(Autism and) The Predictive Brain Theory (in Tech)

Speaker 1 [00:00]

Either way, we still have great talks left, and we will hear now a really great talk from Danny. He's a Microsoft MVP.

Speaker 2 [00:08]

MVP.

Speaker 1 [00:09]

and also has a non-profit in Belgium and he focuses works mostly on accessibility and social impact and there I want to present you Danny with his talk then the slides are up about the predictive brain theory Danny there you go

Speaker 2 [00:38]

Thank you very much, everybody, for coming to my talk. Happy to see so many people. I am Danny de Klerk, and I am president and IT coach at EDSoft, a non-profit I founded myself. And we are connecting the IT industry and the social industry via accessible technology, digital inclusion, and training about people with disabilities, and also training of IT for them. I'm autistic myself. Here I say perks with autism. And autism has a spectrum, and even my thought has a spectrum, Someday I am autistic, someday I am a person with autism, someday I am a person with spectrum. And it's all okay if you change your mind. And next to that I miss dreaming and believing. I really believe in dreams can come true. But the content for today will not be me nagging about my dreams. It will be about a predictive brain theory. First in general, then for autism. And then we are going to apply it at the workplace. So we are going to apply some things that might occur at the workplace and how we can explain it and maybe solve some things using the Predictive Brain Theory. And we start from the past, the good past or the not so good past, because our old beliefs are coded in computers. So how we thought computers worked in the past were input, process, order. So let's say if you use it by your computer, the input can be on a web application where the button is pressed, then the process gets triggered, there is a check-in if there are any errors. If there are no errors, then you're going to put some data in the database, put it on the server, and the output is, after all the things, the data is there on the server, saved. That is input, process, output for a computer. But humans work the same, and I always try to make the joke, why are humans thinking the same or working the same? I guess what computers are invented by humans? So the input for, let's say, a very human thing, not IT thing, can be, what's your name? As an input you get the question, what's your name? And then there needs to be processing. Maybe a possibility of translation. If you are not native English speaking and I ask you what your name in English, you need to translate what it means. So and then you need to think about your name and it's all very fast, it's at a split second, but that process happened for us and then the output is saying, for example, I am Danny the clerk. So that's how we believed computers and also our brain worked. But reality is, this is the biggest lie ever. So I'm going to shock you today, because one of the big problems is, if you're following the input-process-output program, you consider that there's nothing before the input. But in real life, nothing starts from zero. Nothing starts from zero, so having nothing before an input is already something that's not possible. So if you are saying, for example, input tell me, then thinking, then output, it's not working. And guess what we did? On computers we found that it was working, so what we did, we invented, we created as humans AI. It's simplified. And there are other names which I choose here for stimuli, but let's say we have a model. Also what they call in the AI the black box, we have the model, on top of that you have predictions. It's not called the same in AI, but the concepts are the same. There's a layer on it, and then there's an input, the stimuli, and there's an output. So if we look to an AI system, we have the model. By example, we create an AI system to predict the name of our baby. Let's say we have a new baby, we expect a new baby, and there's nothing better to do than using AI to predict the name. Why not? There are name generators, by the way. So let's say in the model, then you should have all the names in one region. The stimuli will be, you are asking, triggering when the AI is asking, should I call my name, maybe give a gender or not, doesn't matter, then the prediction will be the model will do his thing and the output will be, by example, ban. So the AI system is predicting ban. As I said, humans are working the same on a new way of thinking. We have a model. This consists of all our past experiences and learnings. A stimuli is a situation, a question or a trigger, something is this to us or something or someone. We will predict what the result will be and then we have an output that is our behavior, what we are doing, what we are acting, what we are saying, what we are committing. So we're going deeper in both of those, in the big things, the model, the prediction, the input and the output, for humans right now. So not AI, now we are on the humans. Our model is also called the model of the world or our view of the world. And that's everything that we learned and everything that we experienced. And of course, all the result of that, if we experience A, we react on that, what will the result be? Okay, also all in the model and What we have in our model of the world is to be say the good the bad and the ugly Sometimes even the very ugly if we are honest about this Then you have the stimuli or we could say the the the input the stimuli You know prediction is coming later because first we have an a stimuli coming in Someone they say something some party behaves to us good or bad. We get some news. We get some notifications notifications, announcements, there are some rules maybe that are changing, restrictions that are coming, restrictions that are falling away. You get questions, everything changes, we get tasks, there's all stimuli coming into the model via the prediction layer. Because then we make a prediction. And the prediction can be right, so if we expect A and we get A it's right, if we expect B and you get A, it's wrong. So we make a prediction that might be right or wrong. We're going further on that later on in this talk because it's very important. But based on our model and the empathy we predict what the result will be. And as humans we are constantly predicting. We are, it goes so fast that we are not always, humans are not used in our society constantly mindful and thinking of with thought, but in theory we are continuously making predictions on facts. And of course we want to make the predictions as accurate as possible, because accurate predictions will cause better predictions in the future, and better predictions is a nicer experience, a more