HCC: Small: Interactive Causal Networks

如果我们拉动一根绳子，某物立即移动，我们就可以推断出我们的行为和结果之间的因果关系。把一根棍子扔进灌木丛，如果有什么东西没了，我们就认为这也是我们造成的。抓取、击打、按压或挤压的行为都会导致世界状态的直接、偶然的视觉变化。正是通过这种时间偶然性，婴儿获得了对世界状态的基本理解，认知科学家认为，这种经验形成了概念基础，即使是非常抽象的概念也是建立在这种基础上的。因果关系无处不在，是我们理解许多不同现象的基础，包括人们如何相互影响，生态系统如何运作以及物理系统如何运作。在许多情况下，理解因果关系是探索性数据分析的全部目的。然而，目前，我们还没有很好的通用互动方法来探索因果关系。目前，这种联系一般是用简单的箭头用静态图表表示的；在某些情况下，这些箭头被标记。这样的图表既难以理解，也无法直观地表达存在的各种现象。PI在这个项目中的目标是开发视觉上引人注目且易于学习的因果关系的交互式表示。这些将建立在简单的基于物理的视觉隐喻之上，理论认为这些隐喻构成了我们对因果关系进行认知建模的能力的基础。交互符号将能够表达因果现象，包括负因果、因果阻尼、因果放大、因果阻塞以及简单的正因果。在之前的工作中，PI已经开始开发一个视觉思维设计模式系统，该系统捕获简单，有效的交互技术，以及基本的感知和认知过程，以支持视觉思维构建工具的设计过程。目前的项目将有助于进一步发展这一理论，除了提供实际的解决方案来表示因果网络的问题。研究活动将包括交互因果图符号的设计、设计实施和评估研究。多点触控屏幕将提供界面。更广泛的影响：现代认知理论才刚刚开始考虑到这样一个事实，即大多数现实世界的思维都是由人们使用交互式思维工具（如电子表格或计算机辅助设计程序）进行的。交互式图表已被证明在数据分析中是非常有效的。刷屏、动态查询和拓扑范围高亮等技术提供了强大的分析工具，由于iPhone等设备的广泛采用，多点触控交互变得越来越可行。我们需要交互式图表，以一种容易理解的方式有效地表达不同类型的因果关系。这样的图表将在各种各样的知识领域中得到应用，包括教育媒体和科学家和工程师用来探索因果网络的界面。

If we pull on a string and something immediately moves, we infer a causal relation between our action and the result. Toss a stick into a bush and if something runs out we assume we caused that too. The acts of grabbing, hitting, pressing or squeezing all result in direct, contingent, visual changes in the state of the world. It is through such temporal contingencies that infants gain a basic understanding of the state of the world, and cognitive scientists argue that such experiences form the conceptual substrate on which even very abstract concepts are built. Causal linkages are ubiquitous and fundamental to our understanding of many diverse phenomena, including how people influence one another, how ecosystems function, and how physical systems function. Understanding causal linkages is, in many cases, the whole purpose of exploratory data analysis. Yet, currently, we do not have good general purpose interactive methods for exploring causal linkages. At present such linkages are generally expressed with static diagrams using simple arrows; on some occasions these arrows are labeled. Such diagrams are both difficult to understand and incapable of expressing the different kinds of phenomena that exist in an intuitive way. The PI's goal in this project is to develop visually compelling and easy to learn interactive representations of causal linkages. These will be built on simple physics-based visual metaphors that theory suggests form the substrate of our ability to cognitively model causal relationships. The interactive notation will be capable of expressing causal phenomena including negative causation, causal damping, causal amplification, causal blocking, as well as simple positive causation. In prior work, the PI has begun to develop a system of visual thinking design patterns that capture simple, effective interactive techniques, as well as basic perceptual and cognitive processes, in order to support the design processes of building tools for visual thinking. The current project will help further develop this theory, in addition to providing practical solutions to the problem of representing causal networks. Research activities will include design, design implementation, and evaluation studies of interactive causal diagramming notations. A multi-touch screen will provide the interface.Broader Impacts: Modern theories of cognition have only just begun to take into account the fact that most real-world thinking occurs by people using interactive thinking tools, such as spreadsheets or computer aided design programs. Interactive diagrams have been shown to be extremely effective in data analysis. Techniques such as brushing, dynamic queries, and topological range highlighting provide powerful analytic tools, and multi-touch interaction is becoming increasingly available due to the widespread adoption of devices such as the iPhone. There is a need for interactive diagrams that effectively express different kinds of causal linkages in a way that can be easily understood. Such diagrams will find application in a wide variety of knowledge domains, including educational media and the interfaces that scientists and engineers use to explore causal networks.