Practices

Alarm-and-warning evaluation is a practice for assessing whether a system's alerts, warnings, and alarm states communicate correctly under the conditions in which the system is used. It distinguishes technical visibility from operational effectiveness: an alarm can appear on screen and still fail if it cannot be perceived, prioritised, interpreted, or acted on in time.

Behavioural requirements definition translates trust, oversight, and governance analysis into specific requirements for how an interface handles AI outputs. It defines what the product must show, prevent, record, and require from human users when AI recommendations or outputs are involved.

Cognitive load analysis is a diagnostic practice for separating irreducible task complexity from cognitive demand introduced by an interface. It produces specific findings about where extraneous load accumulates and how that load affects different roles and operating conditions.

Decision boundary design is a practice for assigning AI-assisted product decision points to appropriate control categories and implementing the interface behaviours that make those boundaries meaningful. It covers boundary identification, boundary calibration, and boundary implementation, with emphasis on consequence, reversibility, confidence, governance requirements, and operational context.

Design governance through delivery keeps design decisions from fragmenting during implementation. It treats the design system as a reasoning record, provides structured development support, reviews implementation divergence, and prepares the client team to maintain the system independently.

Edge-case-and-degraded-mode analysis is a practice for designing interfaces that remain understandable under fault states, environmental stress, abnormal data, unusual configurations, and partial system operation. It complements nominal-case design by producing a condition inventory and specific design requirements for each non-nominal state.

Error-likely interaction review examines interface and workflow conditions that make errors foreseeable, including memory dependency, ambiguous state, insufficient confirmation friction, information absence, single-channel communication, and foreseeable misuse patterns.

Evidence-led prioritisation makes the evidence behind competing design options visible and comparable. It produces a structured ranking that combines expected operational effect with calibrated evidential confidence.

State-and-transition review is a structured practice for cataloguing system states, mapping transitions, checking spatial consistency, identifying mode changes, and evaluating state communication under the conditions in which users actually operate the system.

Task criticality mapping identifies which tasks are critical to a system's operational purpose, then specifies why they are critical, what successful completion means operationally, and which conditions must hold for success.

Trust-and-oversight analysis identifies interface conditions that cause users to over-trust or under-trust AI outputs, and it assesses whether governance consumers can verify AI behaviour. The practice is used in Sandbox Experiments, Concept Convergence, and before enterprise or regulated deployment contexts.

Usability testing under realistic constraint changes participant selection, task design, environment design, and timing so that usability evidence reflects the conditions that determine operational performance. Creative Navy applies the practice as formative evaluation during design, not as final summative validation.

Workflow analysis is a research practice for reconstructing how work is actually done in complex professional systems. It is used before design decisions and during audits to identify the gap between designed workflows and actual operational behaviour.