Mapping data literacy trajectories in K-12 education

summary

Video file (mp4)

The gist

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

In short

The research mapped how K-12 students engage with data systems by using a framework based on two dimensions: logic (knowledge-based vs. data-driven) and explainability (transparent vs. opaque models). This created four learning quadrants and four distinct pathways, or trajectories, showing how students move between these different ways of interacting with data.

Key concepts

Logic
This dimension distinguishes between two main ways students approach problems: knowledge-based systems, where they use logical proofs to find correct answers, and data-driven systems. In the latter, students train models using actual data to see if the system works statistically.
Explainability
This refers to how easy it is for a student or teacher to understand *why* a model makes a certain decision. Models are categorized as transparent, meaning their workings are clear like linear regression, or opaque, meaning they are complex like recurrent neural networks.

Terminology used across episodes

This episode discusses

The paper

Mapping data literacy trajectories in K-12 education · Read on arXiv

ROBERT WHYTE, MANNI CHEUNG, KATHARINE CHILDS, JANE WAITE, SUE SENTANCE

Raspberry Pi Foundation · University of Cambridge

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.

Transcript

Introduction to the show: ident: AI Radio. Generated commentary on the latest Artificial Intelligence papers.

Tom: Today's paper: "Mapping data literacy trajectories in K-12 education".

Jane: 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.

Tom: First, who's behind it and why it matters.

Title and authors: Tom: So, we're talking about the paper "Mapping data literacy trajectories in K-twelve education," and the authors are Whyte, Cheung, Childs, Waite, Sentance from the Raspberry Pi Foundation and Cambridge University <ref:2603.28317#pg1>. They’ve done a systematic literature review of eighty-four studies to figure out how students engage with data across all these different subjects <ref:2603.28317#pg1,a systematic literature review of 84 studies to>.

Jane: That systematic review is key because it gives them a solid foundation before they propose this framework, showing that there's definitely an area for more focused research in K-twelve education <ref:2603.28317#pg3>.

Lu: It’s about looking at the existing work and trying to find common language—a vocabulary—to describe these very different learning environments across all disciplines <ref:2603.28317#pg5>.

Meng: I wonder what kind of data they found in those eighty-four studies <ref:2603.28317#pg1>? Does it show a clear pattern of how students move between these logic and explainability dimensions?

Tom: They did, and the results led them to create this data paradigms framework, which is essentially a way to visualize those pathways across the four quadrants we talked about—KB+T, DD+T, KB+O, and DD+O <ref:2603.28317#pg5>.

Jane: So this isn't just a theory on paper; it’s an attempt to categorize real learning activities based on the specific systems students are interacting with and how much they can understand about those systems <ref:2603.28317#pg5>.

Lu: It’s trying to give us a common structure for what "data literacy" means in this context, moving beyond just looking at individual interventions <ref:2603.28317#pg6>.

The paper's summary: Jane: Now that we know the framework, the paper really walks us through their findings on four specific learning trajectories that students can take based on those quadrants <ref:2603.28317#pg5>. They call them Keeping transparent, Keeping data-driven, Jumping, and Bridging <ref:2603.28317#pg5>.

Tom: Let's break down the first one: "Keeping transparent," which is moving from knowledge-based and transparent to data-driven and transparent—that means going from hand-authored rules to training classifiers using data <ref:2603.28317#pg5>.

Meng: That sounds like a very structured progression, but I wonder if that transition is smooth for a student? Moving from explicit rules to statistical demonstrations can feel like a big jump in thinking <ref:2603.28317#pg5>.

Jane: It is, and the paper suggests that when we introduce more opaque models later on, like recurrent neural networks, it gets harder to explain how they reach their decisions <ref:2603.28317#pg6>.

Lu: And that's where the trajectory "Jumping" comes in—that’s when learners shift from teaching computers through programming to actually teaching them to learn <ref:2603.28317#pg5>.

Tom: The paper also points out that this jumping trajectory is a pedagogical challenge because students might bypass really critical skills about how data influences model behavior <ref:2603.28317#pg6>.

The paper's improvements: Tom: The authors suggest a few ways to make this framework more useful, like focusing more on the scaffolding needed at each step <ref:2603.28317#pg5>. For instance, they argue that introducing opaque models needs explicit attention to things like model confidence scores <ref:2603.28317#pg6>.

Jane: That makes sense from a communication standpoint; if you're showing a black box, you have to give the student tools to check its work, maybe by using confusion matrices for evaluating performance <ref:2603.28317#pg5>.

Lu: And they suggest that when we're moving into more complex systems, like generative adversarial networks, we need sequences that bridge those gaps—moving from rule-based explanations to visualizing data for classification ideas <ref:2603.28317#pg5>.

Meng: From a production side, if the trajectory involves jumping into opaque learning, I’d want to see how we can keep the operational costs manageable while still exposing some level of interpretability during training <ref:2603.28317#pg6>.

Tom: They are pushing for these specific scaffolding techniques—like feature weighting when dealing with transparent models or visualizing relationships between variables when you introduce new complexity <ref:2603.28317#pg5>.

Conclusion: Jane: So, to wrap up, the core idea of this paper is that by mapping these data literacy trajectories across logic and explainability dimensions, we get a much clearer picture of the different kinds of data experiences K-twelve students have <ref:2603.28317#pg5>.

Lu: It provides a shared vocabulary for researchers and computer science people to describe these epistemic shifts that learners encounter in different contexts <ref:2603.28317#pg6>.

Tom: It gives us a starting point for anyone designing data literacy learning environments, telling them exactly where the friction points are in the learning path <ref:2603.28317#pg6>.

Meng: But they did note a limitation: their results were based on early examples of research-led interventions, not necessarily full curricula, and most interventions stayed within just one paradigm <ref:2603.28317#pg5>.

Jane: That means future work should really focus on applying this framework to broader curricula and materials to see if those trajectories hold up across a wider range of learning experiences <ref:2603.28317#pg6>.

Lu: It’s a useful tool for researchers, but we need more data from actual classroom instruction to fully validate these four trajectories <ref:2603.28317#pg5>.

Tom: Exactly. The paper "Mapping data literacy trajectories in K-twelve education" gives us a way to visualize the journey learners take through data science concepts, and that’s something we need to keep looking at as we try to prepare students for a more data-driven future <ref:2603.28317#pg5>.

More episodes

← Home