When are children most receptive?

[Transcript generated automatically]
I want to apologize for the lack of cute animal videos and pictures. There are some pictures, but it’s going to be a bit more dry than what you just had. So I don’t know if this was a good order, especially with the lack of coffee. What I can give you, though, is I would say interdisciplinarity. So for the past few years, I’ve been working in developmental psychology departments, but I have a background in cognitive science and computer science, and now I’m at IBED, which is the Institute for Biodiversity and Ecosystem Dynamics. So I’ve been hopping around quite a bit and in my presentation there will be a lot of human and non-human examples. So this might explain where this is coming from. So what I’m going to do today is present you some of my dissertation work. Specifically, I’m focussing on the evolution and development of sensitive periods and changing environments. And yeah, I will start by just defining a lot of things and setting the stage. Okay, so all individuals are shaped by their experiences and the time periods during which the impact of experience is greatest are called sensitive periods. Sensitive periods exist across the entire tree of life and play a crucial role during development. The sensitive periods are periods of heightened phenotypic plasticity, and plasticity is the ability of our genes to produce changes in our bodies and behaviour, our phenotypes based on experience. Ontogeny refers to the time period during which experiences can have an impact on the development of a specific trait. Although sensitive periods appear to be more common early in ontogeny, they also exist at later developmental stages such as adolescence across species and traits. Interest in such data sensitive periods has grown vastly. Empirical studies suggest, for example, that some personality traits and rodents are shaped during adolescence. To chew in adolescence are particularly sensitive to social feedback or that mites are more specifically evolved. Might Sharpless, as a teen, responds to nutritional conditions towards later developmental stages, these examples illustrate the variation that exists in the onset and duration of sensitive periods, and this variation exists at different levels of biological, organised fashion and exists, for instance, between traits within a single individual. It also exists in the same trait between individuals within species. So individual differences, which is something I’ll be coming back to later and it and it exists between species by what explains this variation in the timing and duration of sensitive periods. Over the past decades, we have made tremendous steps towards understanding the neuropathies, the logical mechanisms, underlying sensitive periods, and understanding this variation, at least I think so, is key to manipulating when and how organisms are shaped by experiences. An extreme case of such manipulation would be to reopen sensitive periods for specific experiences, so to speak, to turn back to developmental the learning clock. And today, neuroscientists have developed experimental interventions which can erase signatures of early adversity in rodents. So just a prospect of achieving something like this in humans, it’s just very. Yeah, big. Aside from neurophysiological processes, the timing and duration of sensitive periods also depends on properties of the environment. For example, do we expect children who are exposed to caregiver unpredictability to lose plasticity, assist at a similar age as children who experience consistent parenting? At present, we know actually relatively little about how environmental factors shape sensitive periods. In the past decade, there has been research and mathematical models that explore under which environmental conditions. We would expect sensitive periods to evolve and like maps, these models help us zoom in on some factors and processes while leaving out others. In order to navigate this complex space, DAF can present a much simpler version of reality, not despite, but because of their simplicity. We’ve actually learnt quite a bit about evolution and development of sensitive periods. So we have learnt, for example, that plasticity depends on information about the environment that is available to developing organisms, specifically models have found that plasticity tends to be higher early in ontogeny to higher organisms. Uncertainty about the environment at the onset of ontogeny and the better experience is during development. How to reduce this uncertainty. Models have also found that plasticity tends to decline more slowly when experiences are noisy and do not provide as much information in order to reduce uncertainty about the environment. These models have predominantly found sensitive periods early on, and one explanation for this pattern may be that the majority of these models have explored environments that are stable within the lifetime of organisms, and that these conditions, there’s typically no need for organisms to maintain costly plasticity. But as we know, most of species, such as humans or animals, are likely to experience changes in environmental conditions throughout their development. And these changing conditions may favour sensitive periods later in development. Today I present two models for my dissertation which formalise these ideas, while one explores how changes in equality of information within an organism’s lifetime shape sensitive periods. So I like why might we expect such changes in the value of information? It could be, for example, that certain experiences accused which are able to transmit information are