Delphi Risk Assessment Mockups

Delphi Risk Assessment Mockups

This collection demonstrates different approaches to visualizing and collecting expert assessments of AI risks. Each example explores different aspects of risk communication and data collection for Delphi method research.

Examples Overview

Example 1: Risk Exceedance Curve

Risk 3.1 - False or Misleading Information Exceedance Probability Curve Interactive probability curve showing likelihood of exceeding different severity thresholds with detailed data breakdown.

Example 2: Probability Bands Distribution

Risk 3.1 - Probability Distribution by Severity Level Horizontal band visualization showing probability distribution across all severity categories with interactive tooltips.

Example 3: Comparative Risk Ranking

AI Risk Comparative Ranking by Moderate Harm Probability Horizontal bar chart ranking all risks by probability of moderate or worse outcomes.

Example 4: Risk Progression Analysis

Multi-Risk Severity Progression with Interactive Selection Line chart showing probability trajectories across severity levels for all risks with responsive design.

Example 5: Expert Survey Interface

Interactive Risk Assessment Data Collection Tool Survey-style interface for experts to input probability assessments with real-time validation and interpretation.

Example 6: Probability Allocation Survey

AI Risk Assessment Survey - Probability Allocation Method Alternative survey interface using probability allocation across severity levels with cumulative probability calculations and real-time validation.

Design Principles

  • Cumulative Probability Logic: All visualizations respect that P(≥Severe) ≤ P(≥Moderate) ≤ P(≥Limited)
  • Responsive Design: All examples scale appropriately across viewport sizes
  • Interactive Feedback: Users receive immediate visual and textual feedback on their inputs
  • Validation: Built-in logical consistency checks for expert assessments
  • Accessibility: Clear typography, sufficient contrast, and intuitive navigation

Technical Implementation

  • Chart.js: Used for responsive, interactive visualizations
  • Vanilla JavaScript: Clean, dependency-minimal implementations
  • CSS Grid/Flexbox: Modern responsive layout techniques
  • Progressive Enhancement: Functional without JavaScript, enhanced with it

Risk Categories

The examples focus on AI risks across four main domains:

  1. Bias & Fairness (1.x): Discrimination, toxic content, unequal performance
  2. Privacy & Security (2.x): Data breaches, system vulnerabilities
  3. Misinformation (3.x): False information, ecosystem pollution
  4. Malicious Actors (4.x): Disinformation campaigns, cyberattacks, fraud

These mockups support research into expert elicitation methods for AI risk assessment and governance policy development.