adult-model-card

Adult Income Fairness Model

Model Details

Model Description

This model predicts whether a person's income belongs to the positive income class using the Adult dataset.

• Model type: Random Forest Classifier
• Framework: scikit-learn
• Developed for: Human-Centered AI tutorial

---

Intended Use

This model is intended for:

• fairness tutorials
• responsible AI demonstrations
• educational purposes

Out-of-Scope Use

This model should NOT be used for:

• hiring decisions
• loan approval
• insurance decisions
• healthcare decisions

---

Dataset

Dataset used:

• adult-clean.csv

Sensitive attribute:

• gender

Protected groups:

• Male
• Female

Target column:

• Class-label

Class interpretation:

• 1 = positive class
• 0 = negative class

---

Training Procedure

Preprocessing

• missing value imputation
• one-hot encoding

Classifier

• RandomForestClassifier
• n_estimators = 100
• random_state = 42

Train/Test Split

• test size = 20%
• stratified split enabled

Training Timestamp

• 2026-08-08T19:26:24.912828 UTC

---

Evaluation Results

---

Fairness Assessment

Fairness Metric

Statistical Parity (SP)

SP formula:

SP = P(Ŷ = 1 | Male) - P(Ŷ = 1 | Female)

Interpretation:

• SP ≈ 0 → more fair
• negative SP → more positive predictions for females

Observed SP:

• 0.1905

Group Statistics

---

Bias, Risks, and Limitations

This dataset may contain historical societal biases.

Known limitations:

• fairness evaluated only on SP
• possible demographic imbalance
• no causal fairness analysis
• fairness only evaluated on gender

Further fairness evaluations are recommended before deployment.

Recommended additional fairness metrics:

• Equal Opportunity Difference
• Equalized Odds
• Calibration
• Disparate Impact

---

Reproducibility

Software Versions

---

Environmental Impact

This model was trained on CPU for educational purposes.

---

Citation

If using this tutorial or model, please cite:

• Adult Dataset
• Hugging Face Hub