Introduction

Digital technologies have not yet fully realized their potential in education, although the findings of meta-analyses show that their implementation into the classroom can have a positive impact on student achievement (e.g., Chauhan, 2017; Higgins et al., 2019; Hillmayr et al., 2020; Q. Li & Ma, 2010; Moran et al., 2008; Tamim et al., 2011). It is commonly agreed upon that this impact on student achievement is not due to digital technology itself, but rather due to specific digitally enriched instructional features (e.g., adaptive feedback, prompting, scaffolding) implemented into those technologies and the specific use teachers and students make of these features (see the media debate, started by Clark, 1994; Kozma, 1994). Still, research often focuses on the effects of the mere presence of digital tools on learning outcomes (Kucirkova, 2014).

When inspecting the cause of such positive impact of digital tools in educational contexts, however, there may not be one single satisfactory answer (Hillmayr et al., 2020): not only does the term ‘digital tool’ encompass a broad variety of different types of software and hardware that can be used in very different ways, research also has demonstrated that different types of digital tools can have specific benefits (Chi & Wylie, 2014; Hillmayr et al., 2020; Koedinger et al., 2012; VanLehn, 2011). That is one reason why measuring learning outcomes of digitally enriched instruction (i.e., instruction that is enriched by digital technology) and comparing these to other types of instruction does not suffice to theoretically underpin the learning mechanisms at play when students utilize technology during learning.

Therefore, research on why learning with digital tools works should refer to theories on cognition during learning, should be specific about the assumed underlying learning activities which the digitally enriched instructional features implemented aim to enhance, and should account for the characteristics of the content. When analyzing such features of digitally enriched learning environments, it is worthwhile to consider that their positive effect on student learning may be rooted in very different psychological constructs. For example, while metacognitive scaffolds inside an online math tutor may help students regulate their learning, simulations inside a digital physics learning environment may guide students into reasoning about the physical phenomenon. Consequently, we suggest to focus not only on learning outcomes, but also on the specific effects of features implemented in digital tools on students’ learning activities and their variation with different learning contexts and content. To draw attention to this connection we use the term “(digitally enriched) instructional features” for a broad variety of different components and functionalities of software (such as, scaffolds, adaptive feedback, simulations of phenomena, prompts), designed and implemented to stimulate and aid learning activities.

In this paper, we present a framework which has the purpose of guiding theoretical analyses and empirical research that aim to explain learning in digitally enriched educational settings via instructional features and learning activities. For this, we draw on three frameworks or models which generally capture the structure and dynamics of learning in such settings. First, we refer to the knowledge-learning-instruction framework by Koedinger and colleagues (KLI; Koedinger et al., 2012) and elaborate how this framework is well suited to describe learning with digital tools by distinguishing between external instructional events and assessment events on the one hand, and internal learning processes and knowledge components on the other hand. Second, the SOI-Model (Selection-Organization-Integration) defines the internal cognitive core processes during learning (Mayer, 2014). According to the model, the learner actively constructs knowledge representations in the working memory using both incoming material from the environment and prior knowledge in the long-term memory. This active construction includes selection: identification of useful information, organization: understanding how the single information interact with each other, and integration: relating new information with prior knowledge. We further argue that an appropriate approach to capture learning should account for the systematic distinction between digitally enriched learning opportunities and students’ actual learning activities during their use—since these may vary substantially among students. Such an approach is described by so-called utilization-of-learning-opportunity frameworks from classroom research (ULO; Seidel, 2014; see also Brühwiler & Blatchford, 2011). ULO frameworks distinguish explicitly between potentially used learning opportunities on the one hand (e.g., educational settings that have the potential for cognitive activation—a core dimension of instructional quality representing offers of challenging tasks and stimulating ways of interacting with the subject matter and with the peers; see Praetorius et al., 2018) and actual student engagement on the other hand (i.e., the mental effort learners invest in genuine knowledge construction processes—germane processing; Sweller et al., 2019; see Nückles, 2021, for a discussion).

