experiment EX-VE-X-002
Accessibility Without Modes
EX-VE-X-002: Accessibility Without Modes
Question
Does a calibrated mechanism-aware visual profile outperform conventional named accessibility modes while preserving semantic consistency and user control?
Conditions
A. strong universal default;
B. conventional named presets, for example large text, high contrast, dyslexia-oriented typography, or color-vision palette where such presets are normally offered;
C. user-selected manual controls;
D. calibrated Perceptual Envelope profile.
Do not intentionally construct poor presets to make condition D win.
Calibration
Calibration may estimate:
- useful typography ranges;
- preferred polarity and luminance range;
- spacing sensitivity;
- color-category discrimination;
- tolerance for density;
- motion reduction preferences.
Calibration must expose uncertainty and must allow manual override.
Tasks
Use at least:
- continuous reading;
- identifier recognition;
- semantic status recognition;
- dense comparison;
- navigation and recovery from error.
Outcomes
Primary:
- task accuracy;
- comprehension where applicable;
- completion time.
Secondary:
- fatigue;
- confidence;
- preference;
- mode switching;
- manual overrides;
- abandonment;
- calibration burden.
Semantic invariants
The experiment fails implementation review if personalization changes the meaning of:
- success;
- warning;
- error;
- unknown;
- selected;
- disabled;
- legal/illegal action;
- completion;
- uncertainty.
Presentation may adapt. Semantics may not.
Privacy
Prefer a local device profile containing visual parameters rather than medical labels.
Do not require users to disclose a diagnosis to obtain an adaptation.
Profiles should be:
- inspectable;
- editable;
- resettable;
- portable only by explicit user action;
- unnecessary for baseline accessibility.
Falsification
HY-VE-PE-005 is weakened if calibrated profiles do not outperform a strong universal default or manual controls by a practically meaningful amount, if calibration burden exceeds the benefit, or if profile instability creates errors.
A null result may imply that good universal defaults plus direct user controls are preferable to automatic adaptation.
Differentiation test
This experiment is the direct test of whether Visual Engineering should become an adaptive visual system or remain a rigorous evidence-based design method without personalization.