When the world becomes ‘too real’: a Bayesian explanation of autistic perception

<p indent="0mm">Autism spectrum disorder (ASD) is a complex neurodevelopmental condition characterized by social impairments and restricted and repetitive behaviours and interests. Moreover, individuals with ASD often exhibit atypical perceptual features (i.e., hypersensitivity or hyposensitivity to stimuli). Previous perceptual models such as enhanced perceptual functioning (EPF) and weak central coherence (WCC) offer only partial explanations for differences between individuals with ASD and typically developing (TD) individuals. A novel approach to understanding these differences is the Bayesian framework, which conceptualizes perceptual processes through the lens of Bayesian computational modelling and predictive coding theory. This framework offers insight into how prediction errors (i.e., bottom-up sensory input) and priors (i.e., top-down cognitive processes) jointly influence perception in individuals with ASD in various ways. In this paper, we conducted a comprehensive review of prominent Bayesian models of ASD and delved into the intricate explanations provided by these models for both social and nonsocial symptoms of ASD. To evaluate the validity of these models, we also scrutinized a wide range of empirical evidence derived from behavioural and neuroimaging studies. We identified several notable Bayesian models in the literature, two of the most prominent being the “hypo-prior hypothesis” and the “high, inflexible precision of prediction errors in autism theory” (HIPPEA). Empirical studies have examined these theories at different levels of cognitive processing, ranging from higher-level social cognitive functions to sensory perception across multiple modalities, with varying designs and methodological details. Overall, these studies have provided equivocal support for Bayesian theories. While some studies have suggested that individuals with ASD have a lower weighting of prior expectations than TD individuals, other studies have reported inconsistent or even contrasting findings. Similarly, some studies have reported that individuals with ASD exhibit a decreased ability to adapt prediction error signals to varying contexts, whereas other studies have suggested that the neural coding of prediction errors remains intact in ASD. Furthermore, in some studies, behavioural findings were at odds with neuroimaging findings. These mixed outcomes may be attributed to participant heterogeneity, different learning timescales in the task, different presentation probabilities of stimuli material, and variations in how priors were operationalized. In addition, few empirical studies have made comparisons between different Bayesian theories of ASD or between Bayesian theories and traditional perceptual models, and most previous studies have struggled to distinguish between different types of priors. Although Bayesian theories of ASD are promising and may help us better understand atypical sensory perception in individuals with ASD, they face challenges on the empirical front. For example, there is a lack of comparisons between multiple theories within the same study, and there is a relative scarcity of current neuroscience research. At the theoretical level, following the proposal of the “hypo-priors” hypothesis in 2012, scholars have conducted further studies to develop the hypothesis and provide empirical validation. While the empirical findings have been heterogeneous, this hypothesis has the potential to enhance our comprehension of altered sensory perception in ASD individuals. Ongoing research endeavours will provide substantial empirical data, with ample opportunities to refine the hypothesis and investigatory approach. Subsequent research initiatives should include a comparative analysis of theoretical frameworks within Bayesian theories and expand the integration of neuroimaging studies. In summary, Bayesian theories have demonstrated practical utility and are supported by considerable evidence, thereby contributing to an enriched understanding of atypical sensory perception in individuals with ASD. Nevertheless, Bayesian theories remain an evolving concept, necessitating extensive future research to accommodate updates and refinements.

