🤖 machine learning

Self-Regulated Learning in Essay Writing: Consistency of Strategies and Impact on Outcomes

This study analyzes trace data from 93–95 secondary students in Colombia to identify three dominant self-regulated learning strategies during online essay writing, revealing that while most students consistently employed a "read first, write next" approach, the less common "write intensively, read selectively" strategy was uniquely associated with superior learning outcomes.

Original authors: Gloria Fernández-Nieto, Kiyoshige Garcés, Mladen Raković, Tongguang Li, Xinyu Li, Linxuan Zhao, Dragan Gašević

Published 2026-05-15

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Gloria Fernández-Nieto, Kiyoshige Garcés, Mladen Raković, Tongguang Li, Xinyu Li, Linxuan Zhao, Dragan Gašević

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: Learning to Drive Without a Map

Imagine learning to write an essay is like learning to drive a car. Self-Regulated Learning (SRL) is the driver's ability to check the map, watch the speedometer, plan the route, and adjust when they hit traffic.

This study looked at 93 to 95 teenagers (ages 12–15) in Colombia over two weeks. The researchers wanted to answer three questions:

What driving habits do these students have? (What strategies do they use?)
Do they stick to the same habits? (Are they consistent?)
Which habits lead to the best trip? (Which strategies get the best grades?)

The researchers didn't just ask the students what they did; they acted like "digital detectives," analyzing the digital footprints (clicks, typing, scrolling) left behind on a computer platform to see exactly how the students worked.

1. The Three "Driving Styles" (Strategies)

The researchers found that the students naturally fell into three distinct groups, or "styles," of working on their essays. Think of these as three different ways to navigate a maze:

Style A: "Read the Map, Then Drive" (Read first, write next)
- What it is: This was the most popular style (used by about 80% of students). Students spent time reading all the source materials first, like studying a map, and then started writing their essay.
- The Analogy: It's like packing your entire suitcase before you leave the house. You gather everything you need, then you go.
- The Result: It's a safe, common approach, but it wasn't the one that produced the highest scores.
Style B: "Drive and Read the Map at the Same Time" (Read and write simultaneously)
- What it is: About 17% of students did this. They would read a sentence, type a sentence, go back to read another sentence, and type again. They were constantly switching back and forth.
- The Analogy: This is like trying to drive a car while constantly stopping to look at the GPS screen. You are moving, but you are also distracted, switching your attention constantly.
- The Result: This group tended to get the lowest scores.
Style C: "Drive Fast, Peek at the Map Only When Needed" (Write intensively, read selectively)
- What it is: This was the rarest style. In the first week, only 4 students used it; in the second week, zero students used it. These students started writing immediately and only went back to read the source material if they got stuck or needed a specific fact.
- The Analogy: This is like a race car driver who knows the track well. They drive fast and only glance at the map when they need to make a specific turn. They don't stop to read the whole manual.
- The Result: Surprisingly, this was the best-performing group. Even though almost no one used it, the few who did got the highest essay scores and learned the most new information.

2. The "Habit" Check: Do They Change?

The researchers watched these students for two weeks to see if they changed their driving style.

The "Sticky" Habit: Most students who started with Style A (Read first, write next) stuck with it. They were consistent.
The "Switchers": Most students who started with Style B (Read and write at the same time) realized it wasn't working well and switched to Style A in the second week.
The "Dropouts": The few students who tried Style C (Write intensively) stopped using it by the second week. They switched to other styles.

The Takeaway: Students tend to stick to what feels comfortable (Style A), even if a different, more aggressive style (Style C) might actually get them better results.

3. The Missing "Dashboard" (Metacognition)

One of the most important findings was about what the students weren't doing.

In a car, the dashboard tells you your speed, fuel level, and if the engine is hot. In learning, this is called Metacognition (thinking about your own thinking). It includes:

Planning: "How much time do I have?"
Monitoring: "Am I understanding this?"
Evaluating: "Is my essay good so far?"

The Finding: The students were terrible at using their "dashboard." They rarely checked the time, rarely planned their steps, and rarely stopped to ask, "Do I understand this?"

The Analogy: It's like driving a car with your eyes closed, hoping you'll arrive at the destination without checking the speedometer or the road signs.
Why it matters: The paper suggests that because these teenagers are still developing these skills, they often "offload" the hard thinking work. In the modern world, this is risky because they might rely too much on AI tools to do the thinking for them, leading to "metacognitive laziness."

Summary of the Main Points

Most students play it safe: They read everything first, then write. This is consistent, but not the most effective.
The "Winning" strategy is rare: The students who wrote first and read only when necessary got the best grades, but almost no one used this method, and those who tried it stopped using it quickly.
Students are bad at self-checking: Across all groups, students rarely planned, monitored their time, or evaluated their progress. They were driving without looking at the dashboard.
The Lesson: Just because a strategy is popular (like reading everything first) doesn't mean it's the best. The study suggests that schools need to teach students how to use their "dashboard" (planning and monitoring) so they can drive more effectively, rather than just focusing on finishing the task.

What the paper does NOT say:

It does not say that AI should be banned or used.
It does not claim that all students should switch to "Style C" immediately.
It does not offer a specific app or tool to fix this yet; it simply identifies the problem and the patterns.

Technical Summary: Self-Regulated Learning in Essay Writing

Problem Statement

Effective Self-Regulated Learning (SRL) is critical for lifelong learning, particularly during adolescence when metacognitive skills consolidate. While the importance of SRL in online and blended education is well-recognized, there is a scarcity of evidence regarding how secondary school students enact SRL strategies in digital environments. Existing literature predominantly focuses on higher education, leaving gaps in understanding:

How SRL strategies develop and evolve over time in secondary contexts.
The consistency with which students apply these strategies across repeated tasks.
The specific relationship between SRL strategies and learning outcomes in secondary education.
How these insights can inform the design of online support mechanisms.

