Synthetic interactive prototype · Independent portfolio study
Geospatial AI Heuristic Loss Dashboard
A controlled comparison environment for making common AI-generated geospatial interface failures visible, inspectable, and easier to correct.
Interactive demonstration
Compare the same synthetic context in flawed and optimized states.
Use the state switcher, inspect the map and records, and compare changes in aggregation, contrast, information hierarchy, and action clarity. The displayed score is an illustrative design proxy, not a measured human outcome.
If the embedded demonstration does not load, open it full screen in a new tab.
Purpose and method
Turn subjective critique into an explicit review system.
- Problem
- Generated interfaces can appear polished while weakening map legibility, exception visibility, comparison, and spatial-query continuity.
- Method
- Hold the synthetic context constant, expose a small set of heuristic failures, and compare them with deliberately corrected states.
- Interaction
- The state switch makes design consequences visible without implying that a numerical proxy is a validated usability score.
- Design value
- The prototype connects visual inspection to review language that can support critique, redesign, and a future validation plan.
Evidence boundary
Working prototype, deliberately bounded claims.
This is a synthetic, independent portfolio prototype. It demonstrates an implemented comparison interaction and documented design reasoning. It is not a deployed AI model, live GIS feed, client system, trained RLHF policy, human-factors experiment, accessibility certification, or safety result.