Through perception, an organism arrives at decisions about the external world, decisions based on both current sensory information and prior knowledge concerning the environment. Unfortunately, the study of perceptual decision-making is distributed across sub-disciplines within psychology and neuroscience. Perceptual psychologists and neuroscientists focus on the information available in stimuli, developmental researchers focus on how knowledge about the environment is acquired, researchers interested in memory study how this knowledge is encoded, and other cognitive psychologists might be primarily interested in the decision mechanisms themselves. Any perceptual task evidently involves all of these components and, at rst glance, it would seem that any theory of perceptual decision making must draw on heterogeneous models and results from all of these areas of psychology. Many researchers have recently proposed an alternative (see chapters in [8]). They suggested that Bayesian Decision Theory (BDT) is a convenient and natural framework that allows researchers to study all aspects of a perceptual decision in a unied manner. This framework involves three basic components: the task of the organism, prior knowledge about the environment, and knowledge of the way the environment is sensed by the organism [6]. In this chapter, we summarize the key points that make the Bayesian framework attractive as a framework for the study of perception and we illustrate how to develop models of visual function based on BDT. We emphasize the role played by prior knowledge about the environment in the interpretation of images, and describe how this prior knowledge is represented as prior distributions in BDT. For the sake of terminological variety, we will occasionally refer to prior knowledge as iprior beliefsi or iprior constraintsi, but it is important to note that this prior knowledge is not something the observer need be aware of. Yet, as we shall see, these implicit assumptions can be revealed through psychophysical experimentation. To introduce the Bayesian approach, we illustrate how to model a simplied problem of threedimensional perception. The problem is not realistic but our intent in presenting it is to introduce the terminology, concepts, and methods of Bayesian modeling. Following the example, we illustrate how the framework can be used to model slightly more realistic problems concerning the perception of shape from shading and from contours. We conclude with a general discussion of the main issues of the Bayesian approach. Other tutorials on Bayesian modelling of visual perception with more technical details include Knill, Kersten & Yuille [7], Yuille & B ¤

Spatial navigation is a complex cognitive activity that depends on perception, action, memory, reasoning, and problem-solving. Effective navigation depends on the ability to combine information from multiple spatial cues to estimate one's position and the locations of goals. Spatial cues include landmarks, and other visible features of the environment, and body-based cues generated by self-motion (vestibular, proprioceptive, and efferent information). A number of projects have investigated the extent to which visual cues and body-based cues are combined optimally according to statistical principles. Possible limitations of these investigations are that they have not accounted for navigators' prior experiences with or assumptions about the task environment and have not tested complete decision models. We examine cue combination in spatial navigation from a Bayesian perspective and present the fundamental principles of Bayesian decision theory. We show that a complete Bayesian decision model with an explicit loss function can explain a discrepancy between optimal cue weights and empirical cues weights observed by (Chen et al. Cognitive Psychology, 95, 105-144, 2017) and that the use of informative priors to represent cue bias can explain the incongruity between heading variability and heading direction observed by (Zhao and Warren 2015b, Psychological Science, 26[6], 915-924). We also discuss (Petzschner and Glasauer's , Journal of Neuroscience, 31(47), 17220-17229, 2011) use of priors to explain biases in estimates of linear displacements during visual path integration. We conclude that Bayesian decision theory offers a productive theoretical framework for investigating human spatial navigation and believe that it will lead to a deeper understanding of navigational behaviors.

