3 of 23

Problem Statement

Domain of Learning: Social sciences (or Social Work)

Target learners: Graduate students who enrolled in an online (asynchronous) course

Problem Statement:

Studies show online learners often struggle with motivation, cognitive load, and self-monitoring (Broadbent & Poon, 2015; Jansen et al., 2020);
Graduate students are expected to be self-directed, yet often receive less structured guidance than undergraduates;
Raises the question: How can we better support and motivate SRL for graduate students in asynchronous settings?

Why it Matters: As many universities adopt online programs, our solution can benefit a growing and diverse population of advanced learners.

Graduate students in asynchronous online courses face greater challenges in managing their own learning compared to those in face-to-face settings, making self-regulated learning especially critical. Research has shown that online learners often experience higher levels of cognitive load, decreased motivation, and weaker self-monitoring due to limited instructor presence and peer interaction (Broadbent & Poon, 2015; Jansen et al., 2020). These challenges are particularly acute for graduate students, who are expected to engage with more complex and autonomous learning tasks compared to undergraduates. Unlike undergraduates who often receive more structured guidance, graduate students are assumed to be self-directed—but this assumption can leave many without adequate support in online environments. This raises the question: how can we better motivate and scaffold SRL for graduate learners in asynchronous settings? As universities expand their online offerings, addressing this issue has the potential to benefit an increasingly diverse population of advanced learners.�This study focuses on social sciences—specifically UNC’s online graduate course in social work—which offers both a relevant context and a rich dataset to explore how we might design effective SRL support for this learner group.

4 of 23

Online learners lack interaction with instructor/classmates
Online learning environments often provide learners with high levels of autonomy and low levels of instructor presence (Jansen et al., 2020).
Canvas lacks features that actively support or track students’ SRL behaviors. It provides limited tools for goal setting, progress monitoring, or reflection, and does little to motivate students to develop or sustain SRL skills.

Student-facing Learning Analytics Dashboards (LADs)

LADs help visualize SRL behaviors, goals, and progress (Uysal & Horzum, 2021)

Integrating adaptive AI elements

AI has been shown to enhance:

Engagement and metacognitive awareness (Holstein et al., 2020);

Self-monitoring and strategy use (Roll & Winne, 2015);

Students also perceive AI as motivating and cognitively supportive (Jin et al., 2023)

Challenges

Solution

Visualize SRL behaviors

Motivate learner autonomy and reflection

Provide intelligent, personalized feedback

Add into Current LMS Canvas

One key challenge in asynchronous online learning is the lack of regular interaction with instructors and peers, which can undermine students’ ability to self-regulate. As Jansen et al. (2020) note, online environments often offer high autonomy but low instructor presence, placing greater responsibility on students to plan, monitor, and reflect on their learning independently. However, platforms like Canvas provide limited support for these SRL processes—offering few tools for goal setting, progress tracking, or reflection, and doing little to actively motivate students. To address this, we propose embedding a student-facing Learning Analytics Dashboard (LAD) into Canvas. These dashboards can visualize students’ learning behaviors and contexts (Uysal & Horzum, 2021), making SRL processes more transparent and actionable. Additionally, by integrating adaptive AI elements, we can deliver personalized feedback and goal recommendations tailored to each learner. Prior studies have shown that AI-driven tools can significantly improve student engagement, metacognitive awareness, and learning performance in digital environments (Holstein et al., 2020; Roll & Winne, 2015). Students generally view AI as supportive of cognition and motivation (Jin et al., 2023), making it a promising mechanism for scaffolding SRL and improving learning outcomes in asynchronous settings.

5 of 23

Current Learning Environment

Module Progress Indicator

Discussion Assignments

Other Assignments (Padlet, Reflection, Quiz)

6 of 23

Current Learning Environment

Course #	Discussions/Module
SOWO 510	1.8
SOWO 730	2.9
SOWO 770	3.1 (including video responses)
SOWO 833	2.5
SOWO 841	3.2
Total	2.7

Discussions/Padlets are central to student engagement.

Course #	Ungraded Assignments	Graded Assignments
SOWO 510	11	14 (5 if similar assignments grouped)
SOWO 730	8	3
SOWO 770	21	4
SOWO 833	7	7
SOWO 841	9	4

Other assignments vary. How can we compare if number, type, and %-of-grade differ?

