User Capability in an Adaptive World
Robert Dodd, Steve Green, Elaine Pearson · 2009 · MSIADU '09: Proceedings of the 1st ACM SIGMM International Workshop on Media Studies and Implementations that Help Improving Access to Disabled Users · doi:10.1145/1631097.1631110
Summary
This paper presents a fundamental rethinking of how user profiles should be constructed for adaptive and accessible computing. Published at MSIADU '09 (co-located with ACM Multimedia), it critiques the dominant Access For All standard (ISO/IEC 24751) for conflating user capability with user preference, and proposes three distinct models — Capability, Capacity, and Preference — that together provide a more powerful and flexible foundation for adaptive user interfaces. The paper begins by examining how existing systems, from mobile phone settings to the Access For All standard, record user preferences (screen reader choice, font size, Braille device configuration) rather than underlying user capabilities. While preferences tell a system what the user wants, they fail to capture what the user can do — a critical distinction when interfaces need to adapt dynamically to changing conditions. The authors argue that the Access For All approach, with its hierarchy of XML containers holding literal preference values, becomes unwieldy when users operate across multiple contexts (classroom, fieldtrip, home) because every setting must be duplicated for each context. The proposed alternative models capability as the effective capacity to interact within specific design spaces (visual, sonic, haptic), using quantifiable properties rather than either medical diagnoses or pre-selected solutions. Colour-blindness, for example, is modelled not as a medical condition (protanopia, deuteranomaly) nor as a list of colours to avoid, but as percentage values of colour perception and intensity across low, medium, and high frequency ranges — giving interface designers a direct, quantifiable input for adaptation decisions.
Key findings
The paper introduces three interconnected information models using Executable UML (Shlaer-Mellor method). The Capability Model organises properties by subject ontology (visual, sonic, haptic) with typed values, composite properties, precedence hierarchies, and grouping into templates for user-facing acquisition wizards. The Capacity Model maps these properties to context-specific settings for individual users or groups, crucially introducing functionally dependent settings through Actions triggered by external influences — enabling dynamic adaptation based on environmental changes or observed user behaviour. For example, a system could detect repeated keystroke errors and automatically adjust haptic key guard settings, or modify font sizes in response to changing ambient lighting. The Preference Model separates arbitrary user choices from capability, using Shlaer-Mellor bridges to connect preferences to settings across domains. The paper also introduces a versioning mechanism (the Adaptation Model, borrowed from the authors' CISNA work) that allows profiles to be composed incrementally — "Fred is like Jim except..." — supporting template-based initialization and efficient handling of users with complex, multi-dimensional disability profiles such as Multiple Sclerosis.
Relevance
This paper articulates a distinction that remains largely unaddressed in mainstream accessibility practice: the difference between what a user prefers and what a user can do. Modern operating systems still present accessibility as collections of preference settings (font size, contrast, screen reader on/off) rather than modelling underlying capabilities. The concept of functionally dependent settings — where one capability value changes in response to environmental or behavioural triggers — anticipates the kind of context-aware adaptation that mobile platforms are only now beginning to explore with features like automatic brightness and adaptive display. For practitioners building accessible systems, the paper challenges the assumption that static preference profiles are sufficient, particularly for mobile and context-variable use cases. The vision of user profiles as autonomous agents, capable of self-adjustment through game-theoretic negotiation, points toward a formal computer science foundation for accessibility that has yet to be fully realised.
Tags: user profiling · capability modeling · adaptive interfaces · user modeling · accessibility standards
Standards referenced: ISO/IEC 24751 · IMS Access For All