This paper aimed at examining simulation modeling in Bayesian Decision theory and its application in day to day decision making as well as planning in water resources and Environmental engineering. It also gives more insight in the validation of prior probability. The research objectives deals with the multi-objective value of water for its wide range of purposes such as Power generation, water supply, Navigation, Irrigation, and Flood control, in the Cross River basin using Bayesian Modeling. In line with foregoing objectives, the research aim to achieve the following: (i) to lay bare the usefulness of the Bayesian theory that gives more than point estimation. It measures the magnitude of the difference between alternative actions and provides a variety of estimates for consideration, (ii) to present selected empirical results of a study employing decision-making theory as a framework for considering decision making under uncertainty. (iii) to evaluate the optimal policy or strategy or action that maximizes the expected benefit in the River Basin within the available limited resources and funds over the planning period of a course of action or alternatives. The multi-objectives arising from the development that were optimized include: Economic Efficiency, Regional Economic Distribution, State and Local Economic Redistribution, Youth Employment and Environmental Quality Improvement, which are primarily essential in Cross Rivers State and Nigeria. Methodology applied involving methods, experiments and data were collected for the River Basin Engineering Development, from Parastatals and Ministries. The conceptual framework on Bayesian Decision Model (BDM) as presented captured the iterative updates of prior probability toward achieving an optimum solution of a set problem. The analysis and presentation of results were based on simulation of Bayesian Models Iterations. Chi-square, Contingency and association and Pearson Product Moment Correlation were carried out as Interaction, reliability and Validity tests respectively. The study applied Bayesian Decision Model, where the following parameters were obtained:: (a)Posterior Probabilities of the States of Nature (b) Marginal Probability of the Courses of action, (c) Maximum Expected Monetary Value[EMV*] (d) Expected Profit in a Perfect Information[EPPI], (e) Expected Value of Perfect Information[EVPI], and (f) Expected Value of System Information[EVSI]. In the process of Iteration, and at some point the Prior becomes equal to the Posterior Probability, when this occurs an optimum solution is said to be achieved. However, the correlation of prior and posterior probability is equal to one (1) at the optimum solution. In conclusion, the efficiency of system information is 50%. Table 25 indicates monetary allocation to the multi-objectives which gave a clear indication that the life wire of the watershed/dam lies on it; and therefore should be comparatively considered; because without it, it will be difficult to maintain the watershed. The Basin Authority is expected to pay the researcher the Expected Value of System Information (EVSI) value of = ₦0.1billion for information generated using the Bayesian Decision theory model spreadsheet. The value of Economic efficiency optimized from 1 st iteration to 2 nd Iteration with the EMV values of ₦2.54billion to ₦2.74billion respectively as in [ Table 4 & 15] Keywords : Optimum Solution, Prior-posterior, Probability, River Basin. DOI : 10.7176/CER/11-2-11 Publication date :March 31 st 2019

Sensory perception in autism is strikingly heterogeneous, with individuals showing both hypo- and hypersensitivity across different sensory domains. While sensory differences are widely recognized as a core feature of autism, the structure and underlying patterns of this variability remain poorly understood. Previous studies have yielded mixed findings, often examining sensory processing in isolation within single domains, thereby limiting a comprehensive understanding of sensory sensitivity in autism. We compiled psychophysical data from 107 autistic and 408 age- and IQ-matched non-autistic individuals across 32 experimental conditions spanning multiple perceptual domains, including size, brightness, orientation, pitch, and face processing. Two complementary statistical approaches were used: segmented regression and a Bayesian hierarchical model. Despite substantial inter- and intra-individual variability, both models revealed a consistent domain-specific pattern: on average, autistic individuals showed reduced sensitivity to faces and speech, while performance on basic non-social tasks was comparable to or exceeded that of the comparison group. Bayesian modelling further indicated that social relevance, rather than domain alone, accounted for the primary source of divergence between groups. This study focused on sensory sensitivity thresholds and did not assess perceptual biases or changes in subjective appearance of the stimuli. A full account of perception in autism requires considering these broader alterations. The current findings suggest that sensory differences in autism reflect a structured perceptual profile shaped by social relevance, stimulus complexity, and individual variability. The results highlight the importance of individualized sensory profiling and may inform both theoretical models and personalized approaches to intervention in autism.

A Bayesian perspective on delusions: Suggestions for modifying two reasoning tasks