With regard to student interactions, discussions and Padlets are the most common across courses. From a sample of five courses in the MSW program, we can see that students must contribute to almost three discussions per module. Though discussions are not often graded, they provide a substantial amount of written data which can highlight our target SRL processes and motivation indicators. The content of a discussion post reveals the aspects of a student’s comprehension and their strategies for building it, such as connecting to other sources/readings, asking questions, or responding to peers’ thoughts and ideas. The content can also reveal insights to students’ emotions and implicit feelings by assessing their tone and word choice. Compared to discussions, other assignments vary in quantity and type, making it difficult to adequately compare student performance across courses. Discussions, on the other hand, appear in more consistent amounts, reflect similar learning processes, and are graded similarly (generally for participation).

7 of 23

Current Learning Environment

Author, date and time posted

Same information present in replies

Canvas Discussion Boards

Padlet Discussion Boards

Author, rough date, full text content including replies

With our focus on discussion activities, we must acknowledge the differences between the two types. Canvas discussion boards are specially designated within the LMS and provide detailed information regarding the author, date and time of posting for both original posts and replies. This precise time provides insight to how students schedule and regulate their activities across a module. Padlets on the other hand, are from an outside resource and must be embedded within a standard page, meaning if we were to only examine pages with the “Discussion” label in Canvas, Padlets would be missed. Padlets are also much more challenging to properly analyze. Users are anonymous by default, meaning students must remember to put their names in their posts. Dates and times are also far less precise, and the content itself is not easily transferable to a Canvas-based data system. For these reasons, we focus primarily on discussions within Canvas itself, though we do aim to use the text content from Padlet posts, as it should reflect the same quality of writing as traditional discussions and will expand our working data set.

8 of 23

Learning Theories

Zimmerman's cyclical phases model (Zimmerman, 2000)

Planning: set goals and select strategies

Execution monitoring: track performance and adjust as needed

Reflection: evaluate outcomes and refine approaches

Self-Determination Theory – Intrinsic Motivation (Ryan & Deci, 2000)

The dashboard is designed to meet the student's:

Autonomy: the ability to customize goals and paths;

Competence: seeing their progress;

Relatedness: the encouragement that comes from seeing peer learning.

Control-Value Theory - Motivation/Autonomy

When students are in control of the learning process, their motivation and emotional experience will be enhanced (Alibeigi et al., 2024).

Our solution—a student-facing Learning Analytics Dashboard (LAD) enhanced with adaptive AI—draws upon multiple learning theories. At its core, the dashboard is structured around Zimmerman's cyclical phases model of SRL (Zimmerman, 2000), which outlines three key phases: Planning, where students set goals and select strategies; Execution monitoring, where they track progress and adjust as needed; and Reflection, where they evaluate outcomes and refine future approaches. Each panel of the dashboard is designed to support one or more of these phases through interactive and visual elements.

To guide motivation and engagement within this structure, we integrate Self-Determination Theory (SDT), which offers a solid framework for designing LADs to promote student engagement. This theory posits that learners are intrinsically motivated when their three basic psychological needs (BPNs)—autonomy, competence, and relatedness—are satisfied (Ryan & Deci, 2000). Students are likely to engage in their learning when motivated (Reeve, 2012). Our dashboard fosters autonomy by allowing students to customize weekly goals and select from AI-generated suggestions. It promotes competence through visualizations such as progress charts, login frequency, and task completion rates, helping students recognize their achievements. It also incorporates relatedness by showing anonymized peer benchmarks, allowing learners to situate themselves within a learning community.

In addition, we apply the Control-Value Theory of achievement emotions, which emphasizes the importance of perceived control and value in learning. By offering AI suggestions, LIWC emotion analysis, and goal-setting tools, the dashboard enhances students’ sense of control over their learning journey. When students feel they can influence their outcomes and see personal relevance in their tasks, their motivation and emotional engagement improve (Alibeigi et al., 2024). Together, these three theories inform a solution that supports SRL behaviorally, cognitively, and emotionally.

