At the Crossroad: AI in Science Assessment

On January 30, 2026, Dr. Xiaoming Zhai delivered a powerful keynote at the 9th IEEE Eurasian Conference on Educational Innovation, titled “At the Crossroad: AI in Science Assessment”. He urged the education community to move beyond the sidelines and proactively shape AI applications to ensure they align with core educational values. While acknowledging the promises of AI, Dr. Zhai also focused on navigating its potential pitfalls, particularly the need for rigorous investigation into algorithmic bias and fairness in assessment. 

In his keynote remarks, Dr. Zhai showcased the National GENIUS Center’s focus on using generative AI and large language models to transform science education. The National GENIUS Center is an example of how AI can be a strategic partner in the classroom, with its development of multi-agent environments (such as EngageAgent and LabAgent) to support student engagement in science practices and provide professional resources to help educators leverage GenAI for assessment and classroom interactions. 

The key takeaways from the presentation for educators and researchers included: 

• Shift from “Whether” to “How”: The conversation has moved beyond whether to use AI; the focus is now on determining for what purposes it should be used and how to implement it effectively. 
• AI as a Partner, Not a Replacement: Tools like ArguAgent, developed by the National GENIUS Center, can serve as a “partner in the classroom,” multiplying a teacher’s presence by managing structured argumentation at scale and nudging students toward evidence-based reasoning. 
• The Multi-Agent Advantage: Moving toward multi-agent solutions can significantly reduce errors like “over-praise” or “over-inference” common in single-agent AI feedback systems. 
• Addressing “Pseudo AI Bias”: Educators must distinguish between actual algorithmic bias and “pseudo-mechanical bias,” which often stems from disconnected interpretations of labels or unrealistic expectations of the technology. 
• Proactive Research: To serve learners—not just systems—the education community must co-create solutions that prioritize human-centered design and intellectual merit.