predictable experience. So we want them having as good as possible. And then we have our output, this might be our behavior, the decisions that we made, and also what the outside world says, Denny is behaving good, Denny did a great job, Denny is behaving bad, Denny is good, I'm never good. So a very short executive briefing, we don't receive input on a blank canvas, also the older model says this, we don't just process from fresh input, we have a huge mental of the world, we make predictions with that model, and it can be right or wrong, and we want to optimize it, because our brain consumes energy. And now, just for everyone, what you show here has nothing to do with autism, but now the most important part is, what's the difference between autistic individuals and non-autistic individuals? And there's a very important hypothesis, still a hypothesis, but it's so nice, and you can already use it in daily life, so that's why I'm already bringing it. A Belgian researcher, Peter Vermullen, is very hard working on this. It's one of the important and famous people in Belgium. So the hypothesis suggests the following, a too accurate of the world is causing more prediction errors. So it's too accurate, it's too detailed, in fact. Limited information about a new situation can, for this reason, trigger anxiety. And behavior based on a prediction error is often seen as bad behavior. Make it tangible. I predict that you are going to give an applause to me. Maybe my talk is so bad that you won't give an applause to me, and you will not give an applause. maybe I can be very offended for that maybe I can say something good or I can just here cry like a cat as well, not happen by just explaining but that thing is seeing as bad, if I cry here on stage if I give an angry let's say a rude statement so because there is a prediction error and we need to have more good prediction results to have a good model and a good realistic model of the world And this is more difficult in this hypothesis for people with autism than people without autism. But now we are going to explain autism at the workplace, because that's what you are here for. And we are explaining it by using the predictive brain theory. So first of all, our model is trained from past experiences. So if you look at, for your right column, those are the past experiences. Maybe at the previous workplace you get bullied a lot. Well, if you look to the left column, you will see that it still can have an impact on your current job. Maybe we are hiding our autism, we are not disclosing that we are on the spectrum, and we are hiding our true selves. Not having the position we wanted for years, maybe due to a resume that was not what leaders wanted at the company. Maybe that can result in, once we're having a job, not daring to go for promotion. Being happy as we are, we are not daring to ask for promotion because of what? Because I've never been a leader of people, for example. A lot of times, people with autism, and that's not nice, but we are honest. No, we are not honest as people with autism. Most of the time we are too honest, the word too is never good. So this can happen if we get blocked a lot about it, that we don't dare to say anymore what we are thinking if we are feeling overwhelmed, that we are blocked. Theme changes felt brutal in the past, we having panning attacks once the leader moves. The same for software. For example, if you are updating some kind of open source software package and you had a big failure once in that package, following all this when the predictive brain is with it can cause it not daring to update anymore. Oh, that's not so good because not being updated with the recent thing can also be not being secure. It's not that by the intent it is not daring and not daring is not the same as not wanting. So how do we fix this? By understanding what can happen. And of course, in addition to that, we all have a confirmation bias. So we are searching for the confirmation, we are looking for that confirmation, we have a bias and we will have it. Let's say last software crashed by the system, this will always be the case. Make it practical, we are very nervous before the update, we know we need to update the system. We have lack of sleep because we know it already that we need to update it tomorrow. Lack of sleep, we have bad concentration, we forget to check that stupid box to save something. First, system crash. The last lay-off, I was one of the unlucky ones and this will always be the case for example. Confirmation bias being so unsure and less productive due to stress, unfortunately is leading to a new layoff where you are the victim. Our customer explained at last print review that I did something wrong. He complained it was missing the feature that they really wanted. They had ten features on the list. I implemented nine features. One I didn't implement. Which one did they want? That one was, people are unpredictable, but we want to have it predictable. So the customer complies again, this time not about the future, but maybe because my behavior, because I am behaving, maybe let's say you have it being assertive, but you can also be a little bit of too assertive that you are already counting in some things and that is sort of the tiny edge of assertiveness and goodness. And our last off-site team activity gave me a panic attack because I was too overwhelmed. Unfortunately, it's mandatory, the team activity at your job. For some reason, companies make it mandatory, off-site teams. And by example, I heard a lot of people that, by example, having autism or neurodivergence or other conditions is not an excuse for not doing it. But if you have a week already before the off-site already bad sleep because you are stressed, the day will not be great to be honest. We need to be honest about this. So the stimuli we get are a fuel. So when a new stimuli get triggered, for example, an unexpected demand, a long leave for a certain colleague, we get on a one-on-one, what will happen right now? We make a prediction. As I said, I was coming back on right or wrong predictions. And the prediction can be true or can be false. A false prediction, we don't call a false prediction, we call it a prediction error. So, here you see I made it, I didn't work with colors because I want to have also much accessible for people who are color blind, so don't see in color contrast. So I didn't choose, for example, for green and red, but just for rectangles and triangles for that reason, a way to make it also accessible beyond autism. And let's say the triangle here is a wrong prediction and the rectangle is a right prediction. What can happen? You predict a good result and you have a good result, so that is a positive thing of something positive. You expect