Fóthi Ábel, Soorya Latha, Lőrincz András
Institute of Enzymology, Research Centre for Natural Sciences, Budapest, Hungary.
Department of Artificial Intelligence, Faculty of Informatics, Eötvös Loránd University, Budapest, Hungary.
Front Psychiatry. 2020 Dec 2;11:503462. doi: 10.3389/fpsyt.2020.503462. eCollection 2020.
Autism spectrum disorder (ASD) is a heterogeneous neuropsychiatric condition traditionally defined by core symptoms in social behavior, speech/communication, repetitive behavior, and restricted interests. Beyond the core symptoms, autism has strong association with other disorders such as intellectual disability (ID), epilepsy, schizophrenia among many others. This paper outlines a theory of ASD with capacity to connect heterogeneous "core" symptoms, medical and psychiatric comorbidities as well as other etiological theories of autism in a unifying cognitive framework rooted in neuroscience and genetics. Cognition is embedded into an ever-developing structure modified by experiences, including the outcomes of environment influencing behaviors. The key constraint of cognition is that the brain can handle only 7±2 relevant variables at a time, whereas sensory variables, i.e., the number of sensory neurons is orders of magnitude larger. As a result, (a) the extraction, (b) the encoding, and (c) the capability for the efficient cognitive manipulation of the relevant variables, and (d) the compensatory mechanisms that counteract computational delays of the distributed components are critical. We outline our theoretical model to describe a Cartesian Factor (CF) forming, autoencoder-like cognitive mechanism which breaks combinatorial explosion and is accelerated by internal reinforcing machineries and discuss the neural processes that support CF formation. Impairments in any of these aspects may disrupt learning, cognitive manipulation, decisions on interactions, and execution of decisions. We suggest that social interactions are the most susceptible to combinations of diverse small impairments and can be spoiled in many ways that pile up. Comorbidity is experienced, if any of the many potential impairments is relatively strong. We consider component spoiling impairments as the basic colors of autism, whereas the combinations of individual impairments make the palette of autism. We put forth arguments on the possibility of dissociating the different main elements of the impairments that can appear together. For example, impairments of generalization (domain general learning) and impairments of dealing with many variable problems, such as social situations may appear independently and may mutually enhance their impacts. We also consider mechanisms that may lead to protection.
自闭症谱系障碍（ASD）是一种异质性神经精神疾病，传统上由社交行为、言语/沟通、重复行为和兴趣受限等核心症状来定义。除了核心症状外，自闭症还与许多其他疾病密切相关，如智力残疾（ID）、癫痫、精神分裂症等。本文概述了一种自闭症谱系障碍理论，该理论能够在一个植根于神经科学和遗传学的统一认知框架中，将异质性的“核心”症状、医学和精神共病以及其他自闭症病因理论联系起来。认知嵌入到一个由经验不断塑造的结构中，这些经验包括环境影响行为的结果。认知的关键限制在于，大脑一次只能处理7±2个相关变量，而感觉变量，即感觉神经元的数量则要大几个数量级。因此，（a）相关变量的提取，（b）编码，（c）对相关变量进行有效认知操作的能力，以及（d）抵消分布式组件计算延迟的补偿机制至关重要。我们概述了我们的理论模型，以描述一种笛卡尔因子（CF）形成的、类似自动编码器的认知机制，该机制打破了组合爆炸，并由内部强化机制加速，并讨论了支持CF形成的神经过程。这些方面中任何一个方面的损伤都可能扰乱学习、认知操作、互动决策和决策执行。我们认为社交互动最容易受到多种微小损伤组合的影响，并且可能以多种方式累积而受到破坏。如果许多潜在损伤中的任何一种相对严重，就会出现共病。我们将组件损坏性损伤视为自闭症的基本颜色，而个体损伤的组合则构成了自闭症的调色板。我们提出了关于分离可能同时出现的不同主要损伤元素的可能性的论点。例如，泛化损伤（领域通用学习）和处理许多变量问题（如社交情境）的损伤可能独立出现，并且可能相互增强其影响。我们还考虑了可能导致保护的机制。
