A learner can know what would help and still not do it.
They may understand that retrieval is better than another reread. They may accurately notice that the material is weak. They may even plan to begin early.
Then fatigue, anxiety, competing goals, time pressure or a poorly designed environment changes what happens.
Self-regulated learning, usually abbreviated SRL, studies this larger system.
SRL concerns how learners set goals, plan, monitor, select strategies, regulate motivation and emotion, manage behavior and adapt to feedback while learning. Metacognition is central, but it is not the whole construct.
SRL is an umbrella with several mature models
Ernesto Panadero's 2017 review compared six influential models of self-regulated learning and identified commonalities as well as differences.
Across models, regulation is typically cyclical rather than a one-time act. Learners anticipate and plan, act and monitor, then reflect and adapt. Motivation, self-efficacy, strategy use, emotion and behavior enter the cycle in different ways depending on the model.
The existence of multiple models is not evidence that SRL is vague beyond use.
It does mean that researchers should name the model and component being tested rather than treating "self-regulation" as a single intervention ingredient.
Metacognition sits inside SRL
Metacognitive monitoring asks how well learning is going. Metacognitive control uses that information to regulate cognitive action.
SRL includes both and adds more.
A student may monitor correctly and know what strategy to choose, but low self-efficacy can reduce persistence. Anxiety may narrow available control. A competing social goal may dominate the learning goal. Environmental distraction can overwhelm an otherwise sound plan.
This distinction prevents a common educational error: assuming that poor learning reflects poor self-awareness.
Sometimes the learner knows exactly what is happening and lacks the conditions to act differently.
Regulation is cyclical
In many SRL models, learning unfolds through linked phases.
Before action, learners interpret the task, set goals and plan. During action, they use strategies, monitor progress and manage effort. Afterward, they evaluate outcomes and update future plans.
The cycle is not always orderly in real life. Learners revise goals mid-task, abandon plans, receive unexpected feedback and shift strategies.
That messiness is not a failure of the construct. It is why process-sensitive measurement matters.
A questionnaire administered once at the end of a course cannot reveal every dynamic transition that occurred during learning.
Motivation is not an optional add-on
A purely cognitive account can identify an effective strategy and still fail to explain why it is not used.
SRL models incorporate motivational variables such as goals, value, self-efficacy and persistence because regulation is effortful and selective. Learners decide not only how to learn but whether the goal is worth pursuing and how much effort to invest.
Emotion matters too. Frustration, boredom, anxiety and curiosity can alter attention, strategy choice and persistence.
This does not mean every emotion problem should be psychologised as self-regulation failure. It means learning occurs in an organism whose motivational and affective systems are part of the task.
Measurement is moving from traits toward processes
Traditional SRL research often uses self-report questionnaires. They can efficiently capture beliefs and typical strategy use, but they are limited by memory, interpretation and social desirability.
More process-oriented work uses trace data, think-aloud protocols, observations, diaries, log files and physiological measures to study regulation as it unfolds.
The measurement challenge is substantial because a behavior can have multiple functions. Pausing might indicate reflection, confusion or distraction. Spending more time can indicate persistence or inefficiency.
Good SRL measurement therefore needs theory, task context and often multiple data sources.
Executive functions are related, not interchangeable
A 2024 systematic review by Dörrenbächer-Ulrich and Bregulla examined relationships between self-regulated learning and executive functions.
The literatures overlap around goal-directed behavior, monitoring and control, but they originate in different traditions and operate at different levels of analysis.
Executive functions typically refer to cognitive control capacities such as inhibition, updating and shifting. SRL includes motivational, metacognitive and contextual processes extending far beyond those functions.
The same caution applies to metacognition. Related constructs can interact without becoming synonyms.
Regulation can be shared
One of the most important developments in SRL research is the move beyond the isolated learner.
Sanna Järvelä and colleagues at the University of Oulu have been central to research on socially shared regulation of learning. Panadero and Järvelä's 2015 review examined evidence for regulation emerging at group level in collaborative learning.
Groups can negotiate goals, monitor shared progress, manage collective motivation and adapt strategy together.
This is not merely several individuals regulating side by side. The object of regulation can become the group's shared activity.
The Nordic contribution is especially relevant to an Institute that wants to understand human capacity without reducing it to private self-management.