Since the perspectives of these frameworks appear highly relevant for empirically investigating and theoretically explaining learning in digitally enriched educational settings, we introduce the CoDiL framework—a Cognitive process framework on the Learning mechanisms of Digital tools in educational settings that incorporates both the core ideas of the KLI framework, and the ULO framework—i.e., core ideas from empirical classroom research, cognitive psychology, and subject-specific as well as content-specific education research (Fig. 1). For an analogous framework for composite instructional designs (i.e., instructional designs with multiple phases) see Loibl et al. (2023).

Fig. 1
figure 1

The CoDiL framework, a cognitive process framework on the learning mechanisms of digital tools in educational settings. The dashed separates the external situation and external behavior on the left, and the internal person characteristics and cognitive processes on the right. The structure of the utilization-of-learning opportunity framework (ULO) is represented by the dashed boxes. The structure of the knowledge-learning-instruction framework (KLI) is indicated by the three columns

Full size image

In the context of the CoDiL framework, instruction is understood as a digitally enriched educational setting (see the “Acquiring Knowledge Components in (Digitally Enriched) Educational Settings” section), in which the used digital tools can be characterized by their implemented digitally enriched instructional features (see the “Digital Instruction—Instructional Features as Opportunities Stimulating Learning Activities” section). If utilized by the student these features (which offer learning opportunities) are considered to initiate and guide learning activities (see the “Learning Activities When Learning With Digital Tools” section). These learning activities have a behavioral side (i.e., external, and directly observable, e.g., interacting with the digital tool, solution formulation, or written self-explanations; see the “Evaluation—Linking Students’ On-Task Behavior and Learning” section) and a cognitive side (internal and latent, e.g., adaption, discrimination, or revision; see the “Learning Activities When Learning With Digital Tools” section). In interaction with prior knowledge—and mediated by cognitive processes (i.e., selecting, organizing, integrating)—the learning activities cause evolving knowledge during learning—leading to final knowledge at the end of the learning process.

By representing a high-level conceptual integration of different theories, the CoDiL framework serves to systematically describe (see the “Learning as Individual Content-Specific Cognitive Processes to Generate Knowledge” section), develop (see the “Development—Designing Digitally Enriched Educational Settings” section), and evaluate (see the “Evaluation—Linking Students’ On-Task Behavior and Learning” section) research on digitally enriched instruction: we show how it can be utilized to theoretically ground the design of digital tools for educational settings, why it is suitable to frame correlational or causal research endeavors, and how it can guide the evaluation of students’ learning activities in digitally enriched educational settings based on appropriate operationalizations of student-tool interactions. We exemplify the applicability of our framework by framing recent studies regarding the effect of learning with digital tools with regard to our framework.

Learning as Individual Content-Specific Cognitive Processes to Generate Knowledge

Acquiring Knowledge Components in (Digitally Enriched) Educational Settings

In line with established learning theories, we understand learning as an active, constructive, and content-specific cognitive process (Koedinger et al., 2012; Mayer, 2014). Consequently, a framework which aims at explaining learning should explicate on the one hand external, observable, and manipulable elements—e.g., instructional events and observable student-tool interactions—and on the other hand model internal and unobservable learning processes and learning outcomes—e.g., knowledge components (Koedinger et al., 2012; Yeo & Fazio, 2019).

Knowledge components are connected in multiple ways to other elements in the CoDiL framework. Similar to the construct of schemata (Schweppe & Rummer, 2014), Koedinger and colleagues (2012) “define a knowledge component … as an acquired unit of cognitive function or structure that can be inferred from performance on a set of related tasks” (p. 764). The knowledge components define the structure of the final knowledge, but they also specify the prior knowledge that can be activated during learning and the evolving knowledge (including incomplete or erroneous knowledge components) that develops throughout the learning process. Since the knowledge components available at any point in time influence the learning activities, a content-specific analysis of these knowledge components is a fundamental basis for understanding the learning mechanisms at play, and thus a fundamental basis for the design and development of any educational environment that aims to achieve this knowledge acquisition. In particular, we emphasize the argument that a subject-specific analysis of the content (as part of, e.g., adaptive control of thought-rational, Anderson et al., 1997; construction-integration model, Kintsch, 1991; cognitive modeling, Ritter et al., 2007) is relevant for the development of digitally enriched educational settings and the design of digital tools.

