Route divergent or linguistically weak AI-scored explanations to teachers as diagnostic signals for mindful human review¶

Strategy · All strategies
Evidence · no claims cited

Description¶

As a mitigation for language bias, the article suggests human–AI collaboration in which disagreements between a teacher and an AI scorer serve as diagnostic signals flagging explanations for closer inspection. It further proposes "a dedicated AI -based scoring approach that identi fies explanations of low linguistic quality and routes them to the STEM teacher with an explicit cue to assess conceptual understanding mindfully independently of linguistic quality". The article notes this design fails where both assessors make the same biased decision.

Design Implications¶

Context¶

Requirements¶

• A human assessor available to review flagged explanations
• A mechanism for detecting divergent or low-lingistic-quality cases

Constraints¶

• Whether such awareness-raising can overcome the language bias remains an open question for future research; the design fails where both assessors make the same biased scoring decision

Target Learners¶

• secondary-school physics students
• multilingual and international learners

Target Learning Goals¶

• fair scoring of conceptual understanding in text-based explanations

Affordances¶

• Language Bias Task Inherent Explanation Scoring

Related Strategies¶

• Awareness of Implicit Biases

Examples¶

-

Key Sources¶

• Markus S. Feser & Paul L. Tschisgale. (2026). AI-based scoring systematically underestimates conceptual understanding of linguistically weak students' explanations in physics. Physics Education, Leibniz Institute for Science and Mathematics Education, Kiel. https://orcid.org/0000-0001-8503-0951