Results¶

The Results page displays the outcomes of your AI jury evaluation, including quantitative results, detailed justifications, and reference information for future access.

Results Overview¶

Query Configuration Summary¶

The top section displays your evaluation setup: - Query Text: The question that was evaluated - Outcomes: Number of possible results - Iterations: How many evaluation cycles were run - Jury Members: Number of AI models that participated - Supporting Files: Count of included reference materials

If your query used a pre-built package, this shows the package's configuration rather than your local settings.

AI Jury Configuration Details¶

For package-based queries, you'll see detailed jury information: - Provider and Model: Which AI services were used - Runs: How many times each model evaluated the query - Weight: Relative influence of each model (shown as percentage)

Understanding Results¶

Results Bar Chart¶

The bar chart visualizes the AI jury's conclusions:

Reading the Chart¶

• X-axis: Shows your defined outcome labels
• Y-axis: Likelihood values (0 to 1,000,000)
• Bars: Height represents the jury's confidence in each outcome
• Colors: Different colors distinguish between outcomes

Interpreting Values¶

• 1,000,000: Maximum confidence (100%)
• 500,000: Moderate confidence (50%)
• 0: No confidence (0%)
• Total: All values sum to 1,000,000 (100%)

Percentage Calculation¶

Hover over bars to see percentages: - Percentage = (Bar Value ÷ 1,000,000) × 100%

Example Interpretations¶

Scenario 1: Clear Decision

Outcome A: 850,000 (85%)
Outcome B: 150,000 (15%)

Scenario 2: Close Call

Outcome A: 520,000 (52%)
Outcome B: 480,000 (48%)

Scenario 3: Multiple Outcomes

Outcome A: 100,000 (10%)
Outcome B: 600,000 (60%)
Outcome C: 300,000 (30%)

AI Jury Justification¶

Understanding the Justification¶

The justification section contains the AI jury's reasoning: - Detailed Analysis: Step-by-step evaluation process - Evidence Review: How supporting materials influenced decisions - Consensus Building: How different models reached agreement - Uncertainty Discussion: Areas where models disagreed or were uncertain

Justification Features¶

Pagination¶

• Long justifications are split into readable pages
• Navigate using page controls at the bottom
• Page numbers show current position

Content Types¶

Justifications may include: - Executive Summary: High-level conclusions - Detailed Analysis: In-depth reasoning - Evidence Citations: References to your supporting materials - Model Perspectives: Individual model viewpoints - Confidence Levels: Certainty indicators for different aspects

Reading Tips¶

• Start with Summary: Look for conclusion sections first
• Check Evidence Usage: See how your supporting materials were used
• Note Disagreements: Areas where models differed indicate uncertainty
• Consider Context: Relate findings back to your original query

Reference Information¶

Content IDs (CIDs)¶

Two important CIDs are displayed:

Query Package CID¶

• Purpose: References your original query and configuration
• Use Cases:
• Share your query setup with others
• Rerun the same evaluation later
• Reference in documentation or reports
• Access: Click the link to view on IPFS

Result CID¶

• Purpose: References the evaluation results and justification
• Use Cases:
• Share results with stakeholders
• Reference in future analyses
• Archive evaluation outcomes
• Access: Click the link to view raw results on IPFS

Copying CIDs¶

• Click the 📋 button next to any CID to copy it to clipboard
• Use copied CIDs in other applications or documentation
• CIDs are permanent references to specific content

Evaluation Timestamp¶

When available, the timestamp shows: - Date and Time: When the evaluation was completed - Time Zone: Includes local time zone information - Precision: Accurate to the second

Viewing Past Results¶

Looking Up Previous Evaluations¶

Use the "View Past Results" section to access historical data:

Finding Result CIDs¶

• Check your browser history for previous sessions
• Look in blockchain transaction records
• Reference saved documentation or notes
• Ask colleagues who may have the CID

Loading Process¶

1. Enter the Result CID in the text field
2. Click "Load Results"
3. Wait for the system to fetch and parse the data
4. Results will replace current display

What Gets Loaded¶

• Outcome Distribution: Bar chart with historical results
• Justification: Original AI jury reasoning
• Configuration: Query setup information
• Timestamp: When the evaluation occurred

Managing Multiple Results¶

Comparing Evaluations¶

• Load different results in separate browser tabs
• Note differences in configurations and outcomes
• Look for patterns across similar queries

Result Organization¶

• Save important Result CIDs in external documentation
• Create naming conventions for different query types
• Maintain a log of evaluations for reference

Taking Action on Results¶

Next Steps Based on Outcomes¶

Clear, Confident Results¶

• High Consensus: Proceed with confidence in the decision
• Document Rationale: Save justification for future reference
• Implement Decision: Act on the AI jury's recommendation

Uncertain or Close Results¶

• Gather More Data: Add supporting materials and re-evaluate
• Refine Query: Clarify ambiguous aspects of your question
• Increase Jury Size: Add more AI models for additional perspectives
• Multiple Iterations: Run more evaluation cycles

Unexpected Results¶

• Review Justification: Understand the AI reasoning
• Check Supporting Data: Ensure materials were interpreted correctly
• Consider Bias: Look for potential model or data biases
• Seek Second Opinion: Run similar queries with different configurations

Starting New Evaluations¶

New Query Button¶

Click "New Query" to: - Clear current results - Return to Query Definition page - Start fresh with a clean configuration - Keep current contract and wallet settings

Building on Previous Queries¶

• Use insights from current results to improve future queries
• Reference successful configurations for similar evaluations
• Learn from justifications to ask better questions

Best Practices¶

Result Interpretation¶

• Consider Context: Relate results back to your original decision needs
• Read Justifications: Don't rely solely on numerical outcomes
• Note Confidence: Pay attention to how certain the AI jury was
• Check Evidence Usage: Ensure your supporting materials were properly considered

Documentation and Sharing¶

• Save Important CIDs: Keep records of significant evaluations
• Document Context: Note why you ran the evaluation and how you used results
• Share Appropriately: Consider what information is suitable for different audiences
• Archive Results: Maintain long-term access to important decisions

Continuous Improvement¶

• Learn from Patterns: Notice what types of queries work best
• Refine Techniques: Improve how you structure queries and supporting data
• Experiment with Configurations: Try different jury setups for comparison
• Build Expertise: Develop skills in interpreting AI jury evaluations

Troubleshooting¶

Display Issues¶

• Missing Chart: Refresh page if bar chart doesn't load
• Broken Links: Check internet connection for IPFS access
• Loading Errors: Verify CIDs are correctly formatted

Data Interpretation¶

• Unexpected Results: Review query clarity and supporting data quality
• Missing Justification: Check if evaluation completed successfully
• Confusing Reasoning: Consider if your query was ambiguous

Access Problems¶

• CID Not Found: Verify the CID was copied correctly
• Slow Loading: IPFS access may be temporarily slow
• Format Errors: Ensure you're using Result CIDs, not Query Package CIDs for past results