Co-regulation also matters
Teachers, peers and tools can temporarily support regulation.
A teacher may model how to decompose a task. A peer may notice that the group is stuck. A digital system may prompt reflection or visualize progress. These supports can scaffold capacities learners later use independently.
But scaffolding can also create dependency if the learner never becomes able to initiate regulation without the support.
That tension becomes central in AI-supported learning.
If a system always chooses the next task, explains every error and decides when mastery is sufficient, it may improve immediate performance while changing what the learner practices regulating.
Whether that is beneficial depends on goals and evidence, not ideology.
AI creates a new regulation partner
Järvelä, Nguyen and Hadwin's 2023 paper explicitly considers human and artificial intelligence collaboration for socially shared regulation in learning.
The area is still emerging.
AI can potentially make learning states visible, provide feedback and support group regulation. It can also over-automate monitoring and decision-making, reduce productive struggle or introduce opaque recommendations.
The Institute should frame AI as a possible participant in regulation loops, not as a proven metacognitive tutor.
A serious research agenda asks which regulatory functions should be scaffolded, which should remain effortful practice, and how control can return to the learner.
Self-regulation language can individualise structural problems
SRL has practical value, but institutions can misuse it.
A school might respond to unreasonable workload by teaching students time management. A workplace might answer constant interruption with resilience training. A learning platform might blame low engagement on motivation while using manipulative design.
These moves relocate system design failures into the learner.
Self-regulation is real, but its opportunity structure is environmental.
A strong institutional application should ask both: what can the learner regulate, and what is the system unnecessarily forcing them to regulate against?
Goal quality matters before regulation begins
A learner can regulate efficiently toward a poor goal.
They may optimise for completing pages rather than understanding, for a grade rather than durable knowledge, or for producing an assignment rather than becoming able to perform the underlying task.
SRL research often begins with goal setting because regulation needs a target. But goal content is not automatically supplied by the construct.
Educational systems influence which goals feel legitimate through assessment and reward.
This means self-regulation can reproduce bad incentives very effectively. A student who learns to game a narrow test may be highly strategic without developing the intended capability.
Context changes the meaning of the same behavior
Spending more time can indicate persistence, confusion or inefficient strategy. Seeking help can indicate adaptive regulation or premature dependence. Switching tasks can indicate flexibility or distraction.
SRL cannot therefore be inferred from surface behavior without context.
Learning analytics systems face this problem acutely. Clicks and pauses are observable; the regulatory purpose behind them often is not.
Models that infer states from trace data need validation against richer measures rather than treating digital exhaust as transparent cognition.
Regulation develops through participation
Children and novices often regulate with others before they regulate independently.
Teachers structure time, label strategies, model monitoring and help interpret feedback. Peers remind one another of goals. Over time, some of these supports can be internalized.
This developmental perspective makes co-regulation more than a temporary convenience. It can be part of how self-regulation is learned.
It also warns against premature autonomy. Removing scaffolding too early can produce failure that is misread as low self-regulation.
Collective regulation introduces power
Groups are not neutral containers.
Whose definition of the goal counts? Whose uncertainty is heard? Who is allowed to slow the group down? Who performs the emotional labor of maintaining collaboration?
Socially shared regulation research creates a vocabulary for group-level processes, but institutional application must also examine hierarchy and participation.
A team can appear coordinated because dissent has been suppressed.
The Institute should therefore connect shared regulation with governance questions rather than treating group synchrony as inherently good.
AI can scaffold regulation or silently appropriate it
A tool that reminds a learner to revisit weak material may support regulation. A tool that automatically chooses every next step may reduce the need to practice planning. An assistant that asks a learner to attempt an explanation before giving one may preserve monitoring. Another that instantly completes the task may optimize production instead of learning.
The distinction is not "AI versus no AI."
It is which regulatory operation is being supported, which is being delegated, and what capacity is meant to remain afterward.
That is a tractable research agenda because each operation can be specified and measured.
SRL interventions are bundles, which complicates causal claims
Many educational interventions combine goal setting, strategy instruction, reflection, feedback and motivational support. If outcomes improve, the package may be useful while the active ingredient remains uncertain.
This matters when programmes claim to train a single capacity. A multi-component intervention cannot automatically identify which component caused the change.
