Toward AI-Resilient Assessment in Computer Science Courses in an AI-Native World Formalism and Construction

Anshumali Shrivastava · 2026

Read Toward AI-Resilient Assessment in Computer Science Courses in an AI-Native World Formalism and Construction by Anshumali Shrivastava (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-30655.txt.

Description

Anshumali Shrivastava designs course assessments that stay fair when students can use AI freely. In senior computer science courses, he argues, grading should reward AI-resilient skill: achieving outcomes beyond a strong AI baseline. A good assessment should also stop bigger private AI budgets from buying grades by themselves. His minimal formal framework has four parts: a real task, an executable evaluator, a declared AI baseline frontier, and a grading rule.

The grading rule uses Pareto surplus, the amount by which a submission beats the baseline's best trade-offs. Surplus gives a measurable certificate that the work achieves something the declared AI baseline did not already supply. Supporting protocol elements, such as design reports, ablations and oral checks, help interpret surplus as student skill. Extensions handle self-improving AI loops, budget neutrality and red teaming. A concrete example is a Bloom filter assignment built for Rice University courses.

Catalog record

CREATORAnshumali Shrivastava
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE68 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.30655

Browse the full library · About the Tower of Babel