Frontier analysis — Operationalizing Product Legibility — Product Genome / Canonical / Product Genome Research Execution Package Run 02

Knowledge boundary

• Source: content/projects/product-genome/canonical/product-genome-research-execution-package-run-02.md
• Status: complete
• Discipline: Theory
• Confidence: Moderate-High
• Primary objective: Operationalize MODEL-001 from RP-PROD-001: text L = f(A, D, I, E, F, C) Where: - A = actual action possibility - D = discoverability - I = interpretability - E = executability - F = feedback quality - C = consequence comprehension The prior model proposed these components but did not specify measurement units, aggregation rules, thresholds, or experimental procedure.
• Primary claims/evidence: Product Legibility is not an intrinsic property of an object. It is a measured relationship among: text Product × Task × User Population × Environment × Time A control can be highly legible for a trained technician standing in good light and illegible for an older first-time user wearing gloves under time pressure. Therefore, statements such as “this product is intuitive” or “this control is legible” are incomplete unless the task and population are named.
• Methodology: Failed because predictive models are conditional and limited to specific portions of interaction.
• Limitations/known uncertainties: Remaining Uncertainty: The protocol can reliably structure measurement, but three major questions remain: - How should scores be normalized across products with radically different tasks? - How should learning over repeated use alter the legibility profile? - Wh… Attempted Falsification: Visual form sometimes produces stable expectations across users, suggesting that product-level properties matter. However, stable cues do not eliminate dependence on bodily capability, task, and environment. Product properties influence le… Attempted Falsification: A purely questionnaire-based scale would be cheaper and could produce a single score. It fails because users may report confidence after incorrect actions, may not recall near-errors, and may adapt to poor design. Direct observation remain… Attempted Falsification: A weighted mean with very high weights on safety-relevant dimensions could reduce this problem. However, weights would need to vary by task consequence, effectively reintroducing gates and context-specific rules. Attempted Falsification: A product interaction could be decomposed into movement, choice, mental preparation, and response operators. This may predict task time after the correct method is known. It still cannot predict discovery, mistaken mental models, or whethe… Risks: 1. False precision: Numerical profiles may appear more validated than they are. 2. Context loss: Aggregation may detach results from the tested task and population. 3. Sampling bias: Young, able, technically experienced participants can in… Tool Limitations: - No direct access to original physical products during this research run - Some standards provide only limited public previews - Historical case records may omit failed use and maintenance experience

Five highest-value opportunities

Challenge and confidence decay

The source was challenged for construct validity, independent replication, boundary conditions, transfer, and comparative baselines. Confidence should decay when the source lacks a dated replication, when its technology or target population changes, or when later artifacts report contradictory findings. Revalidation is recommended before treating context-bound recommendations as universal.