more abundant, only available at some developmental stages, but not at us. For example, mothers transmit information to the developing foetus about the expected long term nutritional environment after birth, and this information is unique to that live stage. It’s only available in the womb. Information that the newborn receives outside the womb tends to be less informative about this long term nutrition environment. So in this example, the quality of information is highest at the foetus stage and then subsequently declines in model. To a look at how changes in the environment has played itself shape sensitive periods changes in climate a seasonality might cause environmental fluctuations, yes. And alternatively, the organism itself may move through different environments and habitats across its lifetime. For example, in humans moving within our cousin countries may drastically change the environments that we experience and that preserved by providing even more technical details about the models. So please bear with me and model one. I explored variation the quality of information in the environment. A state itself remained stable, so organisms can sample information by sampling cues throughout ontogeny and the quality of that information depends on what we call the two reliability. The reliability of Q essentially indicates how well the Q helps the organism to discriminate between different states of the world. The cure reliability can range from low to high. To make this a little bit more concrete, just try to think of a smoke detector. If you have a highly reliable, very expensive smoke detector, it will only go off in case of a fire, but not due to a steamy shower. So in this case, this smoke detector can very well discriminate these two different states. And when you hear the sound, you know, it must be very likely a fire, as indicated earlier, the reliability of cues may change across the lifespan for different reasons. The one reason that I mentioned before was that cues may be more available in some developmental stages, but not others. Or it could also be that the organism’s ability to detect cues changes across the lifespan. For instance, as their sensory systems mature or degrade with age. In Model two, I explore changes in the environment to state itself, whether to reliable or to remain stable. Here, the environment fluctuates between two discrete states, and the rate of this change is a pyramid and a model. Therefore, I can explore a range of environments from relatively stable to rapidly fluctuating. Both models extended previous models of sensitive period evolution. And one important assumption of this model is that development is both gradual and irreversible. And I would come back to that point in a second. Okay. I want to try to walk you through the basic setup of the model, hoping that this will make the results easier to understand. So at the start of ontogeny, organisms are born into one of two possible states of the environment. Just for purposes of illustration, let’s assume those are safe or dangerous and the beginning organisms are only equipped with a prior estimate of the distribution of these states. And the prior can say, for example, that both states safe and dangerous, they could likely audit one status much more likely. So maybe it’s much more likely to be in a dangerous environment in each time period. The organism then sampled secure about the environment and that cue can say it’s safe or it’s dangerous. And remember that for a specific environment there is this Q reliability that specifies how much information the organism can gain from this cue. With all of this in mind, organisms can update their estimate about the environment via patient inference, and then they get the opportunity to adjust their phenotypes and to adjust their phenotypes. They’re given three options. Either they can do nothing, so they just forego specialisation and leave everything as is and continue. Or they can specialise incrementally to what’s a safe environment. For example, that could grow slightly bigger ornaments to attract mates, or they could incrementally specialise to what’s a dangerous environment. For example, by growing bigger weapons to fight. And I’m stressing incrementally because developing these things in the model takes time. So it’s not like within just one time period do your organism can fully develop any of these phenotypic specialisations? So in that sense, development is incremental. Once such an adjustment to the phenotype has been made, it can never be undone. And that sense development is irreversible. However, an organism that had previously specialised to what’s one type of environment that started to invest in ornaments to attract mates can later on still invest into weapons to cycle continues until the end of ontogeny and the next organisms accrue fitness based on how well this phenotype that they spent time building matches their environment. And the underlying idea here is that the better the phenotype matches my environment, the more likely I am to survive and the better my chances at reproductive success. So with this rather simple set up, I can compute optimal developmental trajectories for different combinations of priors and cure liabilities. And his so-called optimal policies can be understood as a blueprint for development. So for each possible state that the organism could be in, it puts out the optimal phenotypic decision in that state to compute his optimal policies. I use a method called spastic dynamic programming and no worries, and I kind of go into details about that. I just thought it might be nice to tell you that it’s a method to take a