Source-linked AI summary
The CAPED Framework: Characterizing User Skills in Chart Authoring through the Lens of Tasks and Tools
Bongshin Lee, Zhicheng Liu, John Thompson, Chenglong Wang
TL;DR
Chart-authoring research lacks a framework for characterizing the diverse skills users need, and existing theories do not adequately capture how tasks and tools shape those skills. The paper develops CAPED, a five-dimension framework, and uses it to describe authoring situations and identify research opportunities. Its scope excludes interaction and animation.
Problem
Chart-authoring systems have received attention, but a conceptual framework for characterizing users’ diverse skills is lacking, and high-level skill definitions are insufficient for identifying those skills.
Source: §1 · p. 1, §4.1 · p. 4Method
Results
CAPED describes differing skills in similar authoring situations and supports identifying types of knowledge associated with chart authoring.
Source: §1 · p. 1Takeaways & Limitations
The framework offers a vocabulary for characterizing authoring skills and a basis for exploring skill assessment and support for users with varying skills.
Source: §1 · p. 1, §7 · p. 9The framework is a first attempt scoped to static chart authoring, leaving interaction and animation outside its coverage.
Source: §7 · p. 9
Abstract
from arXiv · showhide
In this paper, we introduce a new framework for characterizing the user skills required for chart authoring. Despite growing attention to authoring systems, we lack a conceptual framework that characterizes the diverse and multifaceted skills involved and helps identify opportunities for system design and development. Our initial investigation revealed that skill is a complex and elusive concept, and that existing theories on skills from other disciplines are too high-level to capture the importance of tasks and tools in data visualizations. To address this gap, we formulate the CAPED framework that characterizes skill as the ability to create a chart involving five dimensions pertinent to tasks and tools: input data, output chart, paradigm, abstraction, and expression. We demonstrate the framework's descriptive power by highlighting different skills involved in similar authoring situations and identifying the types of knowledge associated with chart authoring. We further present examples illustrating how the framework can be used to identify new research opportunities in supporting users with varying skills.
1 INTRODUCTION
Chart-authoring research lacks a framework for describing the diverse skills users need, while existing systems often characterize audiences with ambiguous labels. CAPED frames those skills through five dimensions that connect authoring tasks and tools.
1 INTRODUCTION
- Existing authoring systems are often system-centric, and labels such as “novices” or “non-experts” do not specify users’ existing or needed skills.The paper seeks a systematic way to characterize those skills to inform system design and evaluation.Source: §1 · p. 1
- A conceptual framework is missing for distinguishing users with different combinations of chart, coding, design, and structured-data experience.Authoring also spans tasks such as data preparation, encoding, construction, and refinement across tools and workflows.Source: §1 · p. 1
- The paper introduces CAPED to characterize chart-authoring skill across input data, output chart, paradigm, abstraction, and expression.The framework describes dimensions that shape required abilities; it is neither a skill taxonomy nor a process model.Source: §1 · p. 1
- Excel can create a waterfall chart from a template, whereas composing a dumbbell chart requires adapting scatter plots and error bars.The example illustrates how authoring demands can differ even within the same tool.Source: Figure 2 · p. 2
2 SCOPE AND MOTIVATING SCENARIOS
The paper scopes chart authoring to constructing static charts with abstract data using digital tools, then uses scenarios to show why skills depend on the task, tool, and strategy. The examples span tool transfer, data preparation, and AI-assisted authoring.
2.1 What Chart Authoring Entails in Our Framework
- Chart authoring is composing a chart with a design in mind and data in hand, within a scope centered on static digital visualizations.The framework excludes design ideation, interaction, annotation, animation, hand-drawn or fabricated visualizations, and statistical analysis without charts.Source: §2.1 · p. 2
2.2 Example Scenarios Requiring Diverse Skills
- Excel’s built-in waterfall template and its scatter-plot workaround for a dumbbell chart involve different authoring approaches and skills.The scenario cautions that fluency with one chart workflow does not establish fluency with another, even in the same tool.Source: §2.2 · p. 2
- Kori transfers visualization concepts from programming libraries to Power BI, but the reverse transfer requires learning additional programming skills.The scenario also shows data transformation becoming important for testing a hypothesis and preparing the desired chart.Source: §2.2 · p. 2
- Jesse’s AI-generated grouped bar chart obscures most job-offer measures, so Jesse diagnoses the design mismatch and prompts a parallel coordinates plot.The example argues that AI changes rather than eliminates skill requirements, including code verification, visualization critique, and targeted prompting.Source: §2.2 · p. 3
3 BACKGROUND AND RELATED WORK
Prior work characterizes authoring systems, users, visualization tasks, and visualization literacy, but leaves room for a user- and skill-centered account. CAPED offers a vocabulary for describing skills across diverse authoring tools and situations.
