Understanding and Reasoning About Cognitive, Metacognitive, and Affective Processes Used During Complex Learning With Advanced Learning Technologies

• In Event: 41.025 - Learning, Engagement, and Assessment in Technology-Rich Learning Environments

Abstract

Objectives and methods: Contemporary research on self-regulated learning (SRL) with advanced learning technologies (ALTs) focuses on the collection and analysis of complex, temporally-unfolding process data using various interdisciplinary methods. In our research, we have used various process measures of SRL data, including concurrent think-alouds, log-files, eye-tracking, physiological sensors, facial expressions of emotions, dialogue moves, etc. to examine the role of cognitive, affective, metacognitive, and motivational (CAMM) processes deployed both by individuals alone and between individuals (e.g., artificial pedagogical agents and learner interactions). The use of these methods yields rich, contextualized, multi-modal data (e.g., utterances of cognitive and metacognitive processes, video streams of facial expressions of emotions, text files from log-files about behavioral sequences, etc.) of temporally unfolding SRL processes, which challenge methodological approaches and traditional statistical analyses (e.g., violate statistical assumptions, unit of analyses, sampling rate, level of granularity, temporal alignment, level of description, and different time scales). These challenges pose significant problems for the advancement of research in the area of SRL. In this presentation, we will present a sample of process SRL data (from several studies) during human learning with various ALTs and discuss the methodological and analytical issues and challenges related to using multi-modal SRL data to understand learning.
Results: The process data to be presented during our presentation will be based on a synthesis of data across several studies that have examined college students‘ SRL while they used MetaTutor to learn about a complex science topic. MetaTutor is a multi-agent intelligent tutoring system that includes four pedagogical agents to detect, model, track, and foster SRL. In a typical study, learners are randomly assigned to either a control (no prompts or feedback from pedagogical agents) or an experimental group (received learning prompts and feedback regarding their use of cognitive and metacognitive SRL processes from four pedagogical agents) during a two-hour session. Several types of data were collected from each participant at various points throughout the learning session. They included multi-channel process data (i.e., concurrent think-alouds, log-files, facial expressions of basic and learner-centered emotions, eye-tracking, and electrodermal data); self-report measures of emotions, motivation, metacognition, and agent persona; knowledge construction activities (i.e., notes and drawings); and, pretests and posttests of the science topic.
Significance: Overall, the results of these studies tend to indicate that participants assigned to the experimental group learned significantly more than those in the control condition. Our results will focus on the process data (specifically on episodes of learner-agent interactions throughout the learning task) since they provide the best evidence regarding the temporally unfolding nature of cognitive, metacognitive, and affective SRL processes deployed by learners in real-time. The focus on this presentation will be on the multi-channel process data.

Authors

• Roger Azevedo, North Carolina State University

• Michelle Taub, North Carolina State University

• Nicholas Vincent Mudrick, North Carolina State University

• Joseph Grafsgaard, North Carolina State University

• Garrett C. Millar, North Carolina State University

• Megan Price, North Carolina State University