PaperlayerView exact revision on arXiv

Source-linked AI summary

Representational Fidelity in Didactic Visualization: Toward a Multidimensional Design Space

Shehryar Saharan, Michele Oliver, Karen Gordon, Gaël McGill, Jodie Jenkinson

arXiv:2609.08037v1cs.HC

TL;DR

Didactic visualization research often treats representational fidelity as a single continuum, leaving limited shared means to describe and compare specific visual choices. This paper develops a multidimensional design space through corpus analysis and a pilot focus group, then illustrates its use in an applied design case study. The framework shows descriptive and generative potential, while its usability across broader participant groups remains to be assessed.

  • Problem

    Researchers lack systematic ways to describe and intentionally manipulate representational properties of didactic visuals, while practitioners and researchers conceptualize fidelity differently.

    Source: §1 · p. 2

  • Method

    The authors derive a multidimensional design space through iterative corpus analysis and use a pilot focus group and applied design case study to explore its application.

    Source: §6 · p. 7, §8 · p. 9

  • Results

    The corpus analysis identifies relationships among fidelity dimensions, including a negative association between Perceptual Fidelity and Model-derived Fidelity (r = –0.54).

    Source: §5 · p. 7

  • Takeaways & Limitations

    The framework offers descriptive power and generative potential for articulating representational choices and identifying patterns and opportunities in contemporary didactic media.

    Source: §8 · p. 9

    The focus group involved highly trained scientific visualizers, limiting generalizability; broader and more diverse participants are needed to assess usability across contexts.

    Source: §7.2 · p. 9

Abstract

from arXiv · show

Representational fidelity is routinely treated as a single abstract-realistic continuum, a simplification that limits how it is described and compared across research and design contexts. We introduce a multidimensional design space of representational fidelity for didactic visualization in science & engineering, inductively derived from a 175-item corpus spanning several disciplines, modalities, and instructional aims. The resulting design space specifies five dimensions: Morphological, Dynamic, Cueing, Contextual, and Interactive Fidelity, with seven sub-dimensions. We demonstrate the design space's descriptive power through successive rounds of expansion and refinement and analyze the corpus to reveal relationships among dimensions and implications for design and research. We further validate the design space through a pilot focus group in which participants applied the dimensions in an open-ended design exercise. Resulting sketches and verbal rationales informed a single-designer applied case study, offering preliminary evidence of the design space's generative potential as a structured aid to design exploration. Together, these contributions lay the groundwork for future research and more intentional design practice.

1 INTRODUCTION

The paper addresses the lack of a shared way to describe and compare representational fidelity in science and engineering didactic visualizations. It proposes a multidimensional design space to clarify representational choices without deciding which configurations are best for learning.

1 INTRODUCTION

  • When multiple visual features shift together under a single “more realistic” label, studies cannot readily attribute effects to specific visual changes.This makes it difficult to compare and synthesize outcomes across studies.Source: §1 · p. 1
  • Researchers lack systematic ways to describe, compare, and intentionally manipulate representational properties of didactic visuals in science and engineering.Existing labels such as “realistic” or “high-fidelity” often do not specify which visual features change.Source: §1 · pp. 1–2
  • The paper defines fidelity as a multidimensional property that mediates access to a referent through choices about what is shown, how it behaves, its context, emphasis, and exploration.Fidelity is not limited to literal resemblance or accuracy, and the framework does not adjudicate which configurations are best for learning.Source: §1 · p. 2
  • The proposed design space identifies five primary dimensions and seven subdimensions from a systematic examination of 175 didactic visualizations across disciplines and modalities.The paper also makes the full coded library available.Source: §1 · p. 2

2 BACKGROUND

The background describes didactic visualizations as varied visual representations shaped by pedagogical choices, while reviewing how fidelity is treated in prior work. It identifies the need for a design-space account suited to didactic visualization.

2.1 Didactic visualizations

  • Didactic visualizations include static diagrams, annotated illustrations, animations, simulations, and interactive models, with design choices made in service of pedagogical goals.They can represent complex, abstract, dynamic, or otherwise inaccessible concepts, processes, or systems.Source: §2.1 · p. 2

2.2 Fidelity in didactic visualization

  • Reviewed fidelity research predominantly treats fidelity as a single continuum, while multidimensional accounts remain comparatively sparse and no reviewed account sufficiently addresses didactic visualization.The authors build on existing accounts to develop a design space for this context.Source: §2.2 · p. 3

2.3 Design Spaces

  • Design spaces support systematic exploration and comparison of alternatives, but their use in learning-focused research is often implicit.Educational design work often relies on convention, intuition, and craft-based knowledge.Source: §2.3 · p. 3

3 METHODOLOGY

The authors developed the design space through staged corpus analysis and complemented it with a pilot focus group and an applied design case study. The study focuses on science and engineering didactic visualizations with real-world referents.