System-Centric Perspectives of Chart Authoring Systems
- Authoring-system studies often assess usability but offer limited insight into authoring processes and required skills, while critical reflection remains system-centric.The paper positions CAPED as a starting point for user- and skill-centered characterization.Source: §3 · p. 3
Target User Characterization
- Labels such as “novices” and “non-experts” can be imprecise, and job roles alone under-specify users’ backgrounds and skills.A cited review found a mismatch between broad novice labels and the narrower populations represented in evaluations.Source: §3 · p. 3
Visualization Task Characterization
- Visualization task frameworks describe user tasks and design processes, while chart-authoring work has begun to characterize how people construct visualizations.CAPED builds on that authoring direction with five dimensions.Source: §3 · p. 3
Assessment of Literacy or Ability
- Visualization literacy tests assess comprehension, while CAPED deconstructs authoring in terms of data, abstractions, and paradigms without providing a skill metric.The framework is presented as a possible basis for authoring assessments and as a way to identify research opportunities at the intersection of AI and chart authoring.Source: §3 · p. 3
4 FRAMEWORK DEVELOPMENT AND SKILL FRAMING
The authors develop CAPED through iterative synthesis, starting from skill as task accomplishment and finding that authoring skills cannot be characterized by tasks alone. They integrate stable tool characteristics with task analysis into a framework of five dimensions.
4 FRAMEWORK DEVELOPMENT AND SKILL FRAMING
- The authors iteratively synthesized skill theory, chart-authoring tasks, and tool comparisons into a framework organized around tasks and tools.Their process included weekly discussions over 16 months and refinement through scenarios, tool comparisons, and skill-transfer discussions.Source: §4 · pp. 3–4, §4.3 · p. 5
4.1 Importance of Tasks in Skill Characterization
- Chart-authoring skill is framed as the ability to create specific charts with given datasets, drawing on conceptual, procedural, and metacognitive knowledge.The authors note that a high-level skill definition alone is not useful for identifying required skills, and that exhaustive enumeration faces an “infinite regress” problem.Source: §4.1 · p. 4
4.2 Analysis of Tasks and the Indispensability of Tools
- Task descriptions alone do not capture skill requirements: tools can change those requirements even when the chart and high-level tasks are similar.Some skills transfer across tools, while others depend on the tool; the same tool can also support distinct authoring strategies.Source: §4.2 · p. 4
4.3 Analysis of Tools and Synthesizing CAPED
- The authors reject UI, modality, device, and rendering as sufficient tool dimensions, identifying abstraction and expression as more useful for chart-authoring skills.These tool dimensions are combined with input data, output chart, and paradigm in the resulting framework.Source: §4.3 · p. 5
5 THE CAPED FRAMEWORK: AUTHORING SKILLS THROUGH THE LENS OF TASKS AND TOOLS
CAPED characterizes chart-authoring skills through five dimensions—input data, output chart, paradigm, abstraction, and expression—covering task and tool requirements.
5.1 Task-Related Dimensions: Input, Output, Paradigm
- Chart authoring tasks focus on what and how; the why is understanding data through exploration or presenting data and insights for communication.Source: §5.1 · p. 5
5.1.1 What: Input Data and Output Chart
- Inputs range from clean and tidy datasets to tool- or output-compatible formats, while the output is a functional chart encoding input data.Source: §5.1.1 · p. 5
5.1.2 How: Authoring Paradigm
- Four common paradigms—compose, instantiate, delegate, and repurpose—vary in scaffolding, are non-exhaustive, and may be sequenced to achieve one chart goal.Source: §5.1.2 · pp. 5–6
5.2 Tool-Related Dimensions: Abstraction and Expression
- Chart authoring tools can be characterized by abstraction and expression.Source: §5.2 · p. 6
5.2.1 Abstraction
- Chart abstractions range from monolithic prototypes and parameterized templates to building-block specifications and data-bound vector graphics, and can vary within a paradigm.Source: §5.2.1 · p. 6
5.2.2 Expression
- Expression is how users articulate a chart abstraction; the same expression can pair with different abstractions, and one abstraction can use multiple expressions.Source: §5.2.2 · p. 6
- Expression types include editing tables, drawing, gesturing, demonstrating structured objects, and manipulating widgets.Source: §5.2.2 · p. 6
- Other expressions include domain-specific languages, natural language, and supplying an example chart in a tool-compatible format.Source: §5.2.2 · p. 6
6 APPLYING THE FRAMEWORK
The section applies CAPED to authoring scenarios and knowledge types, then uses its dimensions to identify design and teaching opportunities for users with different skills.