something that you like and you get it. So this is very good for your flow, in your life expect a good of even tiny good things and you get them, we are getting in a flow state, we are getting more confident, we are getting more productive, the ball is rolling in a good way. But let's say the prediction error, we expect a good result, but it's not good, we get something that we really don't want. Okay, we are internally going, we program our model, our view of the world, we are having a scenario, yeah, even most of the time it works, it doesn't work always, and I felt very disappointed that time. Okay, model will be going very accurate and may predict what causes the disappointment, causes the no and we expect the yet. Okay that's for the rectangles. Going to the triangles. A positive prediction with a negative output. So we expect the disappointed, we expect the no and we get it. This is going for more and more disappointment, more and more delusion, more and more of not trusting yourself anymore, more and more people even who weren't an impostor becoming an impostor or people who are an imposter getting more and more imposter, lower self-esteem, every great stuff. There's a statistic. This is not great, of course. But then we have a very interesting one, and that's something really fascinates me. We're expecting something negative, but we get something positive. It's a prediction error. There is a mismatch between what we predict and what we get. and then my friends people have a monkey mind then I call it welcome monkey mind lots of stuff might happen here it's really something happens if you expect something negative and let's say you want to go for let's say a promotion and you really know for sure you predict your whole prediction instead of saying I will not get it I will not get it you are there and yes, you get promoted it's nice but what you can't do then sometimes I love that kind of idea of the monkey mind so there is a good thing about all my talk we can help with predictions and model fine tuning you can help making that better and if I say you, I mean you if you are on the autism spectrum or if you are a co-worker So you have co-workers with autism. And theoretically, if you are a co-worker, if you are a person with autism yourself, you are also the co-worker of yourself. So it's helpful for everyone. So after each prediction, we know, first of all, that the model needs to update. So it's burning energy. That's also one of the reasons a lot of people with autism have problems with their energy. They are always having a low social battery, a low battery, a low energy battery, because they need to weigh more updates than people without autism. So, if you know that, we can already help with having short feedback loops. So, if you're trying, if something works, and the longer that you need to work, you're expecting more and more, your prediction is more and more in one direction, and then if it's wrong, it's like in software development, like Agile. If you're having small steps and always having feedback, feedback, feedback. It's getting more accurate and the disappointment is not as big. It will be disappointment. Disappointment is never nice, but it will be smaller than one real big disappointment. Also it can be the flying wheel for positive things. You expect positive things, you get it, and the flywheel is getting to spin and you're getting in an amazing flow state. Context is king when an announcement they made or a talk is given. There's also a whole theory from the same Peter, also Peter Vermeulen from Belgium, that is about context blindness. People with autism are context blind. Even if you remember my name, if you Google me on YouTube, I give other talks about autism where I spend way more time on context blindness, but what is context is everything that we don't know, but that most of the people with autism assume normally. So the better the context you give already, the better the prediction will be. Let's say you are making an exception for employer 1. Explain why you are making this exception. Context eliminates the negative circle of doom, let's call it, or the negative loop of self-reflection. So concrete steps right now to make more accurate predictions. When you give a new task, give context and clear expectations. If you are as a customer saying I want this feature, help prioritise it, prioritise, say what's most important and why it's most important and what end goal you want to achieve. When a manager, as a company, as a builder, as a builder of a package, as a person, as an individual contributor. If something happened, there's a big bug, do a post mortem and communicate the post mortem. What you did, what you didn't expect that to happen, but not expected. What are the measures you did to prevent that in the future? Give planned structured reviews if you are an employer, do it planned and do it structured, do it often. Don't say it only when something wrong happens. A very bad thing is never doing a review, but then you have some colleagues saying something bad about the person, I need to talk to you, please come to my desk. Now do that often, weekly, by example, what happens, then it's way more predictable. When another colleague gets an exception, give clear contacts, give clear communication. say you want to always have this special place in the landscape room but you don't get it and another person got it explain it yeah that person is having a problem with his bladder so he needs to sit close to the toilet point if you don't say it we don't know it and you can always think about why does he get that exceptions why does he get it and I also want to sit there but they don't allowing me. Be honest on an interview as an employer, what the job will be and what it will not be. What will you expect of the person and what are you then also expecting? Everyone saying you can work from home for two days, don't make it one day. Very simple, very concrete, very clear. And this will also help for people not on the spectrum, having concrete expectations. Provide a transparent promotion plan for everybody. There's nothing worse than seeing other people getting their promotion, but you don't have it. But if there is a transparent promotion plan, you know what to do, what are the things that you need to deliver, and how well you get evaluated on it. It will be predictable, you have a roadmap, and you can work on it. No surprises. I want to thank you. Then sometimes it's Q&A after this slide. You can contact me, my email address, I'm still on X. A whole story why I'm on X and not on things like Blue Sky yet, to be really honest. Still overwhelming, all those new platforms for me at this time. I'm also on LinkedIn and there is also everybody on. Thank you.