9 of 23

Theory-based Dashboard Design

Learning Processes Enacted: Goal setting, self-monitoring, strategic adaptation, and emotional awareness

Observable through user interactions (e.g., updating goals, completing tasks, writing reflections)

Planning

Monitoring

Reflection

At the top banner, students encounter a daily motivational quote dynamically generated from sentiment analysis (e.g., LIWC score) of their recent discussion board posts. This feature encourages emotional reflection and is informed by Control-Value Theory, which highlights how emotions like enjoyment, frustration, or boredom can significantly influence students’ motivation to learn. By activating positive emotions, the dashboard aims to strengthen both intrinsic and extrinsic motivation.

In the left panel, students can enter their own weekly goals and review AI-generated suggestions that are personalized based on prior behaviors such as submission timing, engagement with optional resources, and assignment grades. This planning function is rooted in the autonomy component of Self-Determination Theory (SDT), which posits that when individuals pursue personally meaningful goals, they experience a greater sense of ownership and are more likely to sustain focus and motivation. Allowing students to define their own objectives also reinforces perceived control, as emphasized in Control-Value Theory.

The main panel supports the monitoring phase. In the upper section, students track their task completion and login frequency through visual analytics, which reinforce SDT’s concept of competence—students feel more capable and confident when they see evidence of their progress. The lower section promotes relatedness by allowing students to compare their behaviors with peers through indicators like completion percentiles and discussion engagement. This supports the SDT principle that social connection fosters sustained motivation.

Finally, the right panel supports reflection and affective awareness. A mood tracker summarizes students’ emotion based on recent reflections, paired with a line chart tracking students' positive and negative tone in reflections, and AI-generated coping strategies. This feedback loop is designed to reinforce emotional regulation and motivation, consistent with Control-Value Theory.

Overall, learning processes enacted through this dashboard include goal setting, self-monitoring, strategic adaptation, and emotional awareness. These are made observable through user interactions such as updating goals, completing tasks, and writing reflections.

10 of 23

Data drawn from Canvas

Canvas LMS can provide trace behavioral data reflecting SRL processes, and the structure of the program provides valuable supplemental data. By aligning SRL processes with elements from SDT, we have identified the relevant data required to support motivation. To capture the planning process and promote autonomy, we need to understand student trends and identify areas for improvement. These trends include grades, submission timing, and broader clicks in the system. To capture monitoring and promote competence, we must assess how a student’s plan is being enacted in an ongoing module. We can capture data such as number of logins, time spent in the module, and percentage of the module completed throughout the week to create a visualization of student progress. We have also paired monitoring with relatedness, which we can visualize using the same data as with competence, but for the class as a whole. By pairing both competence and relatedness, we allow students to monitor their progress individually and with relation to their peers. For the reflection process, we utilize students’ discussion posts. By assessing the content of discussions and reflections for emotion, we can gain insight to how students feel about the class, indicating whether they feel that their current strategies are working or not. We capture this emotion data to track how students sense of success change over time and provide meaningful encouragement.

11 of 23

Raw Data for Top Banner & Right Panel: Reflection & Emotional Quotes

Raw Data for Left Panel: Planning

Raw Data for Main Panel: Monitoring Users Themselves

Raw Data for Main Panel: Monitoring Peers

Raw data

To extend the previous discussion on trace behavioral data, the current dataset provides detailed evidence of how each SRL process may be operationalized in practice.

For planning, we include not only grades and submission timing, but also resource-level interactions such as reading_click_count and video_click_count, enabling the dashboard to infer not just when students act, but what kinds of materials they prioritize. These distinctions allow us to tailor planning suggestions beyond generic time-management nudges.

In support of monitoring, we differentiate between self-tracking and peer-oriented behaviors. Fields like module_time_spent, logins_this_week, and task_completion_rate help model students’ own pacing, while discussion_reply_count, discussion_avg_words_per_post, and group_assignment_interactions capture how students interact with classmates. This dual structure allows the dashboard to scaffold both competence and relatedness in ways aligned with Self-Determination Theory.

For reflection, our use of sentiment data—liwc_positive and liwc_negative—enables real-time emotional awareness modeling. Unlike prior systems that treat discussion forums purely as participation metrics, we extract affective signals from student writing to inform dynamic feedback tied to Control-Value Theory. This approach enables the system to deliver targeted encouragement or regulation strategies in response to how a student appears to feel, not just what they do.

Altogether, these raw data fields enable a layered, responsive system that translates learning theory into personalized learner support.