Most prior studies rely on surveys or think-aloud protocols, or they analyze single-session data, failing to capture the dynamic, longitudinal nature of strategy adaptation. Furthermore, while some work has identified SRL processes, the link between specific strategies (sequences of processes) and learning outcomes in secondary writing tasks remains under-explored.

Methodology

The study employed a micro-longitudinal design involving two secondary schools in Colombia. Data were collected from 93 students in Session 1 and 95 students in Session 2, conducted one week apart.

Task and Environment:
Participants completed a multi-source essay-writing task on a custom Moodle-based platform. The task involved:

Pre-test: Assessing prior domain knowledge (AI in medicine for Session 1; AI in education for Session 2).
Training: Brief familiarization with platform tools.
Main Task: Writing a 300–400 word essay in Spanish within 45 minutes using provided reading materials and embedded tools (annotation, search, timer, planning).
Post-test: Assessing learning gains.

Data Processing and Analysis:

Trace Data Collection: Raw logs included navigation, mouse traces, and keyboard strokes.
Mapping to SRL Processes: Using a process library based on Bannert et al.'s theoretical framework, raw actions were mapped to seven higher-order SRL processes categorized into Metacognitive (Orientation, Planning, Monitoring, Evaluation), Low Cognitive (First Reading, Re-reading), and High Cognitive (Elaboration/Organisation).
Strategy Identification:
- First-Order Markov Model (FOMM): Applied to student-session sequences to generate transition probability matrices.
- Clustering: An Expectation-Maximization (EM) algorithm was used to group students into distinct SRL strategy clusters based on their transition matrices.
Statistical Analysis:
- Consistency (RQ2): McNemar-Bowker's exact test was used to analyze strategy shifts between sessions for 87 students who completed both.
- Outcomes (RQ3): Regression analyses with Bonferroni correction compared learning outcomes (essay scores, pre/post-test scores) across the identified strategy groups.

Key Contributions

Identification of SRL Strategies in Secondary Education: The study successfully adapted a process-mining approach, previously validated in higher education, to identify three dominant SRL strategies among secondary students:
- Read first, write next
- Read and write simultaneously
- Write intensively, read selectively
Longitudinal Evidence: The paper provides rare repeated-measures evidence of SRL strategy stability and adaptation in secondary school settings, moving beyond cross-sectional snapshots.
Strategy-Outcome Association: It establishes a link between specific strategy patterns and learning outcomes in a secondary writing context, challenging assumptions derived solely from higher education research.
Metacognitive Gap Analysis: The study highlights a critical deficit in metacognitive processes (planning, monitoring, evaluation) across all identified strategies in secondary students.

Results

RQ1: Identified Strategies
Three dominant strategies were identified:

Read first, write next: The most prevalent strategy (78.47% in Session 1; 83.15% in Session 2). Students spent roughly equal time on reading and elaboration but showed a trend of shifting from reading to writing over time.
Read and write simultaneously: The second most common (17.0% and 16.84%). Characterized by frequent transitions between reading materials and the essay window.
Write intensively, read selectively: The least common (4.30% in Session 1; 0% in Session 2). Characterized by a heavy focus on writing processes with selective reading.

RQ2: Consistency of Strategies

Stability: 82.61% of students who started with Read first, write next maintained this strategy in the second session.
Shifts: Most students (80%) who began with Read and write simultaneously shifted to Read first, write next.
Abandonment: No students continued using Write intensively, read selectively in Session 2; all transitioned to other strategies.
Statistical Significance: The McNemar-Bowker test indicated that strategy changes were balanced with no directional preference ( $p = 0.392$ ).

RQ3: Impact on Learning Outcomes

Performance: Descriptive statistics suggested that the Write intensively, read selectively strategy yielded the highest essay scores (Mean = 12.33) and post-test gains, followed by Read first, write next, with Read and write simultaneously showing the lowest performance.
Statistical Significance:
- Pairwise comparisons for essay scores showed a statistically significant difference between Read and write simultaneously and Write intensively, read selectively ( $p = 0.045$ , large effect size).
- For post-test scores, Write intensively, read selectively significantly outperformed both Read first, write next ( $p = 0.047$ ) and Read and write simultaneously ( $p = 0.003$ ).
Metacognitive Deficit: Across all strategies, engagement in metacognitive processes (monitoring, planning, evaluation) was limited. For instance, monitoring occurred in only 6.4% to 11.14% of time depending on the strategy.

Significance and Claims

The paper claims that learning success in secondary online writing is shaped not merely by the frequency of strategy use, but by how SRL processes are coordinated. While the Read first, write next strategy is the most stable and common, the less common Write intensively, read selectively strategy was associated with superior learning outcomes, likely due to a more balanced and diverse selection of SRL processes (including orientation and monitoring).

The study emphasizes that secondary students often lack effective strategies for independent planning, monitoring, and evaluation. This gap is critical in the context of digital learning and Generative AI, where "metacognitive laziness" may further erode self-regulation. The authors argue that their findings support the development of Learning Analytics (LA) systems capable of detecting strategy-level patterns rather than just activity counts, thereby providing scaffolds that specifically target the strengthening of metacognitive processes.

The authors modestly note limitations, including the short duration of the study (two sessions), the lack of internal condition measures (e.g., motivation), and the need for validation against alternative detection approaches. They conclude that future research should explore strategy evolution over longer timeframes and the efficacy of analytics-informed scaffolds in fostering productive SRL during adolescence.