Agentic data science with the Data Agent Kit
cloud.google.com·7h ago
AI Research
arXiv CS.AI·Aug 11·5 min read
Positioning Generative Artificial Intelligence in STEM Assessment: When to Require, Scaffold, or Restrict Its Use
TL;DR
A framework for integrating Generative Artificial Intelligence (GenAI) in STEM assessments is proposed, balancing its use to enhance learning while maintaining assessment integrity. It outlines when to restrict, scaffold, or require GenAI based on task characteristics and learning objectives.
✦ Why It Matters
Evaluate your assessment strategies today to incorporate GenAI effectively and maintain educational integrity.
Key Takeaways
- Assess when to restrict GenAI to protect foundational knowledge.
- Use scaffolding to ease cognitive load in assessments.
- Incorporate GenAI for tasks requiring human-AI collaboration.
- Design assessments that prepare students for AI-enabled workplaces.
Full Summary
Generative Artificial Intelligence (GenAI) poses challenges for STEM assessments, as unrestricted use can compromise the validity of traditional evaluations, while outright bans may hinder students' readiness for AI-integrated workplaces. This paper introduces a student-focused framework based on Evidence Centered Design (ECD), which provides guidelines on when to restrict, scaffold, or require GenAI in assessments.
Restrictions are advised when GenAI could undermine the assessment of foundational knowledge and routine skills. Scaffolding is suggested when limited GenAI support can ease cognitive load without sacrificing understanding.
Requiring GenAI is appropriate for tasks that involve collaboration between humans and AI, emphasizing the need for AI literacy. The framework is illustrated with examples from introductory physics, demonstrating how to design tasks under varying GenAI policies, ultimately aiming to preserve educational integrity while preparing students for AI-enhanced environments.