probabilistic and a problem with probabilistic outcomes and breaks it into small set problems. Find solutions for these problems. Put them all back together in order to solve the bigger problem. If you’re interested in this method, I highly recommend the book, a very old book in the corner of the slide, but it’s actually very nice, nice to try out for the method. And two years ago I also gave a workshop on this method and ask material links to the materials for that. If you’re interested. So finally, from the optimal policy, I can simulate entire populations of organisms and their developmental trajectories. And then was just one more step in between. I’m able to quantify changes and plasticity across ontogeny in these data. So after all this effort, I get curves like this. I’m not going to go into detail about how I’m going to quantify plasticity, but if you’re interested, you can ask me later. I’m happy to elaborate. Okay, just as a reminder. So we all back on the same page and the first model, the reliability of cues can change between consecutive time periods. On one time period, it might be a certain value. And then in the next time period it could be a bit higher or lower. In the other model, the Q reliability remains constant. However, between two time periods, the state of the environment may change so it doesn’t have to. But it could. So it could be that the organism is first in a safe environment, then again safe and dangerous, safe and so on and so forth. Okay, with this basic setup in mind, I’m finally going to present some of the results and I’m going to start with the first model. So I’ve explored three different patterns of change. So the reliability of cues may continuously increase across ontogeny, so start out barely reliable and then at the highest point cues are really reliable. So if you see a cue and it says the environment is a safe, then it’s very likely that it’s actually safe. Then I explored another pattern right here reliability. First increases and then decreases resulting in this triangle, hence the triangular shape. And lastly, I explored a pattern where the cue liability continuously decreases across ontogeny I find sensitive periods towards the middle of ontogeny, potentially corresponding to adolescence in some species when the reliability of cues increases at least across some portion of ontogeny. So that is true for the increasing in a triangular pattern when the reliability of cues decreases. I only find sensitive periods at the onset of ontogeny, which makes sense because under these circumstances organisms have access to this highly reliable information early on. So there’s no need for them to remain plastic. They can just go all into specialising towards that environment. I mentioned earlier that some empirical studies suggest that adolescents may be particularly sensitive to social feedback. According to my model, sensitive periods at later developmental stages may occur when the reliability of experiences increases at some portion of crest development. So perhaps a concrete example of that is social feedback as a cue to mate value. Social feedback from the opposite sex may be more informative about your own mate value during reproductive years than in childhood or at old age. So this is very hypothetically hypothetical, by the way, but this just illustrates how very general models like mine may be used to generate specific, testable hypotheses. So let’s have a look at the next model where I explore changes in the environmental state. Here I find sensitive periods tend to evolve towards the end of ontogeny when the environment fluctuates. Frequently design in contrast results from models assuming stable environments and indicates that natural selection may heightened sensitivity to cues towards the end of development. When the environment changes rapidly and organisms develop incrementally, this makes sense when the environment changes, cues towards the end of ontogeny should be better predictors of whatever comes after that live stage. Then cues early in ontogeny. And this is also something we observe and temporal patterns of behaviour and phenotypic plasticity in some species. Going back to my bulk, my earlier involvements in nutritional conditions during the final developmental stages strongly determine whether males mature as fighters, which you can see on the left. So they are bigger and they have this extra fighting lack which I circled in red or benign and defenceless scramblers, which you can see on the right when environment changes slowly. We also observed extensive periods evolve midway through ontogeny and this model, this happens when experiences early in an ontogeny contradict organisms. Estimates of the environment to state this pattern has also been found in other models of sensitive period evolution which make different assumption assumptions. And one of these models is the one I just presented. And another one is a model where the authors assume a stable environment and you also see this delayed sensitive periods because of an incongruence see between the early ontogeny cues and the prior. So it was a specific manipulation that the artist in this model used. But what this indicates is that non early sensitive periods may be a common feature of development which might evolve across a range, across a range of environments, traits and histories. So I just want to pause and just put up like what I think is at least one of the insights I would like you to take away from this talk, which is that changing environmental conditions, more so than stable environments, may favour sensitive periods beyond early ontogeny. And as I just explained, other models