We present here a Bayesian framework of risk perception. This framework encompasses plausibility judgments, decision making, and question asking. Plausibility judgments are modeled by way of Bayesian probability theory, decision making is modeled by way of a Bayesian decision theory, and relevancy judgments are modeled by way of a Bayesian information theory. These theories are discussed in Parts I, II, and III, respectively, of this thesis. Bayesian probability theory is fairly well known and well established. Bayesian probability theory is not only a powerful tool of data analysis, but it also may function as a model for the way we (implicitly) do induction, that is, the way we make plausibility judgments on the basis of incomplete information. In Part I of this thesis we will make the case that Bayesian probability theory is nothing but common sense quantified. The Bayesian decision theory, as proposed in this thesis, derives directly from Bayesian probability theory. In this decision theory we compare utility probability distributions, which are constructed by way of assigning utilities, that is, subjective worths, to the objective outcomes of our outcome probability distributions, which are derived by way of Bayesian probability theory. When the outcomes under consideration are monetary, then we may use the Weber-Fechner law of psychophysics, or, equivalently, Bernoulli's utility function, to assign utilities to these outcomes. This mapping of outcomes to utilities, transforms our outcome probability distributions to their corresponding utility probability distributions. That utility probability distribution which is located more to the right on the utility axis will tend to be, depending on the context of our problem of choice, either more profitable or less disadvantageous than the utility probability distribution that is more to the left. So, we will tend to prefer that decision which `maximizes' our utility probability distributions. This then, in a nutshell, is the whole of our Bayesian decision theory. In Part~II of this thesis, we will apply the Bayesian decision theory to both investment and insurance problems. Not all questions are equal, some questions, when answered, may give us more information than others. Stated differently, questions may differ in their relevancy, in relation to some issue of interest we wish to see resolved. This is borne out by the well known adage that, 'to know the question, is to have gone half the journey'. Bayesian information theory, by way of a mathematical operationalization of the concept of a question, allows us to determine which question, when answered, will be the most informative in relation to some issue of interest. The Bayesian information theory does this by assigning relevancies to the questions under consideration. These relevancies are then operated upon, by way of the information theoretical product and sum rules, in order to determine the relevancy of some question in relation to the issue of interest. The Bayesian information theory constitutes an expansion of the 'canvas of rationality', and, consequently, of the range of psychological phenomena which are amenable to mathematical analysis. For example, we may assign relevancies not only to questions, but also to the messages that are communicated to us by some source of information. The relevancy of a message represents the usefulness of that message, when received, in determining some issue of interest. By assigning a relevancy to the message, we indirectly assign a relevancy to the sources of information itself; possible examples of sources of information being the media, scientists, and governmental institutions. In Part~III of this thesis, we will give an information theoretical analysis of a simple risk communication problem. Bayesian probability has its axiomatic roots in lattice theory, as the product and sum rule of Bayesian probability theory may be derived by way of consistency requirements on the lattice of statements. One may derive, likewise, by way of consistency requirements on the lattice of questions, the product and sum rules of Bayesian information theory. So, if we choose rationality, that is, consistency requirements on lattices, as our guiding principle in the derivation of our theories of inference, then we get on the one hand a Bayesian probability theory, with as its specific application a Bayesian decision theory, and on the other hand we get a Bayesian information theory. In doing so, we obtain a comprehensive, coherent, and powerful framework with which to model human reasoning, in the widest sense.

Bayesian estimation has gained ground after Markov Chain Monte Carlo process made it possible to sample from exact posterior distributions. This research aims at contributing to the ongoing debate about the relative virtues of the Frequentist and Bayesian theories by concentrating on the qualitative dependent variable models. Two Markov Chain Monte Carlo (MCMC) methods have been used throughout this dissertation to facilitate Bayesian estimation, namely Gibbs (1984) sampling and the Metropolis (1953, 1970) Algorithm. In this research, several Monte Carlo experiments have been carried out to better understand the finite sample properties of Bayesian estimator and its relative performance to Maximum Likelihood Estimation (MLE) in probit and poisson models. In addition, the performance of the estimators is compared when inequality restrictions are imposed on the coefficients of the models. The restrictions are imposed within the context of a Monte Carlo experiment for the probit model and applied to the real data in the poisson regression framework. The research demonstrates the ease with which the inequality restrictions on the coefficients of the probit and poisson models via the Gibbs sampler and Metropolis Algorithms, respectively. It has been shown throughout the research that sample size has the largest impact on the risk of the parameters in both techniques. Bayesian estimation is very sensitive to prior specification even in the case of non-informative priors. Lowering the variance of the non-informative prior improves the Bayesian estimation, without significantly changing the nature of the distribution. In the cases where Bayesian prior variance is very large, MLE dominates the Bayesian in the almost all of the experimental designs. Whereas, when the prior variance is lowered, the improvement in the estimation process is remarkable. In the constrained cases, the Bayesian estimator has lower variance and lower MSE when the restrictions are correct. As the specification error increases, the Bayesian estimator suffers more than the MLE. The increase in bias is more than the efficiency gain for the Bayesian case. The effects of changes such as the changes in the distribution of regressors, parameter values, collinearity, and their interactions warrant more investigation.

