In reflexively autocatalytic foodset (RAF)-generated networks, nodes are not only passive transmitters of activation, but they also actively galvanize, or “catalyze” the synthesis of novel (“foodset-derived”) nodes from existing ones (the “foodset”). Thus, RAFs are uniquely suited to modeling how new structure grows out of currently available structure, and analyzing phase transitions in potentially very large networks. RAFs have been used to model the origins of evolutionary processes, both biological (the origin of life) and cultural (the origin of cumulative innovation), and may potentially provide an overarching framework that integrates evolutionary and developmental approaches to cognition. Applied to cognition, the foodset consists of information obtained through social learning or individual learning of pre-existing information, and foodset-derived items arise through mental operations resulting in new information. Thus, mental representations are not only propagators of spreading activation, but they also trigger the derivation of new mental representations. To illustrate the application of RAF networks in cognitive science, we develop a step-by-step process model of conceptual change (i.e., the process by which a child becomes an active participant in cultural evolution), focusing on childrens' mental models of the shape of the Earth. Using results from (Vosniadou & Brewer, 1992), we model different trajectories from the flat Earth model to the spherical Earth model, as well as the impact of other factors, such as pretend play, on cognitive development. As RAFs increase in size and number, they begin to merge, bridging previously compartmentalized knowledge, and get subsumed by a giant RAF (the maxRAF) that constrains and enables the scaffolding of new conceptual structure. At this point, the cognitive network becomes self-sustaining and self-organizing. The child can reliably frame new knowledge and experiences in terms of previous ones, and engage in recursive representational redescription and abstract thought. We suggest that individual differences in the reactivity of mental representations, that is, their proclivity to trigger conceptual change, culminate in different cognitive networks and concomitant learning trajectories.

中文翻译：

---

概念变化的自催化网络模型

在反射性自催化食物集 (RAF) 生成的网络中，节点不仅是激活的被动传递者，而且它们还主动激发或“催化”从现有节点（“食物集”）合成新的（“食物集衍生”）节点）。因此，RAF 非常适合模拟新结构如何从当前可用的结构中生长出来，并分析可能非常大的网络中的相变。RAF 已被用于模拟进化过程的起源，包括生物（生命的起源）和文化（累积创新的起源），并可能提供一个将进化和发展方法整合到认知的总体框架。应用于认知，食物集包括通过社会学习或个人学习预先存在的信息获得的信息，以及食物集衍生的信息项目是通过心理操作产生的，从而产生新的信息。因此，心理表征不仅是传播激活的传播者，而且还触发了新心理表征的衍生。为了说明 RAF 网络在认知科学中的应用，我们开发了概念变化的逐步过程模型（即儿童成为文化进化的积极参与者的过程），重点关注儿童的心理模型地球的形状。利用 (Vosniadou & Brewer, 1992) 的结果，我们模拟了从平面地球模型到球形地球模型的不同轨迹，以及其他因素（例如假装游戏）对认知发展的影响。随着 RAF 规模和数量的增加，它们开始合并，桥接以前划分的知识，并被一个巨大的 RAF（maxRAF）所包含，它限制并支持新概念结构的脚手架。在这一点上，认知网络变得自我维持和自我组织。孩子可以可靠地根据以前的知识和经验构建新的知识和经验，并进行递归的表征重新描述和抽象思维。我们建议个体差异心理表征的反应性，即它们触发概念变化的倾向，最终导致不同的认知网络和伴随的学习轨迹。