It is important for learners to engage in self-regulated learning (SRL), as it predicts academic achievement in a wide range of disciplines (Schunk & Greene, 2018). Even after graduation, self-regulation is important in lifelong learning and workplace performance (Schunk & Greene, 2018). However, SRL is difficult to enact and requires instruction to learn (Perry et al., 2008). Therefore, a longstanding line of research investigates how to scaffold learners’ SRL (Zimmerman, 1986). A number of special issues have been dedicated to scaffolding SRL (Azevedo, 2007; Azevedo & Hadwin, 2005; Azevedo & Jacobson, 2008). Given the ever-changing context of education, such as technological advancements, and the continuing need for effective learning support, we proposed a new special issue. This special issue aims to inform researchers and educators about current approaches to scaffolding SRL. We pay special attention to how scaffolds are designed to inform future research and educational interventions.

SRL is a goal-directed process in which the learner monitors and controls their behaviour to attain specific goals (Winne, 2018). There are different models of SRL (Panadero, 2017), and the contributions to this special issue also base their work on different but related conceptualisations of SRL. Learners capable of self-regulating their learning tend to show improved learning outcomes (Bannert et al., 2015; Deekens et al., 2018). However, students tend to struggle with SRL (Azevedo et al., 2010), suggesting a need to support the use of effective SRL during learning. One way to support students in SRL during learning is via scaffolds (Azevedo & Hadwin, 2005). Relevant studies have reported beneficial effects of scaffolding on the learning process, such as increased metacognitive activities (Engelmann & Bannert, 2021), and learning outcomes, such as a higher score on a transfer test (Leelawong & Biswas, 2008).

The conceptualization of scaffolds and the implementation of a scaffolding mechanism can play a role in the effectiveness of those scaffolds on the learning process and learning outcomes. However, there is little systematic research into the development of scaffolds and how scaffold design decisions are associated with the execution of self-regulation and learning outcomes (Devolder et al., 2012). Meanwhile, there is still a need for effective educational materials to support metacognition and self-regulation within different school levels and educational environments. In this special issue, we collected studies describing how scaffolds can be designed, the potential for implementing scaffolds in different learning contexts, and the effectiveness of scaffolds in supporting SRL.

Scaffolds come in all shapes and sizes. For example, scaffolding and feedback have been implemented in different advanced learning technologies (e.g., Azevedo et al., 2022) or classroom-based studies (e.g., van de Pol et al., 2015). Overall, scaffolds of SRL show a beneficial effect on learning outcomes (Wong et al., 2019; L. Zheng, 2016). Scaffolds can be designed for each phase of SRL (Roll & Winne, 2015). Wong et al., (2019) categorise scaffolds as prompts, feedback, integrated support systems, and other approaches in the context of Massive Open Online Courses (MOOCs). A prompt is a notification that appears during learning, for example, as a pop-up. Feedback can be a question or suggestion to encourage the use of SRL, or both. Integrated support systems can include prompts and feedback, but other tools as well, to support SRL. Azevedo and Jacobson (2008) identify three dimensions in which scaffolds can differ: what to scaffold (which learning process), when and how to scaffold (timing and availability of scaffold), and for whom or how to scaffold (delivery of scaffolds). Numerous empirical studies have studied these relevant aspects of scaffolds (e.g., a special issue by Azevedo & Hadwin, 2005). However, it often remains unclear how decisions were made to choose how to design these (aspects of) scaffolds. It is crucial to understand how the design of scaffolds affects SRL because studies have documented a wide range of effects, including more metacognitive events during learning and better integration of evaluation processes with other learning processes (Engelmann & Bannert, 2021), and improved learning outcomes (Molenaar et al., 2012).

This special issue includes seven studies that have either investigated the effectiveness of implementing scaffolding in various learning contexts or present implications for the design of scaffolding in learning contexts. We describe the key differences and similarities using the SMA (Self-regulated learning processes, Multimodal data, and Analysis)-grid (Molenaar et al., 2023) and the what, when, how, and by whom or what of the use of scaffolding (Azevedo & Jacobson, 2008) across the contributions. The SMA-grid grid describes the characteristics of each study (shown in different table columns or axes) related to how they addressed scaffolding and SRL. Axis (or column) categorizes SRL as CAMM processes (Cognition, Affect, Metacognition, and Motivation). Axis 2 is a categorization of the data stream(s) or channels used, roughly distinguishing physiological data (such as heart rate), behavioral data (such as log data), and contextual data (such as video recordings). In addition, analytical approaches are labelled as either addressing one CAMM process with one data stream (unimodal), multiple CAMM processes with one data stream (horizontal), one CAMM process with multiple data streams (vertical), or multiple CAMM processes with multiple data streams (integrated). Regarding the what, when, how, and by whom or what of scaffolding (Azevedo & Jacobson, 2008), what refers to what to scaffold, when refers to when to provide a scaffold, how refers to the delivery of the scaffold, and by whom or what refers to the delivery mechanism(s).

In the first contribution, learners were scaffolded via four pedagogical agents(Dever et al., this issue). In the experimental condition, these agents were triggered based on rules that consider learners’ actions, such as time spent reading, and more general rules, such as more intense scaffolding at the beginning. The control condition had access to the agents, but the agents were not triggered. The experimental condition showed higher frequencies of SRL activities, less repetitive SRL activities, and higher learning outcomes.