have also provided these kinds of insights. And recently I submitted a paper in which we synthesise findings from both mathematical models and empirical papers which demonstrate sensitive periods beyond early ontogeny. On the bottom of a slide, you can see a link to a preprint in case you’re interested. But I’m also very briefly going to sketch out the term what we observed in that paper. So in a paper we observe that sense of how the development of social behaviours in Midtown charged near adolescence, if you want occur often in mammals, especially in humans and rodents. So we hypothesised that this may be due to increases in the value of social information as organisms start to navigate novel social landscapes during adolescence, they gained independence from their caregivers and they can sound for cues that are more relevant to them. We also observe sensitive periods towards the end of ontogeny across various invertebrate species such as snails, wasps, water, fleas and felt. Here we hypothesise that changes in the physical and social environment of these species may favour phenotypic adjustments towards the end of ontogeny to prepare them for the subsequent developmental stage. So this kind of concludes the first part of my talk and the rest will hopefully be a bit less try early on the time and the talk. I promise to come back to individual differences. More precisely, I wanted to talk about what these evolutionary models can possibly tell us about differences between individuals after same species. Okay, maybe it does get more dry again in this context. It is important. I’m sorry. In this context, it is important to note that the type of findings I presented throughout my talk are typically at the level of species differences. So that is the results indicate what patterns of plasticity may emerge for species that have evolved for specific price and reliability. So you can see the actual findings from my variance reliability model. And each quadrant is a population of species. If you want, corresponding to a unique combination of prior and reliability pattern. But the models also lend themselves to dig deeper into individual differences. What you can see here is the optimal policy that underlies the derived patterns of plasticity for a specific combination of prior IQ reliability. And remember that pattern of plasticity. I guess you can’t remember because I haven’t told you, but the pattern of this dissatisfaction average. Sorry across the population members. So these policies display the optimal developmental trajectories of the entire population for given prior to reliability. And if we look at this, we can gain some insights about individual differences both and phenotypic development and patterns of plasticity. Actually feminism, pension age. Then who started to develop these specific models of incremental development notice really early on that in these models, individual differences emerge from statistics? RAMPLING As I explained earlier, and the models organisms learn about the environment through cues and a q reliability indicates how reliably cues predict a true state after world. So due to chance, some organisms may sample more consistent sequences of cues and others more inconsistent sequences of cues. So what you can see here on the right and the resolution is not great. I apologise for dead is a optimal policy for some model for some parameter combination. But what’s what I would like you to notice that. So whenever a does brownish line goes up, organisms sample one type of cue. When it goes down, they sample the other type of cue. The colours indicate what kind of phenotypic decision organisms make. So what you notice is when it consistently goes up, up, up or down, down, down, or mainly up or maybe down, it’s always the same colour. So organisms consistently specialise towards one type of environment. However, if cues go up, down, up, down or something, some other pattern, then we get like this variability in phenotypes that they develop. And this tastic sampling can result in individual differences in phenotypes and patterns of plasticity. It turns out that this finding is quite robust across models making different assumptions, including those that are presented early on. Just talks. I can observe that same pattern in my optimal policies, leading to key insight number two that I would like to highlight, which is that drastic sampling can result in individual differences, both in phenotypes but also patterns of plasticity. Earlier in the talk, I explained that models have shown the plasticity declines more slowly when experiences are noisy and do not have to reduce uncertainty about the state of the environment. This finding is both true at between species level but also at the between individual level. So if you have populations that in the models, I keep touching this that in a model have evolved with a lower Q reliability than just by virtue of that lower reliability Qs will tend to be more noisy because the information is just less good. So for a species like that and the models we observe that justice a t is prolonged within a species, we also see that in the optimal policy that when individuals receive more inconsistent cues, that plasticity is also prolonged. So this is again the effect. US too has dark sampling. Empirical studies actually support such an explanation. Studies in humans, birds and rodents have, for example, shown that noisy, inconsistent inputs tend to prolong sensitive periods in individuals. The idea of statistics dancing also provides potential explanations for the puzzling observation of individual differences in otherwise genetically identical