Abstract:

Self-regulated learning (SRL) involves a learner’s ability to monitor and regulate their cognitive, affective, and metacognitive (CAM) processes and plays a critical role in learning about challenging domains (e.g., science, mathematics) while using advanced learning technologies (ALTs; e.g., intelligent tutoring systems, simulations, serious games, hypermedia). Additionally, emerging empirical evidence indicates that CAM processes play a critical role in learning and problem solving, and self-regulation with ALTs, however, capturing CAM processes during learning with ALTs poses several major conceptual, theoretical, methodological, and analytical challenges. For example, researchers currently measure CAM SRL processes using several on-line trace methodologies, such as concurrent think-alouds, eye-tracking, log-files, physiological sensors, etc.

While these methods have the potential to advance current SRL frameworks, models, and theories, they still pose serious challenges (e.g., temporal alignment of data channels, lacking analytical techniques and debatable accuracy of inferences made from individual and across data channels) that currently plague the field. Another major challenge is related to the use of real-time CAM trace data to make ALTs adaptive. More specifically, using these trace data can provide support for learners’ CAM processes and domain learning in real-time, by making inferences based on the temporally unfolding deployment of the learners’ CAM processes. However, issues such as the lag time between capturing real-time deployment of CAM processes, inferences made by the ALTs to adapt to learners’ needs, and the ALTs’ ability to effectively monitor and regulate its own external-regulation over time (e.g., a virtual agent self-regulates and modifies the timing, sequencing, and type of scaffolding of metacognitive judgments because it typically induces frustration in learners).

In sum, the educational effectiveness of ALTs hinges on researchers’ ability to collect real-time CAM trace data by using a myriad of interdisciplinary methods (e.g., eye-tracking) and analytical techniques (e.g., data mining, machine learning) to understand them, making accurate inferences regarding the underlying CAM processes,and modeling and embodying them in ALTs in order to enhance learners’ ability to effectively monitor and regulate their own CAM SRL processes and overall learning.As such, my talk will focus on understanding and reasoning about real-time cognitive, affective, and metacogni-tive (CAM) processes to foster self-regulation with ALTs. More specially, my talk will focus on the following: (1) a critical review of various CAM SRL models, theories, and frameworks used to analyze online trace data with ALTs (e.g., Winne and Hadwin’s information processing theory, D’Mello and Graesser’s cognitive disequilibrium model, Scherer’s and Gross’ appraisal theories of emotions); (2) an analysis and discussion of key issues related to investigating CAM SRL processes during learning with ALTs with online trace methodologies (e.g., events vs. episodes vs. states; level of granularity; macro-vs. micro-level processes; mechanisms vs. processes vs. activities; frequency vs. quality of CAM SRL knowledge and skills; learning-centered vs. basic vs. compound emotions; temporal dynamics; multichannel behavioral signatures; differentiating between SRL declarative, procedural, and conditional knowledge and skills during learning; isolating initial CAM SRL knowledge and skills prior to learning vs. those that are acquired during learning and problem solving; the role of contextual factors that may interfere with the effective use of cognitive strategies, lead to inaccurate metacognitive judgements, induce negative emotions; emotion generation and regulation); (3) a critical review of the strengths and weaknesses of several interdisciplinary online trace methodologies, including physiological (e.g., EDA, EEG, ECG, EMG) measu-res, facial expressions of emotions, eye-tracking, linguistics and bodily measurements, learner-system interact-tions (e.g., learner-virtual agent dialogue moves), log-files, etc. in relation to issues raised in (1) and (2); and, (4) a succinct synthesis of the empirical research on each of these method’s accuracy in capturing CAM processes and their use in ALTs to detect, track, model, and foster SRL and domain learning (e.g., MetaTutor, nSTUDY, Betty’s Brain, Crystal Island, AutoTutor).

Lastly, I propose future directions that will significantly augment our understanding of the role of CAM SRL processes during learning with ALTs and enhance the instructional effect-tiveness of these systems based on their ability to detect, track, model, and foster learners’ CAM SRL in positive ways (e.g., an intelligent virtual human’s ability to accurately monitor and regulate the timing and sequence of its scaffolding over time based on its impact on learners’ ability to demonstrate emotion regulation flexibility).