Consider, for example, students learning how to compare the size of fractions, such as 2/3 and 4/5. Here, decades of theoretical and empirical research draw a very clear picture of the prior knowledge (natural number concepts, the part-whole concept, and the concept of fraction equivalence; Post & Cramer, 1987) influencing the students’ proficiency, various faulty, or still evolving knowledge aspects (isolated comparisons of numerator or denominator as natural numbers in different forms; Gómez & Dartnell, 2019), and the final knowledge (repertoire of various correct comparison strategies; Clarke & Roche, 2009). These different knowledge components can be used to design an explanatory cognitive model of how students learn to compare fractions: given the influence of different kinds of prior knowledge, the pathway to a desirable change in knowledge may be very different for students who overgeneralize natural number concepts, or students who already have acquired the part-whole concept. Given such cognitive student models, content-specific educational settings (as opportunities for learning) can be designed to initiate and stimulate specific learning activities (“Learning Activities When Learning With Digital Tools” section) that may lead to the final knowledge.

In general, content-specific prior knowledge should be considered relevant when discussing educational settings (Kalyuga, 2007; Simonsmeier et al., 2022), since every learner has some domain-general prior experiences, domain-specific knowledge facets, or an attitude or interest toward the learning domain (Tobias, 1994). Prior knowledge is the key predictor of how information is selected, organized, and consequently, how evolving knowledge arises (Kalyuga, 2013 and Simonsmeier et al., 2022 for a meta-analysis). This evolving knowledge is further elaborated and consistently integrated into already existing knowledge structures (Mayer, 2014). Consequently, all knowledge components have to be carefully considered when discussing the learning process from the initial start of the learning phase and the individual perquisites of the learner to the processing of new information that gets integrated into existing knowledge and constantly updated (i.e., evolving knowledge) until the final knowledge is reached. Here, cognitive (content-specific) student models may serve as guidelines.

Learning Activities When Learning With Digital Tools

Yet, building the design of digitally enriched learning environments on such rich cognitive models of student knowledge does not guarantee learning success for each student. Therefore, the utilization-of-learning-opportunity framework (ULO) seeks to integrate structural aspects of learning environments and actual processes that students undergo during instruction (Seidel, 2014; see also Brühwiler & Blatchford, 2011). The ULO framework is commonly used as a domain-general framework in educational effectiveness research and focuses on teacher-student interaction on a classroom level, describing why students differ in how they succeed in learning scenarios (Alp Christ et al., 2022; Seidel & Shavelson, 2007). This domain-general description works very well for domain-general measures of how students utilize learning opportunities, such as classroom engagement (Fredricks et al., 2004; Guertin et al., 2007; Henrie et al., 2015; Huang et al., 2019; Lo & Hyland, 2007). In its usual application, the main argument of the ULO framework is that for learning to be effective (in terms of students demonstrating high learning outcomes), students have to actively engage in the learning opportunities offered. This underpins that learning is considered to be an active and generative process—in line with generative learning theory (Fiorella & Mayer, 2016; Roelle & Nückles, 2019) and theories about self-regulation (Azevedo, 2020; S. Li et al., 2022; Molenaar et al., 2023).

However, for disentangling each student’s utilization of learning opportunities provided by digital tools, the conceptualization of learning mechanisms needs to focus on individual learning activities and to be content-specific (e.g., Reinhold et al., 2020a). For that, the CoDiL framework describes learning opportunities in digitally enriched educational settings, which are characterized via specifically designed digitally enriched instructional features (“Digital Instruction—Instructional Features as Opportunities Stimulating Learning Activities” section). These instructional features may initiate and stimulate learning activities. We understand such learning activities in digitally enriched educational settings as (1) activities that students do (or do not) engage in when working within the digitally enriched educational setting and as (2) theoretical predictors of learning outcomes. Moreover, the CoDiL framework differentiates between the external, behavioral side and the internal, cognitive side of a learning activity. The external side of a learning activity encompasses specific student-tool interactions that occur when students work with the implemented digitally enriched instructional features; the internal side of a learning activity encompasses the non-observable mental processes initiating and accompanying the external student actions. Among others, such learning activities relevant for digitally enriched educational settings are the following:

  • Abstraction: detachment from example-bound or contextualized knowledge to a context-free, generalizable level (Arnon et al., 2014; Lehtinen & Repo, 1996; Renkl, 2023; Rittle-Johnson & Star, 2011).

