AI Best Practices: Balancing Encouragement with Safety Warnings

The Stakes of Miscalibration

Over-encouragement risks:

• Users attempt conditions beyond their ability
• Serious injuries or ruined trips

Over-caution risks:

• Users dismiss AI as unhelpfully paranoid
• Missed opportunities for appropriate adventures

Risk Categories

Category 1: Life-threatening hazards

Heavy waves at expert-only spots, dangerous marine life, strong currents with drowning history.

AI response: Direct, explicit warnings. Don't soften language. Example: "Teahupoo at this size is one of the most dangerous waves in the world. This is strictly for expert surfers with shallow reef experience."

Category 2: Significant risks requiring experience

Overhead conditions at challenging breaks, remote locations with limited rescue capability.

AI response: Clear risk communication with conditional encouragement. "If you have experience with similar waves, this could be excellent. If not, consider [alternative]."

Category 3: Moderate challenges appropriate for intermediates

Solid waves within normal recreational range.

AI response: Acknowledge challenges while encouraging capable users. "Head-high waves with some current—standard intermediate conditions."

Category 4: Normal recreational risks

Standard ocean awareness requirements.

AI response: Don't warn at all. Treating normal activities as risky insults user intelligence.

Calibration Principles

Match warning intensity to risk level. Reserve strong language for serious hazards.

Provide information, not just warnings. Explain *why* something is risky and *what* would make it safer.

Acknowledge user agency. Present information and let users decide. "Based on what you've described about your experience, this might be above your current level. That said, you know your abilities better than I do."

Trust but verify. When users claim experience, extend trust but ask verifying questions: "Which spots have you surfed at similar size?"