3 METHODOLOGY

  • The methodology combines staged corpus analysis with a pilot focus group and applied design case study to develop and validate the design space.The corpus strand provides the empirical foundation; the second strand offers early practical validation.Source: §3 · p. 3, Figure 3 · p. 4

3.1 Boundaries of our Design Space

  • The design space covers science and engineering visualizations with real-world referents, including phenomena perceptible through instrumentation or beyond unaided vision.It excludes tactile, physical, or VR/AR modalities and data or statistical graphics without real-world referents.Source: §3.1 · p. 3

3.2 Strand 1 - Corpus Collection and Development of Design Space

  • The corpus evolved through initialization, expansion, and refinement, with coding and collaborative review sharpening dimensions before the final collection reached 175 items.Refinement tested under-represented formats and edge cases; no new distinctions emerged, and all added examples fit the design space.Source: §3.2 · pp. 4–5

3.3 Strand 2 - Pilot Focus Group & Design Case Study

  • Four second-year MScBMC students explored the dimensions in a focus group, and their sketches and rationales informed a single-designer applied case study.In the open-ended exercise, pairs generated variants of blood-flow visualizations for first-year anatomy or physiology students and made fidelity choices explicit.Source: §3.3 · p. 5

4 THE DESIGN SPACE

The design space treats representational fidelity as multidimensional, organized into five dimensions and seven sub-dimensions. Its dimensions distinguish form, change, attention, context, and user engagement rather than reducing fidelity to resemblance.

4 THE DESIGN SPACE

  • The design space organizes fidelity into Morphological, Dynamic, Cueing, Contextual, and Interactive dimensions, with seven sub-dimensions that can vary independently.Dimensions span theoretically infinite continua; inclusion does not require high fidelity on any dimension.Source: §4 · p. 5

4.1 Morphological Fidelity; 4.1.1 Spatial Fidelity

  • Morphological Fidelity covers form and appearance; Spatial Fidelity concerns geometric structure and relationships, which can remain abstract while preserving structural information.Morphological Fidelity includes spatial organization, perceptual qualities, and constructed appearances; Spatial Fidelity ranges from ordering and continuity to literal spatial correspondences.Source: §4.1 · p. 5, §4.1.1 · pp. 5–6

4.1.2 Perceptual Fidelity

  • Perceptual Fidelity measures whether visual cues correspond to a referent’s sensory qualities, and applies only when that referent has an observable appearance.For imperceptible referents, visual qualities are better treated as model-derived conventions or cueing strategies.Source: §4.1.2 · p. 6

4.1.3 Model-derived Fidelity

  • Model-derived Fidelity constructs a coherent visual proxy for an imperceptible referent without claiming that the proxy reproduces its true appearance.Such appearances may draw on scientific data, computational models, disciplinary norms, or established conventions.Source: §4.1.3 · p. 6

4.2 Dynamic Fidelity; 4.2.1 Functional Fidelity; 4.2.2 Temporal Fidelity

  • Dynamic Fidelity concerns depictions of change, distinguishing Functional Fidelity’s causal and mechanistic relationships from Temporal Fidelity’s encoding of before-and-after states.Temporal change may appear through animation, ordered sequences, staged progression, or other interpretable temporal ordering; viewpoint rotation alone does not encode system change.Source: §4.2 · p. 6, §4.2.1 · p. 6, §4.2.2 · p. 6

4.3 Cueing Fidelity

  • Cueing Fidelity uses perceptual variables such as color, contrast, motion, and labels to direct attention, even when those cues do not resemble the referent.A muted overall rendering can provide a substrate for locally foregrounded elements.Source: §4.3 · p. 6

4.4 Contextual Fidelity

  • Contextual Fidelity situates a subject within a frame, environment, or reference system, which may be partial, stylized, or abstract.Context can also arise from relationships among elements or from a figure whose primary function is to establish a reference system.Source: §4.4 · p. 6

4.5 Interactive Fidelity; 4.5.1 Observational Fidelity; 4.5.2 Simulation Fidelity

  • Interactive Fidelity requires meaningful engagement: Observational Fidelity enables inspection of the representation, while Simulation Fidelity lets user changes produce consequential system outcomes.Changing a viewpoint or revealing hidden structure is observational; changing parameters or conditions to generate outcomes reflecting the modeled system’s logic is simulation.Source: §4.5 · p. 6, §4.5.1 · pp. 6–7, §4.5.2 · p. 7

5 ANALYSIS OF DESIGN DIMENSION DISTRIBUTIONS

The corpus analysis describes how often the dimensions appear and which tend to co-occur. Spatial and Cueing Fidelity are ubiquitous, while Interactive Fidelity is less common; the dimensions also show characteristic associations and tensions.