12 of 23

Logic Diagram

= Starting Points/Key Metrics

= System Displays

= AI Generated Outputs

= Variables/Data Sets

13 of 23

Logic Diagram Branch 1 (Grades & Planning)

= Starting Points/Key Metrics

= System Displays

= AI Generated Outputs

= Variables/Data Sets

= Data used in multiple branches

Our first branch reflects the planning phase. To properly assist students in planning, we must first understand their performance in the class, such as with grades. If students are struggling in the class (defined as currently having a C or worse), we must understand which activities or behaviors are involved and whether they relate to SRL. For example, if a student has a C in a class and is spending less time on a module than their peers, it is reasonable to assume that there is some error in implementing SRL, whereas if the same student is spending more time on a module than their peers, the challenge may come from the material itself, not the student’s behaviors. However, planning to spend more time working on a module is not an effective strategy on its own, as it does not address the are that needs attention. Therefore, time-spent must be further supported by data on other student behaviors. For example, if a student spends less time than peers, and we have data suggesting they are less engaged in discussions (i.e. fewer interactions/replies or the content is not adequately substantive/analytic according to LIWC), then we may conclude that students are not adequately comparing their own understanding to their peers. Therefore, the dashboard can recommend the student takes more time reviewing their peers’ thoughts, asking questions, or reviewing topics they did not contribute to meaningfully. A similar process is outlined with “interactions with optional materials”. The system can recommend specific materials for students to review if they are not spending enough time on the module and if they are not viewing specific content.

We also hope to provide recommendations based on predictions gathered from previous coursework. For this process, we must understand the types of assignments students have/will complete and how much they were/are worth. It is impractical to use quiz grades to predict success on a presentation, as they require inherently different skills and are worth widely different amounts of a student's grade. It may be practical, however, to compare a research paper to a presentation. By examining what kinds of assignments students have completed in the past, and their performance on them, we can provide appropriate planning recommendations to students for their current course.

14 of 23

Logic Diagram Branch 2 (Progress & Monitoring)

= Starting Points/Key Metrics

= System Displays

= AI Generated Outputs

= Variables/Data Sets

= Data used in multiple branches

Our second branch focuses on the monitoring phase of SRL, where we compare student task completion with that of their peers and the course progression. Because the program we are operating in is fully asynchronous, most activities, graded and ungraded, are due at the end of their respective module, with few select activities having due dates at the midpoint of a module. This structure allows for a variety of spacing strategies to be employed, with some students potentially opting to do small chunks each day, while others complete larger chunks on some days and do no work on others. To measure the extent to which students are spacing their work, we examine the dates on which they submit assignments and their overall progress through the module as a percentage. If students consistently wait until the end of a module (the last one or two days) to submit their work, then that may be an indicator that students are procrastinating or not appropriately planning out their work schedule. However, a multitude of factors contribute to when students submit assignments, so we want to ensure that this behavior is truly a result of failing to implement SRL. For example, a single instance of late submissions does not reflect a trend, and so a reminder is all that is sent out in the dashboard. However, if students turn in most of their assignments in the last days of a module for several modules, that indicates a pattern that may require more direct intervention. Still, the assignments themselves may influence when students complete them, therefore, we see fit to compare a student’s progress with that of their peers. If all or most students submit their work in the last days of a module, that may be an indicator that the assignment or module structure is flawed, not that students are failing to implement SRL. This branch also connects to the planning branch through past assignments. By using submission trends and grades from previous classes, we can have a better approximation of a student’s habits and provide relevant recommendations for the current course. This connection to the previous branch also gives insight to student performance, meaning we can have a greater understanding of whether a student’s spacing strategies are effective or not (reflected in their grades).