  • Conducting experiments: engagement in laboratory work to answer a question based on empirically obtained data (Hart et al., 2000; Lazonder & Harmsen, 2016; Wörner et al., 2022).

  • Creating examples: generation of different instances of a phenomenon for the purpose of experiencing structure, extending the range of variation, experiencing generality and the constraints and meanings of conventions, or extending example spaces and exploring boundaries (Guerrero et al., 2023; Watson & Mason, 2002).

  • Exploration: initial active, own examination of the learning objects to activate relevant prior knowledge and to raise questions (M. Lachner et al., 2022; Loibl et al., 2017).

  • Formulating hypothesis: formation and evaluation of theory to formulate a claim that can be confirmed or refuted following experimental scientific endeavor (Klahr & Dunbar, 1988; Park, 2006).

  • Refinement: improvement of the quality of knowledge by making it more accurate, appropriately general, or discriminating (Booth et al., 2017; Koedinger et al., 2012).

  • Revision: fundamental restructuring of the learners’ pre-instructional conceptual structures to allow understanding of the intended final knowledge (Chi, 2008; Duit & Treagust, 2003; Schroeder & Kucera, 2022; Vosniadou, 1994).

  • Self-explanation: explaining the new-to-learn content to oneself to achieve deepened processing of the learning materials (Bisra et al., 2018; Renkl et al., 1998).

The cognitive processes that are typically assumed to describe how these learning activities result in the acquisition of knowledge components are selecting, organizing, and integrating (Mayer, 1999) information to construct new knowledge—which is integrated into the already existing prior knowledge structure of the long-term memory. This integrated knowledge structure allows learners to apply the acquired knowledge in new situations. Internally, these cognitive processes lead to linking the new information to prior knowledge and thereby constructing new knowledge beyond the given information (Mayer, 1984).

Although such learning activities relating to concrete domain-specific content are of key interest for developing learning environments, they are often not specified in detail in educational research (Yeo & Fazio, 2019), especially when research focuses on learning with digital tools (Hillmayr et al., 2020; Kucirkova, 2014). Yet, we consider the understanding of these learning activities necessary to answer the question why digital tools work—and what makes some learners succeed while others fail. In context of the CoDiL framework, such learning activities function as a “link” between students’ external, behavioral interaction with the digital tool and students’ internal, cognitive processes that lead to knowledge acquisition (which allows for a theory-driven operationalization of cognitive activities via student-tool-interactions; see the “Evaluation—Linking Students’ On-Task Behavior and Learning” section). More specifically, the CoDiL framework understands students’ engagement in these learning activities as necessity for successful learning—in line with the ULO framework and generative learning theory (Fiorella & Mayer, 2016; Roelle & Nückles, 2019).

Digital Instruction—Instructional Features as Opportunities Stimulating Learning Activities

When referring to digital instruction, a decisive argument is the media debate initiated by Clark and Kozma (Clark, 1994; Kozma, 1994). They argue that on the one hand “media are mere vehicles that deliver instruction but do not influence student achievement” (Clark, 1994, p. 22), but on the other hand that certain media “possess particular characteristics that make them both more and less suitable for the accomplishment of certain kinds of learning tasks” (Kozma, 1994, p. 8). Such ‘particular characteristics’ have been broadly discussed and investigated since the 1980s, among them ways to direct student attention to learning goals (Scardamalia et al., 1989) or ways to promote students help-seeking behavior (Aleven et al., 2003)—to name just two. The CoDiL framework acknowledges these perspectives; we argue in line with the media debate that it is not the digital tool itself that initiates learning activities (i.e., Clark’s argument) but the implemented digitally enriched instructional features (i.e., Kozma’s argument—the ‘particular characteristics’).

Such instructional features can be regarded as the subject of affordances and constraints, which frame the activity patterns of students and thus the learning activities. Greeno (1998) in fact used the notion of affordances and constraints much broader, incorporating social and material dimensions of the learning situation; however, these notions have been also successfully introduced into media design with a focus on the features of the environment which support and guide cognition (e.g., Hartson, 2003; Norman, 1999).

