AutoResearch over a Reflexive Corpus: Design Note for the First Automated Research Experiment

Key Points

• The research aims to examine the differences in coding AI provenance labels done by authors versus those done by assistants.
• Conducted a comparative analysis of author-coded and assistant-coded provenance labels.
• Utilized the Reflexive Laboratory corpus for analysis.
• Evaluated contributions classified as human, cyborg, or AI.
• Identified systematic divergence in interpretations of contributions.
• Showed that assistant-coded labels differ from author-coded ones.
• Highlighted varying perspectives on AI contributions across the corpus.