5 ANALYSIS OF DESIGN DIMENSION DISTRIBUTIONS

  • Spatial Fidelity and Cueing Fidelity appear in all 175 corpus items; Contextual Fidelity appears in 142, Functional Fidelity in 139, and Perceptual Fidelity in 110.Temporal Fidelity appears in 95 items, Model-derived Fidelity in 81, Observational Fidelity in 29, and Simulation Fidelity in 12; the authors note the sampling was selective and purposeful.Source: §5 · p. 7
  • Perceptual and Model-derived Fidelity are negatively related (r = –0.54), while Observational and Simulation Fidelity are strongly associated (r = 0.61).Functional and Temporal Fidelity are moderately correlated (r = 0.47); the authors report hybrid designs despite the negative Perceptual–Model-derived relationship.Source: §5 · p. 7

6 DESIGN SPACE IN PRACTICE

The authors use a focus group and an applied case study to examine how designers engage with the fidelity dimensions. Participants used them to explore alternatives and articulate choices, while the case study illustrates how dimension interactions can complicate interpretation.

6 DESIGN SPACE IN PRACTICE

  • The applied case study was a single-designer proof-of-concept using a blood-flow-through-the-heart prompt, not independent evidence of the design space’s generative power.The designer was the first author, an experienced visualization practitioner, and domain familiarity was required for both the focus group and case study.Source: §6 · p. 7
  • Participants used the dimensions to generate variants, revisit initially unsuccessful designs, and explore trade-offs beyond their initial sketches.The fidelity dimensions typically entered after sketches began from the communication goals, helping designers interrogate and extend those initial choices.Source: §6 · p. 7
  • The focus group and case study suggest the design space can support critique and communication as well as deliberate generation of alternatives.Participants reported that shared vocabulary helped them negotiate disagreements and explain design choices that might otherwise remain tacit; novice use is suggested as a possibility.Source: §6 · pp. 7–8
  • Although dimensions can be adjusted independently, their combinations can have compounding effects that make observed outcomes difficult to attribute to particular representational factors.The authors state that this ambiguity can undermine internal validity and limit interpretability; the design space lets researchers specify manipulated and controlled dimensions.Source: §6 · p. 8

7 DISCUSSION

The discussion assesses the design space’s descriptive, generative, and preliminary evaluative potential, then bounds its claims and identifies open questions for future research. It highlights corpus and participant limitations alongside questions about adoption, research use, and extensions of the framework.

7.1 Descriptive, Generative, and Evaluative Power

  • The dimensions and subdimensions remained stable through iterative corpus expansion, reaching theoretical saturation while supporting consistent description of examples beyond the original corpus.New examples sharpened definitions without expanding the design space; the pilot focus group and applied case study also used its dimensions to articulate and justify design choices.Source: §7.1 · p. 8
  • The corpus analysis, pilot focus group, and single-designer case study provide preliminary evidence that the design space can support idea generation and explicit discussion of design choices.Figure 4 places five case-study examples approximately along Spatial and Perceptual Fidelity; other dimensions also vary but are omitted from that projection.Source: §7.1 · p. 8, Figure 4 · p. 8
  • The authors do not yet claim evaluative power; support for evaluative use remains preliminary and requires further development, testing, and validation.Source: §7.1 · p. 8

7.2 Limitations

  • The selectively and purposefully assembled corpus was designed to probe conceptual boundaries, not to represent instructional materials used across science and engineering education.Source: §7.2 · pp. 8–9
  • The focus group’s highly trained scientific visualizers limit generalizability, while thumbnail sketches and time constraints may have narrowed the breadth or expression of exploration.The authors note that the extended applied case study partly compensates for the focus group’s time constraints.Source: §7.2 · p. 9
  • Because the corpus is English-language and mostly Western, other educational cultures may introduce fidelity distinctions or dimension relationships not captured here.Source: §7.2 · p. 9