In context of the CoDiL framework, instructional features are understood as (1) digitally enriched instructional events, which (2) have the potential to stimulate learning activities; due to them being (3) subject of affordances and constraints; they should be (4) designed in order to foster learning activities—i.e., to lower extraneous load (Sweller, 2020) and to increase generative processing (Schumacher & Stern, 2023) when compared to non-digitally enriched educational settings. This is in line with Cognitive Theory of Multimedia Learning (CTML; Mayer, 2014). We consider, among others, the following digitally enriched instructional features as essential:

  • Adaptive feedback (e.g., Aleven et al., 2017; Reinhold et al., 2020b) may stimulate refinement activities, as it may activate prior knowledge and aid discriminating one’s potentially too narrowly defined knowledge structures. It may initiate revision activities by revealing the limits of students’ prior knowledge, when students are not only confronted with the correct solution, but also with their most-likely misconception.

  • Opportunities to try and fail by posing problems prior to instruction (PS-I; e.g., Boomgaarden et al., 2023; Kapur, 2008; Loibl et al., 2017) may stimulate exploration activities. Here digitally enriched learning environments can aid these learning activities by leading students through the problem-solving process and address students’ failed attempts to make their misconceptions salient and other conceptions more reasonable (Holmes et al., 2014).

  • Prompting students (e.g., Rau et al., 2009; Rittle-Johnson et al., 2017) to use task-specific hints or reconsider specific well-known strategies while solving tasks may stimulate refinement activities—as students may engage in overt elaboration (Fiorella & Mayer, 2016; Weinstein & Mayer, 1986). While elaboration has been shown to foster sense-making and retention, learners often do not spontaneously engage in elaboration processes, but need external guidance, for instance, by prompts, to engage in such processes (Berthold et al., 2007; Chi et al., 1989; Endres et al., 2017; Nückles et al., 2009).

  • Prompting students to summarize and explain central instructional information or draw conclusions from simulations (e.g., Hofer et al., 2018; Reinhold et al., 2020a) may stimulate self-explanation activities, aiding the promotion of active construction of conceptual knowledge.

  • Contrasting cases (e.g., Ma et al., 2023; Schalk et al., 2018) may stimulate revision activities, since learners are encouraged to make connections, identify patterns, and develop their own insights about the topic by comparing two or more examples of a particular concept, issue, or phenomenon—revealing limits of one’s own prior knowledge. Dynamic contrasting cases may make the relevant features more salient.

Of course, some of these (digitally enriched) instructional features can be included in classical (i.e., non-digital) classroom instruction as well, following agreed upon principles of instructional design (Gagne et al., 2005; Merrill, 2001)—yet, the implementation in digitally enriched environments might be particularly promising since affordances of media-based instructions come into play (Mayer, 2014; Sweller, 2020).

Research on Digitally Enriched Educational Settings Utilizing the CoDiL Framework

As a synthesis of the above-mentioned different theoretical frameworks, the CoDiL framework can serve as a holistic framework guiding both the development of digitally enriched educational settings and research thereof. More precisely, structuring research within the CoDiL framework (in the development of the digital tools and the research strategy) aligns research endeavors with (1) theories of cognition during learning of (2) a specific content. It theoretically focuses such endeavors on (3) the assumed underlying learning activities which the (4) digitally enriched instructional features should enhance. Thus, it may help researchers to establish (5) a link between students’ cognition and their on-task behavior which allows for testing causal effects of why specific features of digital tools are beneficial for learning.

Development—Designing Digitally Enriched Educational Settings

In line with generative learning theory (Fiorella & Mayer, 2016; Roelle & Nückles, 2019) and theories about self-regulation (Azevedo, 2020; S. Li et al., 2022; Molenaar et al., 2023), the CoDiL framework highlights the mediating role of students’ engagement in specific learning activities in the cause-and-effect mechanisms of successful learning with digital tools. That is, developing specific digital tools for educational purposes includes the design of digitally enriched instructional features which bear the potential to stimulate learning activities which—mediated by cognitive processes resulting from those learning activities—lead to knowledge acquisition (Fig. 1). By combining and linking content-specific theories of learning with theories of what stimulates relevant learning activities, the CoDiL framework aims at highlighting the following design principles: in order to find the appropriate to-be-implemented instructional features, knowledge about content-specific learner models (i.e., knowledge components and learning activities) should be used to design educational settings (i.e., implement instructional features in the digital tool) that are well-suited to stimulate the relevant generative cognitive processes. The rather broad synthetic structure of the CoDiL framework may serve as a guideline for finding an appropriate theoretical foundation and may illustrate aspects to consider in developmental steps. This may be demanding as it can vary between domains, content, and learning goals, as described in the following different examples, which highlight different aspects of this endeavor:

