understanding is slightly different from what we’ve usually thought about.
[Weber] Well and that sounds too like a lot of times because these predictions are good, this social knowledge is helpful in the way that we’re perceiving the world and just there are times like in the, I mean those were great examples of-the studies you mentioned with the Black faces are great examples of how this is dangerous and how this can malform our perceptions. And so it sounds like we need to be aware of when the social knowledge is changing perceptions in ways or affecting perceptions in ways that are not constructive.
[Otten] Yes, yes, and perhaps also situations in which this can happen. Situations where there’s relatively little sensory input, when it’s dark, when we only have very little time to sort of perceive what’s going on, those are the cases where we might rely much more on our internal assumptions than on the actual outside evidence.
[Weber] Great, well this is really interesting and I appreciate you sharing some of your work with us. What do you think, you mentioned your paper, some things that are next as far as continuing to experiment on this theory, what do you see as kind of the next area to look into?
[Otten] Well for me it’s going to be very important to make this distinction between perceptions and decisions. So, just to make sure that the findings that we have where we think our internal assumptions are influencing perception that’s not really just people making different decisions based on their internal criteria. For me that’s a really exciting part. It sounds maybe a bit nitpicky for the outsider, but I think that will really sort of provide some definitive evidence for this idea that we’ve generated. I think more on a global level, I am really interested in finding more of these interactions where low-level perception is influenced by what we have in mind. So, I’m sort of trying to branch out, outside of visual perception, looking for example in language perception, auditory perception to see whether we hear different stuff depending on what we assume people, sort of as a social background, or that we hear their language differently perceive different things there. So, I think that’s a pretty exciting sort of explored territory.
[Weber] Fantastic, well it sounds really interesting. Well again thanks for sharing your work with us and keep up the exciting research.
[Otten] Thank you. Thank you for having me.
[Weber] Hey, no problem.