Dismantling studies, process measures and mediation analyses can help, but each brings assumptions.
The Institute should therefore report interventions at the level the evidence supports: a package improved specified outcomes under specified conditions, rather than "self-regulation was increased" unless the construct was directly measured well.
Regulation can fail because the task is badly designed
When instructions are ambiguous, feedback is late, goals conflict and workload exceeds available time, self-regulation demands rise sharply.
Teaching learners to compensate can be useful, but redesigning the task may be more effective.
This creates an important institutional distinction between capacity building and friction removal. Not every friction should be removed, because some difficulty supports learning. Not every friction should be internalized as a personal resilience challenge either.
The design question is which difficulties are pedagogically productive and which are merely waste.
Long-term autonomy is a stronger outcome than short-term compliance
A learner who follows prompts perfectly may look highly regulated while the prompts are present.
A stronger test is whether they can later set goals, notice problems and choose strategies without the same scaffolding.
For AI-supported learning, this makes fade-out and transfer conditions especially important. Support quality should be judged partly by what remains when support is reduced.
What the evidence does not show
Self-regulated learning is not simply willpower or conscientiousness.
It is not identical to metacognition, executive function, motivation or study skills, though it includes or interacts with each.
A self-report questionnaire does not fully capture online regulation.
Successful regulation in one subject does not guarantee transfer to another.
More autonomy does not automatically produce better learning if learners lack knowledge, support or feedback.
Socially shared regulation is not just individual self-regulation multiplied by group size.
And emerging AI-SRL research does not yet show that automated regulation support universally improves long-term independent capacity.
Open questions
How should researchers combine self-report, trace and observational measures without creating data-rich but theory-poor models? Which SRL components transfer across domains? When does co-regulation become internalized rather than remain dependency? How do culture and educational systems shape what counts as effective regulation?
AI introduces a particularly consequential question: if a tool becomes very good at planning, monitoring and feedback, which regulatory capacities should learners still practice themselves?
That answer may differ for novices and experts, for temporary scaffolding and permanent assistance, and for learning goals versus productivity goals.
Learning is regulated in a world
The phrase self-regulated learning can sound solitary.
The evidence points to something richer.
A learner brings cognition, confidence, emotion, goals and strategies. A classroom brings tasks, feedback, norms and other people. A digital environment brings prompts, friction and automation. Regulation emerges from their interaction.
At the individual level, SRL helps explain how learning is directed over time.
At the group level, shared regulation shows how learning becomes coordinated.
At the institutional level, design determines whether good regulation is possible or constantly required as compensation.
The purpose is not to manufacture perfectly self-managing people.
It is to understand how humans become active participants in their own learning without pretending they learn alone.
Research references
- Panadero E. A Review of Self-regulated Learning: Six Models and Four Directions for Research. Frontiers in Psychology. 2017;8:422. https://doi.org/10.3389/fpsyg.2017.00422
- Dörrenbächer-Ulrich L, Bregulla M. The Relationship Between Self-Regulated Learning and Executive Functions: a Systematic Review. Educational Psychology Review. 2024;36:95. https://doi.org/10.1007/s10648-024-09932-8
- Dörrenbächer-Ulrich L, Dilhuit S, Perels F. Investigating the relationship between self-regulated learning, metacognition, and executive functions by focusing on academic transition phases: a systematic review. Current Psychology. 2024;43:16045-16072. https://doi.org/10.1007/s12144-023-05551-8
- Panadero E, Järvelä S. Socially Shared Regulation of Learning: A Review. European Psychologist. 2015;20(3):190-203. https://doi.org/10.1027/1016-9040/a000226
- Järvelä S, Nguyen A, Hadwin A. Human and artificial intelligence collaboration for socially shared regulation in learning. British Journal of Educational Technology. 2023;54:1057-1076. https://doi.org/10.1111/bjet.13325
- Veenman MVJ, van Hout-Wolters BHAM, Afflerbach P. Metacognition and learning: conceptual and methodological considerations. Metacognition and Learning. 2006;1:3-14. https://doi.org/10.1007/s11409-006-6893-0
- Fleming SM. Metacognition and Confidence: A Review and Synthesis. Annual Review of Psychology. 2024;75:241-268. https://doi.org/10.1146/annurev-psych-022423-032425