Finding Common Ground: A Method for Measuring Recent Temporal Context in Analyses of Complex, Collaborative Thinking
Finding Common Ground: A Method for Measuring Recent Temporal Context in Analyses of Complex, Collaborative Thinking
复制标题
寻找共同点：在复杂协作思维分析中衡量近期时间背景的方法
DOI：
--
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
D. Shaffer
中科院分区：
文献类型：
--
作者：
A. Ruis;Amanda Siebert;Rebecca Pozen;Brendan R. Eagan;D. Shaffer
复杂的协作思维通常被概念化为在不同参与者的贡献之间发展认知联系的过程。因此，以这种方式建模协作的一个中心问题是，对于讨论的任何贡献，确定用于建模正在进行的连接的适当上下文，也就是说，确定适当的最近时间上下文。最近的时间上下文通常使用固定长度的移动窗口来定义。然而，该长度取决于设置，并且不存在用于可靠地确定适当的窗口长度的现有方法。本文提出了一种实证方法来衡量最近的时间背景，从而定义一个适当的窗口长度用于分析复杂的，协作的思维。重要的是，我们描述的方法最大限度地减少了对人类注释的需求，同时为选择特定的窗口长度提供了定性和定量的保证。在学习科学中，复杂思维通常被概念化为概念之间发展认知联系的过程（DiSessa，1988; Linn，Eylon，& Davis，2004; Shaffer，2012）。在计算机支持的协作学习（CSCL）环境中，个人不仅在自己的贡献中建立这种联系，而且还与合作者的贡献建立联系（Garrison，安德森，& Archer，2001; Shaffer，2017）。因此，从认知联系的角度对复杂的协作思维进行建模的一个核心问题是，为讨论中的任何贡献确定适当的时间背景，以便对正在进行的联系进行建模。学习科学中的先前工作已经使用移动窗口来解决这个问题（Dyke，Rohit Kumar，Hua，& Rosé，2012; Siebert-Evenstone等人，2016），其中每一轮谈话都与形成其最近时间背景的讨论的某个先前片段相关联（Suthers & Desiato，2012）。在这样的模型中，对给定谈话回合的分析既考虑了其自身的内容，也考虑了其相关联的窗口的内容。由于有许多变量可能会影响最近的时间背景的程度，包括域，主题的复杂性，年龄的参与者，和通信介质，重要的是要确定一个适当的窗口长度为每个设置。然而，目前还没有可靠的方法来衡量最近的时间背景的程度，并没有进行研究，显示窗口长度的功能或解释基于窗口的学习分析模型的影响。在这项研究中，我们提出了一个实证方法来衡量最近的时间背景的程度。然后，我们通过使用不同窗口长度分析同一数据集中的概念连接性来评估该方法，以探索窗口长度对所得模型的影响。结果表明，一个适当的移动窗口的长度可以经验确定以最小的努力，而窗口长度可以显着影响模型的功能和解释，我们描述的经验方法产生相对强大的模型复杂的思维。
Complex, collaborative thinking is often conceptualized as a process of developing cognitive connections among the contributions of different participants. A central problem in modeling collaboration in this way is thus determining, for any contribution to a discussion, the appropriate context for modeling the connections being made—that is, for determining the appropriate recent temporal context. Recent temporal context is typically defined using a moving window of fixed length. However, that length is dependent on the setting, and there are no existing methods for reliably determining an appropriate window length. This paper presents an empirical method for measuring recent temporal context, and thus for defining an appropriate window length to be used in analyses of complex, collaborative thinking. Importantly, the method we describe minimizes the need for human annotation while providing both qualitative and quantitative warrants for choosing a particular window length. Introduction In the learning sciences, complex thinking is often conceptualized as a process of developing cognitive connections among concepts (DiSessa, 1988; Linn, Eylon, & Davis, 2004; Shaffer, 2012). In computer-supported collaborative learning (CSCL) contexts, individuals make such connections not only within their own contributions but also to the contributions of their collaborators (Garrison, Anderson, & Archer, 2001; Shaffer, 2017). A central problem in modeling complex, collaborative thinking in terms of cognitive connections is thus determining, for any contribution to a discussion, the appropriate temporal context for modeling the connections being made. Prior work in the learning sciences has approached this problem using moving windows (Dyke, Rohit Kumar, Hua, & Rosé, 2012; Siebert-Evenstone et al., 2016), where each turn of talk is associated with some prior segment of the discussion that forms its recent temporal context (Suthers & Desiato, 2012). In such models, analysis of a given turn of talk accounts for both its own content and the content of its associated window. Because there are many variables that may affect the extent of the recent temporal context, including domain, topical complexity, age of the participants, and communication medium, it is important to identify an appropriate window length for each setting. However, there are no reliable methods for measuring the extent of recent temporal context, and studies have not been conducted that show the effects of window length on the features or interpretation of window-based learning analytic models. In this study, we present an empirical method for measuring the extent of recent temporal context. We then evaluate the method by analyzing conceptual connectivity in the same dataset using different window lengths to explore the effects of window length on the resulting models. The results suggest that an appropriate moving window length can be empirically determined with minimal effort, and that while window length can significantly affect model features and interpretation, the empirical method we describe produces relatively robust models of complex thinking.
职前教师的协作和个人科学推理：通过认知网络分析（ENA）的新见解
DOI：
--
发表时间：
2017
期刊：
Proceedings of Computer Supported Collaborative Learning
影响因子：
--
作者：
Csanadi, A.;Eagan, B.;Shaffer, D.;Kollar, I;Fischer, F.
通讯作者：
Fischer, F.