  • Consider developing an online math tutor to support students in learning fraction arithmetic. Keeping the relevant theoretical mediator for learning success in mind (i.e., engagement in abstraction and refinement activities), digitally enriched instructional features selected to be implemented should bear the potential to leverage students’ knowledge about how to operate with fractions (i.e., to come to more accurate and faster solutions when applying arithmetic operations and to appropriately generate conceptual understanding of those operations) during learning with the tool. For that, adaptive feedback and an adaptive increase in task difficulty regarding difficulty-generating factors may seem appropriate (Reinhold et al., 2020b).

  • In contrast, when developing an appropriate digitally enriched educational setting for students to learn about a physical phenomenon, digitally enriched instructional features should support the a-priori identified purposeful reasoning processes that may establish relevant knowledge facets: to stimulate experimental activities (e.g., formulating hypothesis, conducting experiments) leading to a conceptual understanding of the “control of variables strategy,” a simulation enabling that strategy in a salient way may be an appropriate digitally enriched instructional feature (e.g., Greiff et al., 2015; see also scientific discovery as dual search, Künsting et al., 2011).

Evaluation—Linking Students’ On-Task Behavior and Learning

Studies in the field of digitally enriched learning often focus on the direct path, i.e., the effect of digital tools on learning outcome when compared to another learning scenario—which allows only limited interpretation and explanation of changes in learning outcomes. Following and combining theories of learning (e.g., Fiorella & Mayer, 2016; Mayer, 2014), the CoDiL framework is based on the identical premises that learning activities (and the cognitive processes they result in) during (digital) learning—stimulated and triggered by constraints and affordances that go along with specific digitally enriched instructional features—are what contributes to the final learning outcome. For this reason, we consider the external, behavioral side of students’ learning activities as theoretical as well as empirical mediators of achievement.

In this section, we illustrate how the relevant theories about linking behavioral data to student cognition inform the CoDiL framework in how to investigate these activities directly (and as unobtrusively as possible) during learning: following up on the argument that implemented instructional features in digital tools can stimulate learning activities, students’ interaction with these features (external, behavioral side of the learning activities, implying any process data indicators for student-tool-interaction that can be recorded, e.g., click behavior, or writing-to-learn text prompts, to name two) can be a valid indicator for the cognitive side of learning activities (internal, latent, and thought to result in knowledge acquisition, mediated by cognitive processes).

There is a broad variety of different conceptual ideas for how to establish such a link between behavior and cognition on a theoretical and empirical level (Goldhammer et al., 2021; Greiff et al., 2015; Huber & Bannert, 2023; Mislevy et al., 2012; Molenaar et al., 2023; Sedrakyan et al., 2020). For the CoDiL framework to serve as a widely open framework aiming at causal interpretation of learning effects, we aimed at compatibility with a large variety of different analytical approaches. This is why we consider the term “process data” (as a broader term than “log data”) as an umbrella term encompassing (i) student-tool interactions logged unobtrusively by the digital tool itself and (ii) other process indicators obtained from accompanying assessments during learning, such as, finger-tracing data in open-learning environments (e.g., Moyer-Packenham et al., 2019; Zuo & Lin, 2022), solutions from rather closed unique cognitive items inside a digital learning path (e.g., Boomgaarden et al., 2023; Rau et al., 2017), eye-tracking data (Nückles, 2021; Strohmaier et al., 2020), think-aloud protocols accompanying the use of any kind of digital tool (Ericsson & Simon, 1998; A. Lachner & Nückles, 2015; Renkl, 1997), and journal writing during self-regulated learning with digital tools (Nückles et al., 2020). To sum up, the CoDiL framework emphasizes widely agreed-upon necessities for establishing a link between process data and hypothesized cognition (Goldhammer et al., 2021; Huber & Bannert, 2023; Molenaar et al., 2023; Sedrakyan et al., 2020)—i.e., the need for (a) a solid theoretical foundation, as well as (b) appropriate indicators inside the process data files—to allow for differentiated statements about the effects of digital tools (or more specifically, their implemented instructional features) on learning success. For example,