6.1.1 Describing Authoring Scenarios
- For AI-assisted authoring, a desired output can stay at monolithic-prototype abstraction, but debugging may require building-block construction to inspect generated code.The section also describes transitions between authoring stages and combined abstractions or expressions.Source: §6.1.1 · p. 6, Table 1 · p. 7
- Data Formulator and Falx both delegate data transformation, but Data Formulator exposes transformation code and a transformed table whereas Falx does not expose its inferred logic.Interpreting transformation code is a skill needed for Data Formulator, but not Falx.Source: §6.1.1 · p. 7
- Different input schemas can change the approach to creating the same chart, while Charticulator also supports composing a ribbon chart from scratch or instantiating it from a template.Composition requires tool-compatible input; instantiation requires data exactly compatible with the template.Source: §6.1.1 · p. 7, Table 3 · p. 8
- StructGraphics separates drawing graphics without a dataset, which forms a data-bound vector-graphics abstraction, from later instantiation with compatible data.The first stage uses a drawing interface; the second uses a template.Source: §6.1.1 · p. 7, Table 4 · p. 8
6.1.2 Describing Skills as Knowledge
- The section groups authoring knowledge into conceptual understanding, procedural interface actions, and metacognitive choices about tools and strategies.Examples include understanding chart primitives, preparing data, and switching tools or strategies when authoring stalls.Source: §6.1.2 · p. 8
6.2 Identifying New Research Opportunities
- One research opportunity is to support transitions from AI-generated charts to other paradigms, abstractions, and expressions for refinement.For example, dynamically generated interfaces could let users directly manipulate properties of AI-generated D3 charts.Source: §6.2 · p. 8
- For chart pedagogy, CAPED can help tailor materials to prior skills and focus instruction on abilities not already supported by a tool.For example, business professionals moving from Excel to Power BI may need more focus on composition and visual encoding than on templates or widgets.Source: §6.2 · p. 8
- For designers comfortable with drawing and vector primitives but lacking data transformation skills, the section proposes tighter integration between AI and drawing interfaces.DataInk and StructGraphics offer interaction techniques for those skills but do not help with the needed data transformations.Source: §6.2 · p. 9
7 DISCUSSION AND FUTURE WORK
CAPED is presented as the first theoretical framework for characterizing user skills in static chart authoring through tasks and tools. Its dimensions can interact without being deterministic, while the framework’s scope leaves interaction, animation, and scientific or spatial visualization for future exploration.
7 DISCUSSION AND FUTURE WORK
- CAPED is the first theoretical framework to characterize user skills in static chart authoring through the lens of tasks and tools.The discussion describes the framework as neither exhaustive nor definitive.Source: §7 · p. 9
- CAPED’s dimensions can be interdependent and correlated, but their associations are not deterministic and no dimension can be reduced to another.The authors treat these relationships as reflecting the coupling of tool and task considerations, not as a limitation.Source: §7 · p. 9
- The framework omits interaction and animation to keep this first attempt manageable, although sophisticated interactive exploration may require programming and interface and interaction design.The authors suggest that future extensions may revise the output-chart dimension or add dimensions for non-static charts.Source: §7 · p. 9
- Because CAPED is derived primarily from information-visualization chart-authoring tools, its dimensions may not fully capture abstractions in scientific or spatial visualization.The authors identify extending and validating the framework for those domains as future work.Source: §7 · p. 9
- CAPED can ground future measures of chart-authoring skill and standardized assessments for participant grouping, job descriptions, candidate screening, and skill evaluation.Suggested assessment targets include prompting proficiency and expertise with a visualization library.Source: §7 · p. 9