Making Abstraction Concrete: A Design Space and Interaction Model of Abstraction in Interactive Systems

Abstract: The principle of abstraction guides the design of interactive systems, yet we lack a conceptual framework to understand how it shapes interaction design. Existing models, such as the gulfs of execution and evaluation, do not explicitly model abstractions in the system or in users' mental models, and therefore lack actionable guidance for designing abstractions. To investigate how abstractions are employed in interactive systems, we surveyed 457 papers and synthesized a design space of abstraction techniques along six dimensions. We use this design space to reframe the gulfs through a lens of abstraction, explicitly articulate the cognitive and design processes by which users and systems bridge and navigate the abstraction gap, and demonstrate how this model integrates existing perspectives and surfaces new opportunities for future systems.

• The paper introduces a multidimensional design space that systematically characterizes abstraction techniques in interactive systems.
• The paper presents the Abstraction Spaces model, featuring constructs like the Gulf of Abstraction and misalignment dynamics.
• The paper validates the model through diverse workflows, offering actionable insights for designing and evaluating abstract interactions in HCI.

Authoritative Summary of "Making Abstraction Concrete: A Design Space and Interaction Model of Abstraction in Interactive Systems" (2605.11344)

Introduction and Motivation

The paper addresses a foundational gap in HCI theory: despite abstraction's pervasive influence in interactive system design and user cognition, no existing interaction model explicitly theorizes abstraction as a core process. Prior frameworks, including the canonical Gulfs of Execution and Evaluation, recognize abstraction only peripherally, offering little actionable guidance for its design or systematic understanding. Building on a survey of 457 HCI papers, the authors synthesize a comprehensive design space of abstraction techniques and propose the Abstraction Spaces model, introducing the Gulf of Abstraction and a multidimensional perspective on abstraction misalignment in interaction.

Design Space of Abstraction Techniques

The survey develops a six-dimensional design space for abstraction in interactive systems, grounded in thematic analysis of the literature:

• Unit of Interest: Distinguishes abstraction of content (e.g., data points, text, media) versus operations (e.g., gestures, commands, expressions).
• Granularity: Measures detail in abstraction, spanning techniques for complexity reduction (summarization, attribute hiding), complexity addition (clustering, granular encoding), and specificity/ambiguity management in user input.
• Representation: Concerns the form of abstraction in the interface, from linear lists to spatial canvases, proxy representations, and modalities for accessibility.
• Transformability: Assesses end-user agency in modifying abstractions—ranging from fixed and designer-controlled abstractions, to predefined sets, up to generative and malleable abstractions that users can customize.
• Presentation: Examines how abstractions are situated relative to interface components, including mechanisms for switching, synchronizing, or juxtaposing multiple abstractions.
• Guidance: Describes scaffolding strategies for organizing abstraction exposure, including interface mantras (e.g., Overview First, Details on Demand), concreteness fading, and task-driven abstraction transitions.

This design space spans both technical and cognitive dimensions, offering systematic criteria for creation, evaluation, and extension of abstraction strategies in HCI.

Abstraction Spaces: Interaction Model

The Abstraction Spaces model augments the traditional Gulfs framework by introducing three key constructs:

• Gulf of Abstraction: The vertical misalignment between abstractions in the user's mental model and those presented by the system. Traversing this gulf may require adaptation by either user or system.
• Abstraction Overshoots and Undershoots: Categorical misalignments where the system's abstraction is too high-level (overshoot) or too low-level (undershoot) for the user's needs, requiring compensatory cognitive or system-level adaptation.
• Space of Abstractions: Moves beyond the linear ‘ladder of abstraction’ to conceive of abstraction as a multidimensional space, enabling diverse and non-hierarchical navigation between abstractions.

The model explicitly locates abstraction within all stages of Norman's seven-stage action framework, highlighting reciprocal influence and misalignment dynamics, and surfaces design opportunities for systems to either align or strategically misalign abstractions to provoke user reflection or adaptation.

Workflow and Perspective Modeling

The model is validated against diverse interaction workflows:

• Automating Behavior: Tracks the evolution of mental abstraction alignment through repeated use, lowering the cognitive cost of traversing abstraction gulfs.
• Reifying Operations: Demonstrates systems (e.g., Draco [80]) that encapsulate higher-level abstractions of operations, reducing repetition and empowering users to act at a more expressive level.
• Sensemaking with LLMs: Formalizes the challenge of abstraction undershoots in LLM-driven interfaces, where surfeit of AI-generated content requires the system or user to abstract for sensemaking.
• Multi-level and Malleable Abstractions: Analyzes systems enabling flexible navigation and customization across abstraction levels (e.g., WritLarge [178], malleable overview-detail interfaces [109]), mapping user information needs onto the abstraction space rather than a single ladder.

Additionally, the paper shows that Abstraction Spaces subsumes and extends existing theoretical frameworks such as Instrumental Interaction and the Gulf of Envisioning, effectively capturing abstraction-related design and cognitive challenges that prior models leave implicit.

Implications: Practical and Theoretical

Practically, the Abstraction Spaces model provides a rigorous analytic lens to diagnose, design, and evaluate abstraction in interactive systems. By foregrounding abstraction misalignment, designers are equipped to systematically address user needs across varying levels of specificity and representational form. The model sensitizes researchers to the limitations of heuristic abstraction ladders and encourages exploration of richer multidimensional abstraction spaces.

Theoretically, the explicit modeling of abstraction introduces new cognitive and computational levers in HCI—prompting reconsideration of how systems scaffold, align, or intentionally misalign with user cognition. The model lays groundwork for studying the effects of abstraction design on reasoning, sensemaking, and interaction alignment, especially with the proliferation of generative AI interfaces whose abstraction boundaries are dynamic and difficult to fix.

Future Directions in AI

The paper speculates on advances in human-AI interaction, especially as generative systems (e.g., LLMs) increasingly offer variable abstraction levels. Research must address how to engineer reciprocal abstraction alignment, including deliberate misalignment to provoke metacognition or critical self-reflection. Techniques such as abstraction variation, feedforward previews, and context-driven user modeling offer promising strategies for designing interfaces that responsively navigate abstraction gulfs. The shift from singular or ladder-based abstraction to an expansive abstraction space is expected to unlock fundamentally new forms of co-creative interaction, knowledge discovery, and user agency.

Conclusion

"Making Abstraction Concrete" (2605.11344) provides a systematic, multidimensional analysis of abstraction in interactive system design. The Abstraction Spaces model reframes traditional HCI interaction theory, integrating abstraction as an explicit, actionable component and opening new vistas for designing, analyzing, and evaluating abstraction-driven workflows. The survey and model together equip HCI and AI researchers with precise language and constructs for advancing the study and design of abstraction in interactive systems.