The test of understanding graphs in kinematics (TUG-K) has widely been used to assess students’ understanding of this subject. The TUG-K poses different objectives to the test takers such as (1) the selection of a graph from a textual description, (2) the selection of corresponding graphs, and (3) the selection of a textual description from a graph. Whether test takers follow these task requirements is usually inferred from evaluating the test scores as correct or incorrect, yet the process of how students actually interact with the different tasks remains unknown. Recent studies have shown that eye tracking can provide rich insight into student’s interaction with multiple-choice tasks. In the current work, we analyzed the eye movement patterns of $N=115$ high school students while solving the TUG-K. Each question was divided into a question area (Q) and an option area (O), then gaze transitions between Q and O and between different options were calculated. A cluster analysis using the transition metrics revealed three item groups, containing the aforementioned objectives of the items. The clusters remain stable for different subsamples of our dataset, for instance, considering only the correct or only the incorrect responses, or considering high- or low-confidence responses. We conclude that eye movements can reflect task demands on a procedural level well beyond the classical methods of evaluating test scores, eventually making eye tracking an additional method for item analysis that can be utilized to confirm or explore test and item structures.

中文翻译：

---

运动学中的理解图测试：通过聚类眼动转换来确认项目目标

运动学的理解图测试（TUG-K）已广泛用于评估学生对该主题的理解。TUG-K对应试者提出了不同的目标，例如：（1）从文本描述中选择图形，（2）选择相应图形，以及（3）从图形中选择文本说明。通常通过评估测试分数的正确与否来推断应试者是否遵循了这些任务要求，但学生如何与不同任务进行实际交互的过程仍然未知。最近的研究表明，眼动追踪可以为学生与多项选择任务的互动提供丰富的见解。在目前的工作中，我们分析了$ñ=115$高中生在解决TUG-K的同时。将每个问题划分为一个问题区域（Q）和一个选项区域（O），然后计算Q和O之间以及不同选项之间的凝视过渡。使用过渡指标进行的聚类分析显示了三个项目组，其中包含上述项目目标。对于我们数据集的不同子样本，聚类保持稳定，例如，仅考虑正确或不正确的响应，或考虑高置信度或低置信度响应。我们得出的结论是，眼球运动可以在程序级别上反映任务需求，远远超出了评估测验分​​数的经典方法，最终使眼动跟踪成为了一种用于项目分析的附加方法，可用于确认或探索测试和项目结构。