organisms, a mechanism which is often implicated in the development of individual differences. A feedback loops between an individual’s behaviour and their state to state can include the physical features, the personality, the social wrong of rank, but also what they know about the environment. When individuals express different behaviours during development, this can cause changes in their state, which in turn can shape subsequent behaviours and so on. So creating this loop positive feedback loops can then increase individual differences. So creating more variability while negative feedback loops might lead to behaviours which make individual states converge. Chicken Stochastic sampling of experiences is an additional component that can interact with this loop between behaviour and state, further amplifying individual differences as a consequence of sample text variances. If you are interested to learn more about a kind of theoretical evolutionary underpinnings of these feedback loops, I can recommend a recent model that is building on previous models of incremental development and provide some insights into the selection process which may result in positive and negative feedback loops. I’m just putting this up there in case you’re interested. The last thing I would like to address is the unfolding of individual differences across time. So in the models, I mainly focussed on differences between species, however, and my very own reliability model. I dug a little deeper into the development of individual differences. Specifically, I looked at the development of rank order stability to understand when doing ontogeny, individual differences develop and whether they maintain stable or can still change. For this purpose. I simulated populations of individuals and I started to rank them according how specialised they were towards one environment. So the most specialised individuals at the top and then it goes down to list and this plot I only want you to focus on the black bars so they indicate the proportion of rank switches that occur within the population of simulated organisms between two consecutive time periods. So when the bar is high or large, what this indicates is that from one time point to the next, organisms switched in rank quite a bit. So that means that individual differences have not stabilised yet. So individual population is still kind of figuring out their rank if you want. And the smaller the bars get, the more stable these individual differences have. So like the top individual will remain at the top, the bottom individual will remain at the bottom. So we can compare these bars across the different credits again to get species typical insights, but we can also focus on individual plots in order to get insights about what’s going on in that species. In the leftmost column, we observe a common pattern and empirical literature suggesting that individual differences develop and stabilise over time, causing age related increases in so-called trait repeatability, which is the extent to which individual differences in traits course are maintained over time. However, my model individual differences don’t always continuously stabilise across ontogeny. We observe that in some environments, ranks within a population might first destabilise before stabilising my models suggest hypotheses about the selection pressures that can result in the more common pattern of increasing territory palatability and in the less common pattern of well decrease seeing and then increasing trait repeatability does exist. Empirical studies suggesting that individual differences might not always stabilise continuously across ontogeny. For example, in a paper from 2015, Roots and Cougar observed large variation in repeatability across different traits and life stages in zebra finches, some traits showed no repeatability, while others were only repeatable in some portions of ontogeny. So the authors suggest that trait repeatability may often be overestimated in the literature because studies tend to only measure individuals at two time periods, which is very understandable. It’s extremely expensive to conduct studies like that, but the idea is that when we include more time periods across ontogeny, we might not always find the assumed consistency of individual differences and recent advancements in animal tracking, I think outlined a very promising future for getting a little bit to the bottom of this because they make it easier to continuously track behaviour and individuals and then to see how traits unfold across time. This kind of also brings me to my conclusion. Clearly, my work involves a very theoretical approach to understanding development and change and changing environmental conditions and development of individual differences. But I do think it’s extremely important to kind of ground these results and try to test them empirically in some way, shape or form. That’s future work. My empirically test is different reasons for sensitive periods of theta life stages and development of individual differences and various taxa such as invertebrates, insects, rodents and fish. And I think that understanding why and when sensitive to individual differences and plasticity evolves can advance our understanding and contribute to an integrative science of sensitive periods which may be down the line, could influence human research last, but not least, I wanted to thank my advisors and collaborators on these projects within Frankenheimer’s and Karthik Pension. Nathan And obviously I want to thank you for listening to this very dry talk. And yeah, I’m excited for questions and feel free to email me if anything comes up. Thank you.