Reproduce “Offloading Score: Measuring AI Reliance Through Counterfactual Workflows” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Vishakh Padmakumar, Lujain Ibrahim, Zora Zhiruo Wang, Jennifer Wang, Q. Vera Liao, Diyi Yang
The authors introduce offloading score to estimate how much work AI displaces in coding workflows. It compares observed workflow steps with steps an average person would need to complete subtasks without AI. Among 40 freelance developers, scores averaged 0.451 under one-hour deadlines and 0.308 under four-hour deadlines. This result concerns one reliance measure in a programming study, not productivity.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.