BRIDGE the Gap: Mitigating Bias Amplification in Automated Scoring of English Language Learners via Inter-group Data Augmentation

Researchers developed BRIDGE, a framework to reduce bias in AI-powered automated scoring systems that unfairly penalize English Language Learners (ELLs). The system addresses representation bias by generating synthetic high-scoring ELL samples, achieving fairness improvements comparable to using additional human data while maintaining overall performance.

• →AI scoring systems amplify bias against English Language Learners due to scarcity of high-scoring ELL samples in training data.
• →The BRIDGE framework synthesizes high-scoring ELL samples by combining content from high-scoring non-ELL responses with authentic ELL linguistic patterns.
• →Testing on California Science Test datasets showed BRIDGE reduces prediction bias for high-scoring ELL students without compromising overall scoring accuracy.
• →The method achieves fairness gains comparable to using additional real human data, offering a cost-effective solution for equitable assessment.
• →A discriminator model ensures quality control of the synthetically generated training samples.