Siblings of children with neurodevelopmental conditions have unique experiences and challenges related to their sibling role. Some develop mental health concerns as measured by self-reported surveys or parent report. Few data are available at the population level, owing to difficulties capturing wide-scale health data for siblings. Data linkage is a technique that can facilitate such research. To explore the application of population data linkage as a research method to capture health outcomes of siblings of children with neurodevelopmental conditions. Peer reviewed papers that captured health outcomes for siblings of children and young adults with neurodevelopmental conditions using population data linkage. JBI Scoping review methods were followed. Papers were searched within CINAHL, Ovid, Scopus, and Web of Science from 2000 to 2024 using search terms relating to 'data linkage' 'neurodevelopmental conditions' 'siblings' and 'health outcomes'. The final data extraction included 31 papers. The neurodevelopmental conditions of index children were autism, attention deficit hyperactivity disorder, intellectual disability, cerebral palsy and developmental delay. The mean follow-up time was 31 years, and the majority of studies originated from Scandinavia. Sibling health outcomes observed were psychiatric diagnoses, self-harm and suicide, other neurodevelopmental conditions, and medical conditions such as atopic disease, cancer and obesity. Data linkage can help capture sibling health outcomes quickly across large cohorts with a range of neurodevelopmental conditions. Future research could be enhanced by focusing on siblings as the primary group of interest, increased integration of genealogical data, and comparisons between diagnostic groups and severity levels. Adoption of established rigorous reporting methods will increase the replicability of this type of research, and provide a stronger evidence-base from which to inform sibling supports.

Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by social communication deficits and restricted or repetitive behaviors. Parents play a significant role in research, clinical practice and policy development on autism. Parents' perceptions of autism can affect not only their own well-being, but also their children's development. Nevertheless, few studies have examined the parents’ perception of autism in Chinese context. The parents’ perception of autism questionnaire was applied to collect information from 171 families of children with ASD, mainly to investigate the knowledge of autistic children and the difficulties the family facing. The following conclusions were drawn from this study: (1) Mother as the primary caregiver for children with ASD; (2) Parents’ perception of ASD were various parents of autistic children have a good understanding of the symptoms, causes, age of onset and interventions, and their expectations of the prognosis and the future life of their autistic children are high; (3) The difficulties that faced by the parents are mainly composed of low social acceptance, family pressure and concerns about the effectiveness of interventions for their children. In consideration of the future development of children with autism and the mental health development of their parents, it is particularly crucial to support their parents with multifaceted support.

In this paper, we propose a unified Bayesian decision theory model to integrate various components of a sensor network. We identify the key aspects of the Bayesian decision theory model, the functionalities of each network component, and the nature of interaction of the various network components in the proposed Bayesian framework. We also highlight some of the research avenues that need to be investigated for each network component. Finally, we present the use of Bayesian decision theory to schedule sensors in an energy-bandwidth constrained sensor network for target tracking

