Interfaces must prevent errors before they occur through careful design eliminating error-prone conditions, constraining invalid actions, providing intelligent defaults guiding safe choices, validating inputs before acceptance, and requiring explicit confirmation for destructive operations—prevention proves fundamentally more effective than recovery because errors never enter systems, users maintain confidence avoiding mistake sequences, and cognitive load remains focused on productive tasks rather than error correction. Nielsen's usability heuristic #5 (1994) established "error prevention" as superior to error messages alone emphasizing that "even better than good error messages is a careful design which prevents a problem from occurring in the first place," validated through decades of research demonstrating prevention reduces support costs 40-60%, improves task completion 30-50%, and fundamentally transforms user confidence enabling exploration without fear of irreversible consequences.
Nielsen's foundational usability heuristic #5 "Error prevention" (1994) established prevention as superior to recovery through extensive evaluations demonstrating interfaces preventing errors outperform those merely handling errors gracefully. His research identified two error prevention approaches: eliminating error-prone conditions (designing interfaces making errors impossible through constraints, disabled states, limited options) and presenting confirmation options (checking for potentially problematic actions requesting user verification before execution). Nielsen's evaluations across diverse applications demonstrated that even excellent error messages prove inferior to prevention—users experiencing errors suffer interruption, reduced confidence, and cognitive overhead regardless of recovery ease, while prevented errors maintain flow enabling sustained focus on productive work.
Norman's The Design of Everyday Things (1988) provided theoretical foundation for error prevention through distinction between slips (correct intentions with incorrect execution—pressing wrong button, typing wrong key, selecting wrong menu item) and mistakes (incorrect intentions from faulty mental models—choosing wrong goal, misunderstanding system capabilities, applying inappropriate strategies). Norman's research demonstrated these error types require different prevention approaches—slips prevented through constraints limiting possible actions to valid choices, mode indicators showing current state, confirmation for irreversible actions. Mistakes prevented through clear affordances revealing system capabilities, visible conceptual models showing system organization, and immediate feedback validating actions match intentions before commitment.
Reason's Human Error (1990) established comprehensive error taxonomy distinguishing skill-based errors (slips and lapses during automatic performance—typing "teh" instead of "the," forgetting steps in familiar sequences), rule-based errors (applying wrong rules or misinterpreting situations—using procedures inappropriate for current context), and knowledge-based errors (operating beyond existing knowledge requiring improvisation—incorrect problem-solving when rules don't apply). His research demonstrated effective error prevention matches strategies to error types—skill-based errors prevented through constraints and forcing functions eliminating invalid actions, rule-based errors prevented through clear situation indicators and appropriate rule selection guidance, knowledge-based errors prevented through clear system feedback and exploratory safe environments enabling learning without permanent consequences.
Rasmussen's skill-rule-knowledge (SRK) framework (1983) explained human performance across expertise levels validating error prevention requirements varying with user proficiency. Skill-based performance (automated sensorimotor behavior) generates slips prevented through interface constraints and consistent mappings. Rule-based performance (conscious application of learned rules) generates errors when rules misapply prevented through clear situation recognition and rule guidance. Knowledge-based performance (conscious problem-solving without applicable rules) generates mistakes prevented through visible system models and safe exploration enabling learning. This framework validates prevention requiring different strategies for novice versus expert users—novices need strong constraints preventing exploration errors, experts need flexible systems preventing interruption while maintaining safeguards for truly dangerous actions.
Lewis and Norman's "Designing for error" (1986) established that error prevention serves multiple functions beyond immediate mistake avoidance: system design improvement (frequent errors indicate poor affordances or confusing workflows requiring redesign), user learning (prevention mechanisms teach correct usage through constraints and feedback), and confidence building (prevented errors demonstrate system protects users maintaining trust). Their research validated that overly restrictive prevention creates frustration limiting legitimate functionality, while insufficient prevention creates anxiety reducing exploration—effective prevention balances protection with freedom enabling confident productive work.
Contemporary research on confirmation dialogs and interruption costs (Bravo-Lillo et al. 2013) demonstrated that excessive confirmation creates habituation—users automatically dismiss warnings without reading becoming vulnerable to precisely the errors confirmations intended to prevent. Studies showed confirmation effectiveness depends on frequency (rare confirmations draw attention, frequent ones ignored), specificity (generic "Are you sure?" proves ineffective versus specific consequence descriptions), and user investment (confirmations after substantial work prove more effective than preventive early warnings). This research validates prevention requiring layered approaches from subtle constraints (always active preventing invalid states) through moderate validation (real-time feedback guiding corrections) to explicit confirmation (reserved for truly dangerous rare actions).