7.3 Open Questions & Future Work

  • Open questions include whether practitioners and researchers can adopt the framework in real workflows, and whether it supports stimulus characterization, comparison, or retrospective analysis.Possible research directions include graduated scales for dimensions and examining whether the framework helps explain divergent findings or improve design communication.Source: §7.3 · p. 9
  • Further work could relate fidelity to pedagogical intent, accuracy, affect and trust, or narrower design contexts, and examine variation within a dimension across one stimulus.Source: §7.3 · p. 9

8 CONCLUSION

The paper introduces a multidimensional design space for didactic visualization, derived from corpus analysis and informed by a focus group and applied design case study. The authors report descriptive power and generative potential, presenting the framework as a foundation for future research and design practice.

8 CONCLUSION

  • The five-dimension, seven-subdimension design space derives from iterative analysis of 175 items and is informed by a pilot focus group and applied design case study.Source: §8 · p. 9
  • The framework demonstrates descriptive power and generative potential, and lays groundwork for experiments, pedagogical-alignment research, and design-practice investigations.Source: §8 · p. 9

A AUTHOR CONTRIBUTION

The paper reports contributor roles using the CRediT taxonomy and presents them in a visual matrix.

A AUTHOR CONTRIBUTION

  • The authors summarize contributions with CRediT and provide a visual matrix to make contributor roles transparent.Source: A AUTHOR CONTRIBUTION · p. 12
  • The matrix uses contributor columns and role rows, distinguishing main or substantial contributions from secondary ones; Software and Funding acquisition are omitted as inapplicable.Source: Figure 5 · p. 12

B FOCUS GROUP PARTICIPANTS

The focus group recruited four second-year medical illustration graduate students with scientific and visual-communication training. Participants were recruited voluntarily and had not previously seen the proposed design space.

B FOCUS GROUP PARTICIPANTS

  • Four second-year MScBMC students in medical illustration brought scientific training and experience with illustration, animation, interactive media, and didactic visualization.The study sought participants able to assess scientific departures in representations and articulate possible pedagogical reasons for them.Source: B FOCUS GROUP PARTICIPANTS · p. 12
  • Researchers recruited volunteers through an open call from a cohort they did not directly supervise; participants had no prior exposure to the design space.The study received UofT REB approval, and participants gave written informed consent.Source: B FOCUS GROUP PARTICIPANTS · p. 12
  • Figure 6 shows the Strand 2 methodological flow and sample participant sketches; the complete sketch set is in the supplemental materials.Source: B FOCUS GROUP PARTICIPANTS · p. 12

C DESIGN SPACE - SUPPLEMENTARY GUIDE

The supplementary guide supports consistent coding with discriminating questions, boundary cases, and corpus examples. Its dimensions distinguish what visual form represents, how change is encoded, what is emphasized or situated, and what interaction changes.

C DESIGN SPACE - SUPPLEMENTARY GUIDE

  • The guide pairs each dimension with a discriminating question, boundary cases, and corpus examples to support consistent, reproducible application.It also documents how dimensions evolved through corpus-analysis phases.Source: C DESIGN SPACE - SUPPLEMENTARY GUIDE · pp. 12–13

C.1 Morphological Fidelity

  • Morphological Fidelity separates preserved structure, resemblance to observable appearances, and constructed appearances for unobservable referents.Spatial, Perceptual, and Model-derived Fidelity can occur in combination.Source: §C.1 · p. 13, §C.1.1 · p. 13, §C.1.2 · p. 13

C.2 Dynamic Fidelity

  • Dynamic Fidelity distinguishes showing change over time from encoding the causal or mechanistic relationships that drive change.Temporal Fidelity can occur without Functional Fidelity, while either can occur independently.Source: §C.2 · p. 13, §C.2.1 · p. 13, §C.2.2 · p. 13

C.3 Cueing Fidelity

  • Cueing Fidelity concerns visual variables used to direct attention to specific information, relationships, or structural features.The guide’s example uses targeted color, callouts, and annotation against a muted rendering.Source: §C.3 · p. 13

C.4 Contextual Fidelity

  • Contextual Fidelity is present when a visualization situates its subject, including through relationships among elements rather than an environmental background.A neutral background without relational anchoring does not itself provide context.Source: §C.4 · p. 13

C.5 Interactive Fidelity

  • Interactive Fidelity depends on whether interaction changes how a system is observed or changes the system according to its internal logic.Observational interaction changes the view; simulation interaction produces consequential system-level changes, so interactive components alone do not qualify.Source: §C.5 · p. 14, §C.5.1 · p. 14, §C.5.2 · p. 14

D CORPUS EXAMPLES

Seven corpus examples show how the proposed dimensions characterize static, animated, and interactive visualizations. Their coded features illustrate that examples can combine several kinds of fidelity, including when a dimension is absent.