15 of 23

Logic Diagram Branch 3 (Discussions & Reflection)

= Starting Points/Key Metrics

= System Displays

= AI Generated Outputs

= Variables/Data Sets

= Data used in multiple branches

The third branch is the simplest and focuses on the reflection phase of SRL. For this element, we will analyze students’ discussion board posts for emotional indicators using LIWC. Positive tone may reflect that a student is comfortable or confident with course material, whereas negative tone may indicate that a student is struggling in some areas. Should students’ writing have more negative tone than positive, we reason that they need additional help, as they express more uncertainty than confidence. To understand why students may express more negative tone, we connect this branch to the planning branch, as the students’ behaviors provide greater context for how to provide intervention. For example, if a student is performing poorly in the course, but is expressing positive emotion, this may be an indication that the student is unaware that their strategy is ineffective and that they need more motivation to plan. If a student is performing well yet still expressing negative emotion, this may be an indicator of stressors beyond the course or that the student is unaware/unconfident in their effectiveness as a learner. Emotion is not static however, and the ways it changes over time can indicate changes in a student’s behavior. A student with poor performance and high negative emotion will receive more personalized support from the dashboard, as we will have more warning indicators that the student needs help. If we observe the student’s emotions growing in positivity after receiving planning assistance from the dashboard, we propose that this is an indication that the student has adopted more effective study strategies and is feeling more comfortable with the course.

16 of 23

Mock-up of solution (Scenario 1)

At the start of week 2, student No.14 gets grade C, and his past assignments grade is similar to the current assignments. His time spent on module is below the class average, but the number of replies & quality are good. He hasn’t look at any optional materials. For the current course he submit assignments on the last day of the module. Also, students A gets lower LIWC positive tone score than negative tone score.

Here is a mock-up example demonstrating how raw learning data is processed through the logic diagram and ultimately triggers adaptive changes in the dashboard interface.

Imagine that at the start of Week 2, in a class of 15 students, Student No.14 receives a grade of C. As shown in the raw data tables, his previous assignment scores are similarly low, indicating a pattern. According to the raw data, he has spent only 31 minutes on the module—significantly below the class average of nearly 70 minutes. While his number of discussion replies and the quality of those responses are strong, he has not accessed any optional materials. Additionally, he submitted his assignment on the last day of the module. Separately, his LIWC sentiment analysis shows a lower positive tone score (2.09) compared to a higher negative tone score (4.82).

17 of 23

Mock-up of solution (Scenario 1)

18 of 23

Mock-up of solution (Scenario 1)

19 of 23

Halfway through the module,�this student’s module completion is ahead of class average,�LIWC score has improved, with a neutral emotion

Mock-up of solution (Scenario 2)

Moving on to Scenario 2, halfway through the module, this student’s completion rate has increased from 40.91% to 60.55%, putting him slightly ahead of the class average of 58.03%. His LIWC sentiment scores have also improved, with a higher positive tone (3.56) compared to a slightly lower negative tone (3.21), indicating a generally neutral emotional state.

Because the positive score now exceeds the negative, the dashboard updates the motivational quote to a more affirming message: “Keep up the good work! You’re on the right track!” However, since the difference between the two scores is minimal, the mood tracker displays a neutral emoji. Despite his improved progress, the student still holds a C grade. This combination of effort without significant academic improvement triggers an AI suggestion: “Consider reaching out to your instructor if you’re feeling overwhelmed.” The goal is to prompt the student to seek support while reinforcing their positive behavioral change.

20 of 23

Mock-up of solution (Scenario 2)

As new data are detected, not only are the AI suggestions and motivational quotes updated, but the rest of the dashboard panels have been updated accordingly.

Goals for This Week: The student choose two tailored goals, labeled in green. These goals support strategic planning, and help the student prioritize effectively.

Progress This Week: A line chart shows a steady rise in task completion, now at 60%, with logins at 2. This visual feedback reinforces the link between effort and outcome, strengthening the student’s self-monitoring.

Peer Benchmarking: The dashboard shows the student's task completion, engagement in discussions and group assignments compared to peers. This encourages peer monitoring, which can motivate students complete coursework on time, interact and engage with classmates.

Mood Tracker: The neutral emoji as well as the line chart of positive and negative emotion tracking support emotional awareness. Visualizing emotions can help students adjust their study plans and strategies in a timely manner.

Together, these features work cohesively to guide the student through SRL’s three phases—planning, monitoring, and reflection—while boosting motivation through clear goals, supportive feedback, and peer context.

