Mapping data literacy trajectories in K-12 education

The gist: This position paper proposes a data paradigms framework and learning trajectories to map how K–12 learners engage with data-driven systems across different logic and explainability dimensions.

Theoretical Framework

The authors propose two central dimensions for incorporating data science activities into K–12 computing education: (i) logic (knowledge-based or data-driven systems), and (ii) explainability (transparent or opaque models). This framework categorizes learning activities along these two dimensions to visualize the pathways learners follow across distinct paradigms <ref:2603.28317#pg6>.

The paper distinguishes between Knowledge-based vs data-driven models, noting that students typically solve problems through algorithmic solutions based on “logical proof of correctness” which contrasts with a data-driven approach where students train and evaluate models using data and evaluate systems through a “statistical demonstration of effectiveness” <ref:2603.28317#pg3>. Furthermore, it defines the dimension of explainability, contrasting models that are transparent or opaque <ref:2603.28317#pg6>. Explainability varies across models, with low-dimensional linear regression being highly explainable because its decisions can be accounted for in terms of explicit parameters and relationships <ref:2603.28317#pg5>. Conversely, data-driven models like recurrent neural networks are typically harder to explain because the pathways by which they produce decisions are comparatively opaque <ref:2603.28317#pg6>.

Data Paradigms Framework

The authors analyze learning activities against two dimensions to create four distinct quadrants or data paradigms in Table 1 <ref:2603.28317#pg5>. These quadrants are defined by the combination of logic (KB vs DD) and explainability (T vs O), resulting in: KB+T, DD+T, KB+O, and DD+O <ref:2603.28317#pg5>. For instance, KB+T represents Rule-based models which are explainable by design <ref:2603.28317#pg5>. The framework is used to classify learning activities based on students' engagement with knowledge-based or data-driven systems and the extent to which these systems are explainable within the activity <ref:2603.28317#pg5>.

Data Literacy Trajectories

The authors inductively determined four distinct learning trajectories that characterize student learning within analyzed interventions <ref:2603.28317#pg5>. These trajectories represent pathways learners follow across the four quadrants proposed in the data paradigms framework <ref:2603.28317#pg5>.

1. Keeping transparent: Learners move from knowledge-based/transparent (KB-T) to data-driven/transparent (DD-T) <ref:2603.28317#pg5>. This trajectory involves moving from “hand-authored rules” to training classifiers using data-driven models <ref:2603.28317#pg5>.

2. Keeping data-driven: Learners move from data-driven/transparent (DD-T) to data-driven/opaque (DD-O) <ref:2603.28317#pg5>. This involves progressing from exploratory analysis or simple supervised models to less transparent classifiers or deep learning for prediction <ref:2603.28317#pg5>.

3. Jumping: Learners jump from the knowledge-based/transparent (KB-T) to the data-driven/opaque (DD-O) paradigm <ref:2603.28317#pg5>. This trajectory involves a shift from “teaching computers through programming” to “teaching computers to learn” <ref:2603.28317#pg5>.

4. Bridging: Learners move through more than two paradigms, such as from KB-T to DD-T, then potentially to DD-O <ref:2603.28317#pg5>. This trajectory is characterized by sequencing opaque ML work from transparent reasoning and/or inspectable data work <ref:2603.28317#pg5>.

Discussion and Implications

The authors suggest that the two dimensions—logic (knowledge-based vs data-driven) [36] and explainability (transparent vs opaque) [2]—could serve to delineate the epistemic boundaries of diverse data literacy activities <ref:2603.28317#pg6>. They argue that attention be given to how what approaches to scaffolding [35] are needed to move learners between these paradigms <ref:2603.28317#pg5>. For example, introducing opaque models requires more explicit attention to concepts such as model confidence <ref:2603.28317#pg6>. Conversely, introducing transparent models requires a greater focus on practices such as feature weighting or visualising relationships between variables <ref:2603.28317#pg5>. The paper concludes that the Jumping trajectory indicates a pedagogical challenge as learners ‘bypass’ critical skills in understanding the influence of data on model behaviour <ref:2603.28317#pg6>.

Limitations and Future Work

The results are limited to the extent that they analysed early examples of research-led interventions rather than curricula or grey literature <ref:2603.28317#pg5>. The authors note that most interventions remained within a single paradigm (n=57/84), with few examples of instructional sequences that moved between these paradigms <ref:2603.28317#pg5>. Future work could apply the data paradigms framework to curricula and associated learning materials to identify further trajectories and progressions between different paradigms <ref:2603.28317#pg6>.

