Advancing the study of emotional development through novel modeling and coding approaches

人类情感起着激发适应性行为和促进与他人交流的关键作用，从而塑造个体的环境以实现理想的目标。然而，由于缺乏对儿童和青少年时期关键情感结构的概念化和测量的共识，科学进步受到了阻碍。虽然可观察到的情绪表达（例如，面部情感）用于人类情绪的研究，特别是对儿童情绪的研究，但这些方法既昂贵又耗时。本文提出的三项程序性研究建立在我过去nserc资助的工作基础上，通过使用多种方法和新技术，为早期情绪的本质和最佳评估提供信息。在Study 1中，我将使用新颖的数据分析方法（例如，成长模型）和机器学习来开发儿童情感的新模型。我可以访问从几个大样本儿童中收集的视频记录数据，年龄从3-11岁不等，他们之前参加了观察性实验室任务。在这项研究中，我将利用这些丰富的数据集来检验1)新颖但理论驱动的情感建模方法和2)机器学习方法对情感数据建模的效用。在研究2中，我将利用最近开发的自动化技术对现有数据集进行编码，以获得儿童情绪的新指数，并将这些指数与基于传统编码方法的指数进行比较。现存的与情感相关的自然语言指数依赖于特定的模式（如书面语言），并且主要是利用来自成年人社交媒体使用的信息开发的。这些模型在应用于青少年口语时是否产生有效的数据尚不清楚，尽管从儿童的自然语言中得出有效的情绪指数将通过减少现有评估儿童情绪方法的偏见和成本而带来巨大的好处。在研究3中，我将从一个新的儿童样本中收集数据，通过将它们与心理生理和神经成像数据联系起来，进一步验证研究1和2中得出的情绪参数和编码方法。

Human emotion serves the key functions of motivating adaptive behaviour and facilitating communication with others, thereby shaping individuals' environments to achieve desirable goals. However, scientific progress has been stymied by the lack of consensus regarding the conceptualization and measurement of key emotion constructs in childhood and adolescence. While observable emotion expressions (e.g., facial affect) are used in the study of human emotion and are especially useful for the study of childhood emotion, these methods are expensive and time-consuming to implement. The three programmatic studies proposed here build upon my past NSERC-funded work to inform the nature and optimal assessment of early emotion through the use of multiple methods and new technologies. In Study 1, I will use novel data analytic approaches (e.g., growth modeling) and machine learning to develop new models of children's emotion. I have access to video-recorded data collected from several large samples of children, ranging in age from 3-11, who previously participated in observational laboratory tasks. In this study, I will capitalize on these rich datasets to examine the utility of 1) novel but theory-driven emotion modeling approaches and 2) machine learning approaches to modeling emotion data. In Study 2, I will draw upon more recently developed automated technologies to code existing datasets to derive new indices of children's emotion, comparing these to indices based on traditional coding methods. Extant indices of natural language relevant to emotion rely on specific modalities (e.g., written language) and have been primarily developed using information derived from adults' social media usage. Whether these models yield valid data when applied to youths' spoken language is unclear, although deriving valid emotion indices from children's natural language would be of tremendous benefit by reducing the bias and expense of extant methods of assessing children's emotion.In Study 3, I will collect data from a new sample of children to further validate the emotion parameters and coding methods derived from Studies 1 & 2 by linking them to psychophysiological and neural imaging data.