21 of 23

Considerations, Limitations, and Promises

Considerations for Effective Use

Provide onboarding or guidance to help students interpret dashboard data
Ensure AI feedback is transparent and explainable to build trust
Design tool to support rather than replace educational decisions

Limitations and Cautions

Behavioral data may not fully reflect students’ intentions or challenges
Avoid using data for high-stakes decisions (e.g., grading or tracking)
Protect student privacy and data ethics
Requires iterative testing and user feedback
Co-design with instructors to ensure relevance and usability

Promises and Benefits

Used as a supportive tool, not for grading or surveillance
Designed to promote agency and reflection, not judgment
Makes SRL behaviors visible to students in real time, empowers them to set goals, monitor progress, and reflect
Provides timely, personalized feedback and emotional support
Helps bridge the gap in instructor presence in online learning

While learning analytics dashboards and adaptive AI can enhance SRL in online courses, several considerations must be addressed to ensure their effectiveness. First, students need guidance to interpret the dashboard data meaningfully; without orientation or scaffolding, visualizations may confuse rather than support learning. Therefore, onboarding tutorials or embedded prompts should be developed to help users engage with the tool intentionally. Second, while adaptive AI can personalize feedback, its suggestions should be transparent and explainable to build trust and avoid over-reliance. Third, this tool should inform—not replace—educational decisions. It is designed to support self-awareness and motivation, not to label students or determine academic outcomes.

Additionally, limitations such as data quality, privacy, and overgeneralization must be considered. Behavioral data alone may not fully capture students’ intentions or challenges. To address these issues, the tool should be iteratively tested with user feedback and co-designed with educators to align with pedagogical goals and ethical standards.

To avoid misuse, we commit to using this tool as a supportive resource—not for grading or surveillance. Its purpose is to promote student agency, not judgment. This dashboard becomes a valuable resource by making SRL behaviors visible, offering timely feedback, and fostering reflection. For graduate students in online social sciences programs, it helps bridge the gap left by low instructor presence. By guiding goal-setting, tracking progress, and providing emotional support, it empowers learners to take ownership of their learning. Ultimately, it benefits students by fostering self-awareness, sustained engagement, and better learning outcomes in asynchronous environments.

22 of 23

References

Alibeigi, M., Davoudi, M., Ghaniabadi, S., & Amirian, M. R. (2024). Enhancing Students’ Online Self-Regulation through Learning Analytics: Students' Expectations. Technology Assisted Language Education, 2(4), 1-21.

Broadbent, J., & Poon, W. L. (2015). Self-regulated learning strategies & academic achievement in online higher education learning environments: A systematic review. The internet and higher education, 27, 1-13.

Deci, E. L., & Ryan, R. M. (2012). Self-determination theory. Handbook of theories of social psychology, 1(20), 416-436.

Holstein, K., McLaren, B. M., & Aleven, V. (2019). Co-designing a real-time classroom orchestration tool to support teacher-AI complementarity. Grantee Submission.

Jansen, R. S., van Leeuwen, A., Janssen, J., Conijn, R., & Kester, L. (2020). Supporting learners’ self-regulated learning in Massive Open Online Courses. Computers & Education, 146, 103771.

Jin, S., Im, K., Yoo, M., Roll, I., & Seo, K. (2023). Supporting students’ self-regulated learning in online learning using artificial intelligence applications. International Journal of Educational Technology in Higher Education, 20(1), 37-21. https://doi.org/10.1186/s41239-023-00406-5

Reeve, J. (2012). A self-determination theory perspective on student engagement. In Handbook of research on student engagement (pp. 149-172). Boston, MA: Springer US.

Roll, I., & Winne, P. H. (2015). Understanding, evaluating, and supporting self-regulated learning using learning analytics. Journal of Learning Analytics, 2(1), 7-12.

Ryan, R. M., & Deci, E. L. (2000). Intrinsic and extrinsic motivations: Classic definitions and new directions. Contemporary educational psychology, 25(1), 54-67.

Uysal, M., & Horzum, M. B. (2021). Designing and developing a learning analytics dashboard to support self-regulated learning. In Visualizations and dashboards for learning analytics (pp. 477-496). Cham: Springer International Publishing.

Zimmerman, B. J. (2000). Attaining self-regulation: A social cognitive perspective. In Handbook of self-regulation (pp. 13-39). Academic press.

1 of 23

2 of 23

3 of 23

4 of 23

5 of 23

6 of 23

7 of 23

8 of 23

9 of 23

10 of 23

11 of 23

12 of 23

13 of 23

14 of 23

15 of 23

16 of 23

17 of 23

18 of 23

19 of 23

20 of 23

21 of 23

22 of 23

23 of 23