Overall Score
{{ "%.3f"|format(overall_fairness_score) }}
Protected Attributes
{{ total_attributes }}
Warnings
{{ total_warnings }}
Critical Issues
{{ total_critical }}
📊 Dataset Overview
Sample Size
{{ dataset_info.total_samples|default(0)|int|format_number }}
total observations
Target Distribution
{% if dataset_info.target_distribution %}
{% for class_label, class_data in dataset_info.target_distribution.items() %}
Class {{ class_label }}:
{{ class_data.count|format_number }}
({{ "%.1f"|format(class_data.percentage) }}%)
{% endfor %}
{% else %}
No target distribution available
{% endif %}
{% if dataset_info.target_distribution %}
{% endif %}
Protected Attributes Distribution
{% if dataset_info.protected_attributes_distribution %}
{% for attr_name, attr_data in dataset_info.protected_attributes_distribution.items() %}
{{ attr_name|capitalize|replace('_', ' ')|title }}
Unique Values:
{{ attr_data.unique_values }}
{% for value, value_data in attr_data.distribution.items() %}
{{ value }}
{{ value_data.count|format_number }}
({{ "%.1f"|format(value_data.percentage) }}%)
{% endfor %}
{% endfor %}
{% else %}
No protected attributes distribution available
{% endif %}
⚙️ Test Configuration
Configuration Profile
{{ test_config.config_name|upper|default('CUSTOM') }}
Metrics Tested
{{ test_config.metrics_tested|length|default(0) }}
metrics
Features Enabled
{{ '✓' if test_config.pretrain_enabled else '✗' }} Pre-training Metrics
{{ '✓' if test_config.confusion_matrix_enabled else '✗' }} Confusion Matrix
{{ '✓' if test_config.threshold_analysis_enabled else '✗' }} Threshold Analysis
{% if test_config.age_grouping_enabled %}
📅 Age Grouping Configuration
Strategy:
{{ test_config.age_grouping_strategy|upper }}
{% if test_config.age_grouping_details %}
{% for detail in test_config.age_grouping_details %}
Attribute:
{{ detail.attribute }}
Original Range:
{{ detail.original_range }}
Groups Created:
{% for group in detail.groups %}
-
▸ {{ group }}
{% endfor %}
{% endfor %}
{% endif %}
{% endif %}
📋 Metrics Breakdown
{% for metric in test_config.metrics_tested %}
• {{ metric|replace('_', ' ')|title }}
{% endfor %}
Pre-Training Metrics (Model-Independent)
These metrics evaluate bias in the data before training. They don't depend on the model, only on data and protected attributes.
4 Implemented Metrics:
- Class Balance (BCL) - Sample imbalance between groups
- Concept Balance (BCO) - Difference in positive class rate
- KL Divergence - Difference in label distributions
- JS Divergence - Symmetric version of KL Divergence
📋 Detailed Metrics by Protected Attribute
{% for attr in protected_attributes %}
📊 {{ attr.name|capitalize }}
{% if attr.pretrain_metrics %}
{% for metric in attr.pretrain_metrics %}
{{ metric.name }}
{{ "%.4f"|format(metric.value) }}
{{ metric.interpretation }}
{% if metric.groups %}
Groups:
{% for group_name, group_info in metric.groups.items() %}
{{ group_name }}: {{ group_info.count }} ({{ "%.1f"|format(group_info.percentage) }}%)
{% endfor %}
{% endif %}
{% endfor %}
{% else %}
Nenhuma métrica pré-processamento disponível para este atributo.
{% endif %}
{% endfor %}
Main Post-Training Metrics (Compliance-Critical)
The 5 most critical metrics for legal and regulatory compliance. All include minimum representativity filter (MIN_REPRESENTATION_PCT = 2.0%).
⚖️ 5 Critical Compliance Metrics:
- Statistical Parity - Equal positive prediction rate (EEOC 80% rule)
- Equal Opportunity - Equal TPR between groups (fair benefit)
- Equalized Odds - Equal TPR and FPR (more restrictive)
- Disparate Impact ⚖️ - LEGAL CRITICAL - Ratio ≥ 0.8 (EEOC)
- False Negative Rate Difference - Miss rate between groups
📋 Detailed Metrics by Protected Attribute
{% for attr in protected_attributes %}
📊 {{ attr.name|capitalize }}
{% if attr.posttrain_main %}
{% for metric in attr.posttrain_main %}
{{ metric.name }}
{% if 'disparate_impact' in metric.name.lower() %}
⚖️ LEGAL
{% endif %}
{% if metric.disparity is defined and metric.disparity is not none %}
Disparity: {{ "%.4f"|format(metric.disparity) }}
{% endif %}
{% if metric.ratio is defined and metric.ratio is not none %}
Ratio: {{ "%.4f"|format(metric.ratio) }}
{% endif %}
{% if metric.value is defined and metric.value is not none and metric.disparity is not defined %}
Value: {{ "%.4f"|format(metric.value) }}
{% endif %}
{{ metric.interpretation }}
{% if metric.testable_groups %}
Testable Groups (≥ {{ metric.min_representation_pct }}%): {{ metric.testable_groups|length }}
{% if metric.excluded_groups %}
Excluded Groups (< {{ metric.min_representation_pct }}%): {{ metric.excluded_groups|length }}
{% endif %}
{% endif %}
{% endfor %}
{% else %}
No main post-training metrics available for this attribute.
{% endif %}
{% endfor %}
Complementary Metrics
6 additional metrics for in-depth fairness analysis. Complement the main metrics.
6 Complementary Metrics:
- Conditional Acceptance - Precision/PPV by group
- Conditional Rejection - NPV by group
- Precision Difference - Precision difference between groups
- Accuracy Difference - Accuracy difference between groups
- Treatment Equality - Balance FN/FP ratio
- Entropy Index - Individual fairness (doesn't use sensitive attributes)
📋 Detailed Metrics by Protected Attribute
{% for attr in protected_attributes %}
📊 {{ attr.name|capitalize }}
{% if attr.posttrain_complementary %}
{% for metric in attr.posttrain_complementary %}
{{ metric.name }}
{% if metric.value is defined and metric.value is not none %}
{{ "%.4f"|format(metric.value) }}
{% elif metric.disparity is defined and metric.disparity is not none %}
Disparity: {{ "%.4f"|format(metric.disparity) }}
{% elif metric.ratio is defined and metric.ratio is not none %}
Ratio: {{ "%.4f"|format(metric.ratio) }}
{% endif %}
{{ metric.interpretation }}
{% if metric.testable_groups %}
Testable Groups: {{ metric.testable_groups|length }}
{% if metric.excluded_groups %}
Excluded: {{ metric.excluded_groups|length }}
{% endif %}
{% endif %}
{% endfor %}
{% else %}
No complementary metrics available for this attribute.
{% endif %}
{% endfor %}
{% if has_confusion_matrix %}
Confusion Matrices by Group
Detailed breakdown of prediction outcomes for each demographic group.
{% endif %}