Numerical methods commonly employed to convert experimental data into interpretable images and spectra commonly rely on straightforward transforms, such as the Fourier transform (FT), or quite elaborated emerging classes of transforms, like wavelets (Meyer, 1993; Mallat, 2000), wedgelets (Donoho, 1996), ridgelets (Candes, 1998), and so forth. Yet experimental data are incomplete and noisy due to the limiting constraints of digital data recording and the finite acquisition time. The pitfall of most transforms is that imperfect data are directly transferred into the transform domain along with the signals of interest. The traditional approach to data processing in the transform domain is to ignore any imperfections in data, set to zero any unmeasured data points, and then proceed as if data were perfect. Contrarily, the maximum entropy (ME) principle needs to proceed from frequency domain to space (time) domain. The ME techniques are used in data analysis mostly to reconstruct positive distributions, such as images and spectra, from blurred, noisy, and/or corrupted data. The ME methods may be developed on axiomatic foundations based on the probability calculus that has a special status as the only internally consistent language of inference (Skilling 1989; Daniell 1994). Within its framework, positive distributions ought to be assigned probabilities derived from their entropy. Bayesian statistics provides a unifying and selfconsistent framework for data modeling. Bayesian modeling deals naturally with uncertainty in data explained by marginalization in predictions of other variables. Data overfitting and poor generalization are alleviated by incorporating the principle of Occam’s razor, which controls model complexity and set the preference for simple models (MacKay, 1992). Bayesian inference satisfies the likelihood principle (Berger, 1985) in the sense that inferences depend only on the probabilities assigned to data that were measured and not on the properties of some admissible data that had never been acquired. Artificial neural networks (ANNs) can be conceptualized as highly flexible multivariate regression and multiclass classification non-linear models. However, over-flexible ANNs may discover non-existent correlations in data. Bayesian decision theory provides means to infer how flexible a model is warranted by data and suppresses the tendency to assess spurious structure in data. Any probabilistic treatment of images depends on the knowledge of the point spread function (PSF) of the imaging equipment, and the assumptions on noise, image statistics, and prior knowledge. Contrarily, the neural approach only requires relevant training examples where true scenes are known, irrespective of our inability or bias to express prior distributions. Trained ANNs are much faster image restoration means, especially in the case of strong implicit priors in the data, nonlinearity, and nonstationarity. The most remarkable work in Bayesian neural modeling was carried out by MacKay (1992, 2003) and Neal (1994, 1996), who theoretically set up the framework of Bayesian learning for adaptive models.

Paul Green had it right, and we are seeing evidence of it again 40 years later. Advances in Bayesian computation, the collection of new and unique data, and the development of complex models has been the focus of much research in Bayesian modeling over the last 20 years. Now, armed with these tools, researchers and managers again have the ability to emphasize, as Green did in the early 1960s, the application of Bayesian theory to making improved decisions — Bayesian Decision Theory (BDT).

Precise identification of the time when a process has changed enables process engineers to search for a potential special cause more effectively. In this paper, we develop change point estimation methods for a Poisson process in a Bayesian framework. We apply Bayesian hierarchical models to formulate the change point where there exists a step change, a linear trend and a known multiple number of changes in the Poisson rate. The Markov chain Monte Carlo is used to obtain posterior distributions of the change point parameters and corresponding probabilistic intervals and inferences. The performance of the Bayesian estimator is investigated through simulations and the result shows that precise estimates can be obtained when they are used in conjunction with the well-known c-, Poisson exponentially weighted moving average (EWMA) and Poisson cumulative sum (CUSUM) control charts for different change type scenarios. We also apply the Deviance Information Criterion as a model selection criterion in the Bayesian context, to find the best change point model for a given dataset where there is no prior knowledge about the change type in the process. In comparison with built-in estimators of EWMA and CUSUM charts and ML based estimators, the Bayesian estimator performs reasonably well and remains a strong alternative. These superiorities are enhanced when probability quantification, flexibility and generalizability of the Bayesian change point detection model are also considered.

Probabilistic models in human sensorimotor control