  • Consider Lalley and colleagues’ (2010) study on the comparison of virtual vs. real frog dissection in biology classrooms. The theoretical framework takes into account ethical and health issues, as well as specimen decay—underlining the necessity of virtual alternatives to reach the same learning goals and asking whether “virtual dissection procedure [result in] comparable learning … outcomes when compared to traditional dissection” (p. 191). While the authors could show a positive effect of the digital learning tool provided compared to real dissection, this effect cannot be further evaluated within the theoretical framework of the study, as no hypotheses about the digitally enriched instructional features implemented in the software and their relation to learning activities during the virtual dissection (compared to the real dissection) are stated. Thus, the research framework lacks a theoretical mediator that could explain why the virtual approach is more effective than the traditional approach—leaving essential questions unanswered.

  • In clear contrast, the study by Koedinger and colleagues (2010) is prototypical for the relevant elements of the CoDiL framework. Here, students who use a formative assessment mathematics tool (and students who do not) were compared in terms of learning outcomes measured via a standardized mathematics test. The theoretical framework (a) encompasses well-established effects of adaptive and individualized learning in the context of formative assessment, stating explicitly that “students would benefit from the tutoring, feedback, and [the] design of the … system” (p. 496)—routed down to estimated positive effects of “practice with timely feedback” and “individualized tutorial assistance” (p. 496). Thus, adaptive immediate feedback and individualized tools (which would be considered digitally enriched instructional features providing additional learning opportunities in terms of the CoDiL framework)—and not digital tool use in general—are regarded as the cause for the effect on learning. From the perspective of the CoDiL framework, these instructional features may stimulate refinement or revision activities in students (i.e., the theoretical mediator of the effect of the digital tool). Yet, for them to become predictors of achievement, it is relevant if (or to which amount) students make use of these digitally enriched instructional features—shifting from design features of the digital tool to the above-mentioned specific learning activities students actually engage in while using the digitally enriched learning tool. Koedinger et al. (2010) assessed engagement in these learning activities by behavioral measures of student-tool interaction; (b) students who did not complete a certain amount of items in the tool were labeled “low usage” and students who did were labeled “high usage”—shedding light on an interaction effect in line with their hypotheses.

Discussion

In this article, we summarized and reviewed different existing theoretical and empirical works regarding the role of digital tools in educational settings. We presented the CoDiL framework as a synthesis of different positions about that specific role of digital tools. One main argument from the last decades of research on educational technology which we want to emphasize is that cognitive processes have a mediating role in the cause-and-effect mechanisms of successful subject-specific learning with digital tools. This mediating role can be framed both in a theoretical and an empirical manner: on a theoretical level, the CoDiL framework describes digital tools by focusing on their implemented digitally enriched instructional features. These instructional features offer learning opportunities to students that may (or may not) be utilized; if utilized they trigger learning activities—which mediated by cognitive processes and in interaction with prior knowledge lead to knowledge gains. From an empirical perspective, these learning activities should be operationalizable by measurable student-tool interactions (generating a broad variety of different process data). Yet, we argue that for the framework to be able to guide research on digital tools in education, its ‘degrees of freedom’ need to be specified for the concrete educational scenario of interest: we recognize that our current framework represents a high-level conceptual integration of KLI, SOI, and ULO. This provides a basic starting point, but more detailed and empirical integration into specific (newly conducted) studies is required. This offers potential for further research from various fields.