Figure 8A

  • The surgical illustration combines preserved anatomical arrangement and observable appearance with broader anatomical context and attention-directing labels and callouts.Color, shading, and tissue texture correspond to structures’ appearance in a surgical field.Source: D CORPUS EXAMPLES · p. 14

Figure 8B

  • The hepatic-iron visualization combines muted naturalistic rendering and targeted cues with anatomical context, a constructed magnetic-field depiction, and a causal account of susceptibility.Its color, callouts, and annotation direct attention to the pathology.Source: D CORPUS EXAMPLES · p. 14

Figure 8C

  • The pulley diagram preserves spatial relationships and represents forces through conventions, while encoding mechanics and cueing relevant quantities without contextual anchoring.Its geometric primitives and labels preserve which masses connect to pulleys and ropes.Source: D CORPUS EXAMPLES · p. 14

Figure 8D

  • The breathing infographic combines temporal and functional fidelity with spatial arrangement, organismal context, and cueing across human, bird, and grasshopper respiration.It encodes cyclical phase timing, directional airflow, and sequential progression.Source: D CORPUS EXAMPLES · p. 14

Figure 8E

  • The membrane-fusion animation combines constructed molecular appearances with spatial structure, sequential change, mechanism, cueing, and host-cell context.Selective coloration emphasizes glycans at key moments.Source: D CORPUS EXAMPLES · p. 14

Figure 8F

  • The substrate-concentration interactive combines modeled reaction-rate changes with observational and temporal interaction, molecular representations, contextual relations, and cueing.Changing concentration produces reaction-rate changes in an adjacent panel under an enzyme-kinetics model.Source: D CORPUS EXAMPLES · pp. 14–15

Figure 8G

  • The moveable-joint visualization combines contextual, observational, spatial, perceptual, cueing, functional, and temporal fidelity.Selecting a joint changes what is seen and reveals its characteristic range of motion over time.Source: D CORPUS EXAMPLES · p. 15

E CORPUS ANALYSIS

The corpus analysis is presented through incidence distributions and correlations among dimensions. Both are based on the 175-item corpus and binary presence/absence coding.

E CORPUS ANALYSIS

  • Figure 7A shows the incidence of each design dimension and sub-dimension across all 175 corpus items.The full coded corpus is available in Supplemental Materials, S1.Source: E CORPUS ANALYSIS · p. 15
  • Figure 7B presents Pearson correlation coefficients among dimensions and sub-dimensions, computed from binary presence/absence coding.The matrix supports the interpretation and discussion in Section 5.Source: E CORPUS ANALYSIS · p. 15
  • The figure’s distribution and correlation panels provide corpus-level summaries; item-level detail is available in Supplemental Materials, S1.Source: E CORPUS ANALYSIS · p. 15

F APPLIED DESIGN CASE STUDY

The applied case study uses a practitioner’s self-directed exploration to traverse the design space through situated design choices rather than measurement. The account also notes that binary coding misses variation within a visualization.

F APPLIED DESIGN CASE STUDY

  • The case study treats dimension indicators as reflective records of design intent, not reproducible measurement scores.The first author’s individual judgment and creative risk-taking are central to this exploratory strand; consensus was not sought.Source: F APPLIED DESIGN CASE STUDY · p. 15
  • Binary presence/absence coding misses within-image variation in a dimension, such as a simplified contextual illustration beside a more perceptually detailed focal image.A dimension is coded as present if it appears anywhere in the image, regardless of how much of the image it characterizes.Source: F APPLIED DESIGN CASE STUDY · p. 15
  • The design exploration began from an intuition-driven baseline, then manipulated fidelity dimensions as generative dials across successive variants.Figure 9 maps the variants and marks dimensions altered relative to the preceding iteration.Source: F APPLIED DESIGN CASE STUDY · pp. 15, 17, Figure 9 · p. 18
  • Contextual information was reduced and reintroduced, while shifts between perceptual detail and spatial structure produced variants that raised audience- and learning-stage questions.The author considers whether a variant that scaffolds early conceptual understanding might hinder later reasoning about structural detail.Source: F APPLIED DESIGN CASE STUDY · p. 17
Loading 2609.08037v1…