Conclusion

The authors present the data paradigms framework and propose a set of learning trajectories as a tool to map the landscape of K–12 data literacy initiatives <ref:2603.28317#pg5>. They suggest these tools could provide a starting point for researchers and resource developers when considering the design of data literacy learning environments <ref:2603.28317#pg6>. The framework offers a shared vocabulary for learning scientists and CS researchers to map the epistemic shifts encountered by learners across contexts and disciplines <ref:2603.28317#pg6>. The paper hopes to further discuss how to align data literacy goals across disciplines and better prepare students for a data-driven future.

--- Page 1 ---

The gist: This position paper proposes a data paradigms framework and learning trajectories to map how K–12 learners engage with data-driven systems across different logic and explainability dimensions.

Theoretical Framework

The authors propose two central dimensions for incorporating data science activities into K–12 computing education: (i) logic (knowledge-based or data-driven systems), and (ii) explainability (transparent or opaque models) <ref:2603.28317#pg6>. This framework categorizes learning activities along these two dimensions to visualize the pathways learners follow across distinct paradigms <ref:2603.28317#pg6>.

The paper distinguishes between Knowledge-based vs data-driven models, noting that students typically solve problems through algorithmic solutions based on “logical proof of correctness” which contrasts with a data-driven approach where students train and evaluate models using data and evaluate systems through a “statistical demonstration of effectiveness” <ref:2603.28317#pg6>. Furthermore, it defines the dimension of explainability, contrasting models that are transparent or opaque <ref:2603.28317#pg6>. Explainability varies across models, with low-dimensional linear regression being highly explainable because its decisions can be accounted for in terms of explicit parameters and relationships <ref:2603.28317#pg6>. Conversely, data-driven models like recurrent neural networks are typically harder to explain because the pathways by which they produce decisions are comparatively opaque <ref:2603.28317#pg6>.

Data Paradigms Framework

The authors analyze learning activities against two dimensions to create four distinct quadrants or data paradigms in Table 1 <ref:2603.28317#pg6>. These quadrants are defined by the combination of logic (KB vs DD) and explainability (T vs O), resulting in: KB+T, DD+T, KB+O, and DD+O <ref:2603.28317#pg6>. For instance, KB+T represents Rule-based models which are explainable by design <ref:2603.28317#pg6>. The framework is used to classify learning activities based on students' engagement with knowledge-based or data-driven systems and the extent to which these systems are explainable within the activity <ref:2603.28317#pg6>.

Data Literacy Trajectories

The authors inductively determined four distinct learning trajectories that characterize student learning within analyzed interventions <ref:2603.28317#pg6>. These trajectories represent pathways learners follow across the four quadrants proposed in the data paradigms framework <ref:2603.28317#pg6>.

1. Keeping transparent: Learners move from knowledge-based/transparent (KB-T) to data-driven/transparent (DD-T) <ref:2603.28317#pg6>.

1. The system can be designed to move learners from rule-based logic to statistical demonstration of effectiveness by implementing activities that progress from inspectable data practices such as visualising and interpreting datasets within the data-driven/transparent paradigm (DD-T) to more opaque systems like neural networks, as described in Trajectory 2.

2. The system can incorporate scaffolding techniques derived from the paper, such as introducing concepts like model confidence or evaluating model performance using confusion matrices when encountering more opaque models to support learner understanding.

3. The system can facilitate a learning sequence that bridges paradigms by structuring activities where learners move from rule-based explanations through Denial-of-Service (DoS) attacks (KB-T) to visualizing data for classification ideas (DD-T), and then potentially to exploring generative adversarial networks (DD-O).

Data literacy skills are fundamental in computer science education. However, understanding how data-driven systems work represents a paradigm shift from traditional rule-based programming. We conducted a systematic literature review of 84 studies to understand K-12 learners' engagement with data across disciplines and contexts. We propose the data paradigms framework that categorises learning activities along two dimensions: (i) logic (knowledge-based or data-driven systems), and (ii) explainability (transparent or opaque models). We further apply the notion of learning trajectories to visualize the pathways learners follow across these distinct paradigms. We detail four distinct trajectories as a provocation for researchers and educators to reflect on how the notion of data literacy varies depending on the learning context. We suggest these trajectories could be useful to those concerned with the design of data literacy learning environments within and beyond CS education.