On the Assessment of Process Data During Learning With Digital Tools

Present theoretical frameworks on learning in digitally enriched or multimedia scenarios (Chi & Wylie, 2014; Järvelä, 1995; Mayer, 2014; Sweller, 2020) are based on the premise that cognitive processes during (digital) learning cause knowledge acquisition. Although this can be considered a commonly agreed-upon statement about how learning occurs, the specific underlying (latent) learning processes are not always operationalized and assessed in studies about learning in digitally enriched scenarios. Consequently, some recent publications call for a more concrete investigation of the learning process (de Jong, 2010; Kucirkova, 2014) to gain insight into the mechanisms explaining the effects of digital-tool design (Boomgaarden et al., 2023; Sedrakyan et al., 2020)—or aptitude-treatment interactions (Grimm et al., 2023; Huber & Bannert, 2023; Reinhold et al., 2020b). In order to reflect this in our synthesis model, we have centered the CoDiL framework on learning activities that have shown relevant to (digital) learning (e.g., abstraction, conducting experiments, creating examples, exploration, formulating hypothesis, refinement, revision, self-explanation). As this is an active process (in accordance with generative learning theory, Fiorella & Mayer, 2016; Roelle & Nückles, 2019; and theories about self-regulation, Azevedo, 2020; S. Li et al., 2022; Molenaar et al., 2023), we integrated ULO in our framework to contribute for various student behavior (Seidel, 2014); as a direct result of the presented arguments, students’ engagement in these learning activities serves as a theoretical as well as an empirical mediator in the CoDiL framework. Regarding research on the role of digital tools in education, one main work for future primary research is to develop various ways to unobtrusively capture student-tool interactions—and to link these behaviors to the underlying cognitive processes validly. Given such valid measures of learning activities in digitally enriched educational settings, mediation analyses could empirically underpin the role of these activities as theoretically grounded causes for the effects of learning with digital tools.

On the Development of Digital Tools for Educational Purposes

We consider the design of the digital tool (and its digitally enriched instructional features) a central part of the development of a digitally enriched learning environment: we consider it relevant to start research on digitally enriched learning with analyses of the to-be-learned content, the necessary knowledge elements, and the required learning activities to achieve these learning goals—and, thus, not by picking a to-be-investigated digital tool first. In fact, we consider it necessary for researchers to engage in the development (or selection) of digital prototype tools that are explicitly designed to align with the learning mechanisms assumed. Given that, the link between students’ underlying learning activities and their behavioral student-tool interactions may be established during tool development (Goldhammer et al., 2021).

On the Role of Content-Specificity in Learning With Digital Tools

Whereas multiple aspects are important to be considered as mediators for learning success (e.g., cognitive load; Sweller et al., 2019), the activation of and engagement in specific learning activities is considered to be of particular relevance (e.g., Koedinger et al., 2012; Schumacher & Stern, 2023). By describing the CoDiL framework, we argue that digital tools should be designed to stimulate such learning activities. In line with Koedinger and colleagues’ KLI framework, we consider particular purposeful learning activities (that result in cognitive processes relevant for knowledge acquisition) to be content-specific (Koedinger et al., 2012). In our framework, we stress the need to provide detailed information on the specific knowledge components that students need to acquire to learn a specific content. Based on this specification of knowledge components, digital tools can be developed to not only support learning in general (e.g., meta-cognitive scaffolding or corrective feedback), but also to stimulate the essential content-specific (conceptual) knowledge components itself (e.g., task-specific feedback on the most plausible student error given the students’ answer or relevant and hierarchically ordered contrasting cases). Thus, the digital tool (and its implemented instructional features) should be considered an inherent part of the learning environment—aiming for a concrete learning goal.

Conclusion: Guiding Research on Digital Tools in Educational Settings

Studies in the field of digital learning often focus on the investigation of direct effects of digital tools on students’ learning outcome. With this article, we provide a theoretical framework of the learning activities mediating learning with such digital tools in educational settings—focusing on the underlying learning mechanisms in digitally enriched learning scenarios. Research on digital tools in educational settings guided by the proposed CoDiL framework should follow these steps:

  1. 1.

    Specify the content-specific knowledge components to be learned.

  2. 2.

    Specify the learning activities that lead to the acquisition of these knowledge components.

  3. 3.

    Select or design digital tools with instructional features that stimulate and support these learning activities.

  4. 4.

    Collect process data (e.g., on student-tool interaction) to measure the actual utilization of the digitally enriched instructional features of the digital tool that may serve as indicators for the cognitive processes taking place.