AI in Assessment, Grading, and Feedback: Evidence, Limits, and Recommendations for Institutional Practice

Artificial intelligence (AI) is increasingly used in higher education to support grading and assessment. Instructors may use these tools to generate draft comments, suggest rubric-aligned scores, and assist with evaluation. Research shows measurable benefits in specific areas, particularly mechanical revision, drafting support, and structured feedback. Evidence supporting AI use in autonomous or summative grading, however, remains limited, with concerns related to reliability, cognitive load, over-reliance, transparency, and data governance. AI can support instructional workflows when faculty remain actively engaged and transparent, but current findings do not support transferring evaluative authority or automating final grading decisions. Evidence supports limited use of AI in clearly defined formative and assistive contexts while cautioning against autonomous use in summative grading. Boise State can protect the integrity of assessment by emphasizing human oversight, transparency, alignment with data privacy standards, and continued attention to assignment design and faculty development. Existing studies are often small in scale and concentrated in specific contexts, and the research reviewed here reflects scholarship published through March 2026.