The second contribution addressed how students use a pre-structured self-assessment tool, namely a diagram, and scaffolds, cueing self-assessment, to adapt their learning (Pijeira-Díaz et al., this issue). Three conditions were compared: 1) empty diagram, 2) pre-structured diagram, and 3) pre-structured diagram with self-assessment cues. Learners fill in the diagrams after reading texts, and after the diagrams, they judged their comprehension, restudied, and took a test. No differences were found between conditions regarding the cues they used to monitor comprehension, but diagramming conditions showed higher metacomprehension accuracy.

The third contribution used a person-centered approach to identify subgroups of scaffold users (J. Zheng et al., this issue). Scaffolds addressed conceptual understanding, such as helping learners to organize information; strategy execution, such as guiding learners to solve the problem; and metacognition, such as reflecting on problem-solving processes and solutions. Four subgroups were identified based on how often and with which type of scaffold was interacted. The metacognitive and strategic subgroups solved more problems correctly, and the metacognitive subgroup had better efficacy and confidence scores than the other subgroups (moderate and low scaffold use).

The fourth contribution identifies potential advantages and disadvantages for both event-based and time-based scaffolds based on data from a self-reflective writing task (Taub et al., this issue). Four clusters of students were identified that differed in terms of what they wrote about which stage of SRL (planning, performance, and reflection). The anti-planner cluster produced more than the performer cluster, and the planner cluster revised more than the anti-planner cluster. No differences were found in pausing.

The fifth contribution showed the effectiveness of digital skill training, addressing SRL, on academic achievement, and moderating effects were studied to investigate why skill training is effective (Bernacki et al., this issue). They redesigned a video-based multimedia training. Similar to the first design, the redesign showed beneficial effects on learning. Specifically, learners’ knowledge about behavioral and environmental regulation demonstrated during the training predicted biology exam performance. Cognitive and metacognitive knowledge showed nonsignificant, positive relations with exam performance.

The sixth contribution described the design of five scaffolds during a read-and-write task (Van der Graaf et al., 2023). Scaffold purposes were determined based on the stages of SRL and their timing based on data from a previous study. Scaffold content was personalized based on learners’ preceding SRL activities. The amount of personalization differed across scaffolds. Compared to no and non-personalized scaffolds, personalized scaffolds showed higher compliance, which was linked to better essay scores.

The seventh contribution presented a digital learning companion to learners after solving textbook problems (Sijmkens et al., 2023). The digital learning companion scaffolds learners’ reflection on how they solved the problems by providing multiple-choice reflection prompts. The prompts aimed to elicit metacognitive activities, and the companion provided feedback on learners’ reflection. Learners who used the prompts frequently showed a better strategic problem-solving approach and a larger conceptual understanding. Self-reported effort regulation was positively related to both strategic problem-solving and conceptual understanding.

The purpose of this special issue is to highlight current research investigating scaffolding within an SRL framework. We noted several similarities and differences between the papers from compiling the contributions, as highlighted in Table 1. We discuss these further below. The clear similarities between these papers are the focus on Cognitive and Metacognitive self-regulatory processes (i.e., the C and M of CAMM), assessing SRL with one data stream, the use of log-file data to capture student behaviors, and the samples of University students, except for one study with secondary students (Pijeira-Díaz et al., this issue). However, although they appear similar, these all apply to different contexts. For example, cognitive and metacognitive processes were examined in terms of engaging diagnostic reasoning (Zheng et al., this issue), writing (Taub et al., this issue), metacomprehension tasks (Pijeira-Díaz et al., this issue), digital skill training (Bernacki et al., this issue), interactions with pedagogical agents (Dever et al., this issue), learning in a digital environment (Van der Graaf et al., 2023), and strategic problem-solving (Sijmkens et al., 2023). In addition, all studies used log data but at different levels of granularity (e.g., actions at the millisecond level, keystroke logging). Furthermore, University students were in specific or general (Van der Graaf et al., 2023) courses, such as biology (Bernacki et al.; Dever et al., this issue), writing (Taub et al., this issue), or medical school (Zheng et al., this issue).

Table 1 The six contributions to the current special issue and their scaffold characteristics
Full size table

Regarding the process of designing the scaffolds, the authors in each study designed a unique way of scaffolding learners’ SRL. A similarity is that overall, the studies show beneficial effects of scaffolding SRL, in line with previous studies (e.g., Wong et al., 2019; Zheng, 2016). Taken together, this suggests different ways to support SRL effectively. The finding is not new, but we can investigate how the authors came about their designs in the current special issue. There are several considerations to be made when designing scaffolds. Are the scaffolds mostly based on theory (Bernacki et al., this issue), or are they also data-driven (Van der Graaf et al., 2023)? When are scaffolds triggered: is it an event- or time-based approach, or both (Taub et al., this issue)? Do scaffolds have a single purpose (Pijeira-Díaz et al., this issue), or does each have a unique purpose (Zheng et al., this issue)? Are scaffolds expected to affect the sequential nature of SRL (Dever et al., this issue)? To what extent can and should scaffolds attenuate effects of prior metacognitive abilities (Sijmkens, et al., 2023)?

Instead of providing concrete guidelines, the contributions to the special issue can offer suggestions on how the design process of SRL scaffolds can be shaped. This special issue lists different types of scaffolds, which were designed with different decisions made during the process. The papers in this issue can be used to inform the design process of new scaffolds. It can be assumed that with technological advances and the everchanging educational context, new (types of) scaffolds are desired. Therefore, a focus on the design process, as in this special issue, is helpful in future designs of scaffolds. It should be noted that there are different ways to create effective and efficient scaffolds, similar to how SRL can be executed differently to attain a specific learning goal. Future designers can read the current special issue for inspiration on scaffolds and the decisions underlying their design.