The Moodoo Library: Quantitative Metrics to Model How Teachers Make Use of the Classroom Space by Analysing Indoor Positioning Traces (Extended Abstract)

Roberto Martínez Maldonado, Vanessa Echeverría, Katerina Mangaroska, Antonette Shibani, Gloria Fernández-Nieto, Jurgen Schulte, Simon Buckingham ShumIJCAI 2021 · DOI

TL;DR

This paper introduces "Moodoo," a novel library that quantifies how teachers utilize classroom space by analyzing indoor positioning data. By mapping raw x-y traces to higher-order spatial constructs, Moodoo offers a scalable and objective method to understand and potentially improve teacher-student interactions and pedagogical strategies.

What Problem Does This Paper Address?

Understanding how teachers use physical space in the classroom has long been recognized as crucial for student engagement and learning. However, this aspect of teaching has remained largely opaque to computational analysis due to its ephemeral nature. Existing methods like direct observation or self-assessments are often subjective, prone to bias, and difficult to scale. While there's a growing interest in using AI and analytics to understand classroom activity, most research focuses on student behavior or verbal interactions, leaving the teacher's spatial dynamics under-explored. This paper aims to bridge this gap by providing a quantitative framework to model teachers' spatial behaviors in a way that is meaningful and actionable for educators.

Methods

The core of this work is the "Moodoo" library, which processes raw x-y indoor positioning data from teachers to extract higher-order spatial metrics. The methodology is grounded in the concept of "Spatial Pedagogy." Key techniques include:

  • Defining "Stops": Sequences of positioning data points close in space and time are identified as "stops." These are modeled by grouping data points around a centroid, with parameters for maximum distance (d) and minimum time (t) to define a stop. This allows for the analysis of periods where a teacher is "positioned to conduct formal teaching" or interacting with students.
  • Modeling "Transitions": Trajectories between consecutive stops are defined as "transitions." A linear quadratic estimation algorithm (Kalman filtering) is used to smooth these trajectories, allowing for the analysis of movement patterns between different areas or student groups.
  • Teacher-Student Interaction Metrics: Proximity to students is a key indicator. A teacher standing within a defined "interactional space" (iDis, calibrated to 1 meter) of students is considered a potential interaction. Metrics derived include total attention time per student/group, frequency and duration of interactions, and sequencing of these interactions.
  • Spatial Entropy: To quantify how evenly a teacher's attention is distributed, a spatial entropy index is calculated. This involves creating a grid over the classroom space and analyzing the proportion of data points in each cell, with lower entropy indicating more focused positioning.
  • Co-teaching Metrics: For classrooms with multiple teachers, metrics are developed to identify potential co-teaching instances by analyzing the distance between teachers over time.

Datasets

The paper describes an authentic physics education study where seven teachers wore indoor positioning trackers across 18 classes. This study served as the basis for developing and testing the Moodoo metrics. The paper does not explicitly list a separate, publicly available dataset but rather refers to the data collected during this specific experimental deployment.

Key Insights and Ideas

  • Quantifying the Ephemeral: The most significant insight is the successful translation of low-level, continuous x-y positioning data into meaningful, higher-order spatial constructs that describe teaching behavior. This moves beyond simple location tracking to understanding how that location is used.
  • Spatial Pedagogy in Action: The work demonstrates how the theoretical framework of Spatial Pedagogy can be operationalized through quantitative metrics, providing empirical evidence for spatial teaching strategies.
  • Actionable Feedback for Teachers: By modeling metrics like "stops," "transitions," and "spatial entropy," the library offers potential for providing teachers with objective feedback on their spatial positioning and movement patterns, which can inform professional development.
  • Beyond Simple Proximity: The research goes beyond just measuring distance, incorporating temporal aspects (stops, transitions) and distribution analysis (entropy) to provide a richer understanding of teacher presence and interaction.

Key Results

The paper presents a conceptual framework and the development of metrics rather than specific quantitative results from a large-scale evaluation. The key "results" are the proposed metrics themselves and their mapping to spatial constructs. For example:

  • Teacher-Student Interaction: The paper highlights how proximity metrics (e.g., defining interactional space with a 1-meter radius) can identify potential instances of teacher-student interaction.
  • Co-teaching Detection: Figure 3 visually demonstrates how the distance between two teachers can be tracked over time to detect potential co-teaching instances when they are within a defined proximity (dTeacher = 1m) for a set duration (tTeacher = 10 seconds).
  • Spatial Entropy: The concept of spatial entropy is introduced as a way to quantify focused vs. dispersed teacher attention, with a lower entropy value indicating more focused positioning.

Limitations and Future Work

  • Calibration of Parameters: The paper acknowledges that parameters like distance thresholds (d, iDis, dObj, dTeacher) and time thresholds (t, tTeacher) need to be calibrated according to specific contexts and tracking technologies.
  • Ethical Considerations: The authors explicitly mention the need for training on the use of such tools and the risks associated with the unethical use of data.
  • Further Maturation of Tools: The paper suggests that further development of these tools will open possibilities for more evidence-based teacher professional development.
  • Integration with Other Data: While not explicitly stated as a limitation, the focus is solely on positioning data. Future work could involve integrating this with other data sources (e.g., audio, video, student engagement metrics) for a more holistic view.
  • Generalizability: The study was conducted in a specific physics laboratory setting. Further research is needed to assess the generalizability of these metrics across different subject areas and classroom types.

Summary updated

BibTeX
@inproceedings{maldonado2021moodoo,
  author = {Roberto Martínez Maldonado and Vanessa Echeverría and Katerina Mangaroska and Antonette Shibani and Gloria Fernández-Nieto and Jurgen Schulte and Simon Buckingham Shum},
  title = {{The Moodoo Library: Quantitative Metrics to Model How Teachers Make Use of the Classroom Space by Analysing Indoor Positioning Traces (Extended Abstract)}},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year = {2021},
  doi = {10.24963/ijcai.2021/654},
}
Figure