Speaker 1 [24:20]

Thank you, Danny, for the great talk. Yes, the social media platform is so overwhelming. So we have some great questions already. So you can also write your own question at talks.pycon.de. I will start now with the first question. It's quite a long one. Negative prediction equals positive outcome can persist. The mindset of I should keep my expectation low so that when there's a bad result, then I am not disappointed disappointed, and if it's good, then I'm pleasantly surprised. How can this be avoided? So how can I avoid this expectation to...

Speaker 2 [24:59]

It's already very hard, but it's a very heavy one. I also call it a political one, because you have the positive thinkers, the optimists, and you have the pessimists. Most of the people that I know, also on the autism spectrum, but also not on the spectrum, are pessimists. We have a lot of pessimists. People in the world, so we are expecting negative answers. And sometimes we get it, or we have sometimes a positive answer. For me, it's different. I'm a positive. I'm an optimist by heart. I'm a self-made man, and I did a lot of, of course, the support of communities, different communities. But I was also always, I said, dreaming is believing. I was always sure I would be there. The thing is, you can't change your mind. If you are a pessimist and negative thinker, you expect negative stuff, and you don't like it anymore, you can train your brain. You can reset your brain, step by step. Our brain, so it is the growth mindset. So you need to see what works for you. I know really people being pessimistic and also giving them negative feedback, but they are kind of accepting it. So it's really thinking, do you really want it? Do you expect it in your life? If yes, great, I don't judge you. If you have the feeling, I don't like it for myself anymore, know that things as goal mindset and coaching do exist.

Speaker 1 [26:20]

Like the next question like a short question. What is monkey mind?

Speaker 2 [26:26]

We might, let's say, if you think about our primal brain, this is when we're doing the most stupid things, the caustic monkey that we say, and I was in a talk today also where we were explaining if you give AI and we compare AI to a gun and we give it to a monkey, what will happen then? Let's say this is when the most crazy things, the most uncalled things, the most um chaotic things that we can do or find the most animal without any politeness or for your radical thinking.

Speaker 1 [27:01]

I see. Sorry. The last question is also, again, a rather longer one. So I will start. If you talk about input slash stimuli and output, you look from a behavioralist view. However, there are also psychological models that examine the internal process of the brain. like, for example, cognitivism, contractivism, would using this newer method also change the perspective on autism?

Speaker 2 [27:32]

And to be honest, I have experience with it, and I only had 25 minutes, so I can give the talk a little longer. The thing is, input doesn't always come from outside your body, it's from outside your brain. Your input stimuli can also come from within your brain. In fact, if you need to go to the toilet, you have the feeling that you need to go, as stimuli coming into your brain. So we are looking from the brain and in fact everything has an input, also internal system for your body, stress system, digestive system, if you are getting ill, if you are having pain, these are also stimuli or inputs to the brain, also everything from your own body.

Speaker 1 [28:16]

Super amazing answer We still have time now for one question. I think about a short one all this all Does this also apply to an anxiety disorders and like the question give her? Also a complimentary amazing talk so so just I will repeat a question Does this also applies to an anxiety disorder like the whole concept?

Speaker 2 [28:41]

very interesting. Myself, I am autistic and I have an anxiety disorder due to trauma next to it. Stacking up with each other, very nice. I like to be a little bit dramatic and sarcastic. I didn't see any of that theory, anything about anxiety. I also worked about newer anxiety research, also later research on anxiety, they don't talk about that theory yet. What doesn't mean that it's not working, and in fact we are having it, it is working, even without research. Let's say, let's we think, we are afraid of something bad. We are predicting, we are afraid, so anxiety, being afraid. We are afraid of something bad, we are predicting that it will happen. We expect that it will happen. Next time we will get even more anxious. The moment that it doesn't happen, prediction error, we get also confused. So no real theory, but if you really think about it, it should make sense. But I didn't with something. I did read a lot of, and followed even a course from autism researchers, but if you think Psychologically, yeah, but not proven.

Speaker 1 [30:03]

Thanks for your talk. There's still some open questions left. So if you had a question give up, please approach Danny directly I think you're open always for Questions about this topic and like giving like your Like knowledge about this. This was a great talk and we will now have like a short break for the next talk