DeepSeek and Gemini Get First-Author Bylines in Chinese Journal

East China Normal University's flagship academic journal, the Journal of East China Normal University (Educational Sciences), just published a special issue on a large-scale social experiment it calls "AI First Author." The issue includes a recap of the experiment, academic reflections, and five winning papers — all signed by large language models like DeepSeek and Gemini as first authors.

The experiment was launched in September 2025 to explore a new paradigm for human-machine collaborative research. Over six months, it received 724 submissions from around the world, each following a strict rule: AI leads the writing, humans review and supervise.

The post-mortem report reveals a mixed bag. AI excels at sparking ideas, organizing information, and polishing text, but it also suffers from fabricated citations, hollow logic, and a lack of genuine innovation. On the plus side, AI-generated peer reviews showed some reliability; younger scholars adapted more easily to human-AI collaboration; and the technology helped level the playing field in terms of access to intelligence.

The five published papers demonstrate what AI can actually do in specific research areas. One uses agent-based modeling to simulate how teacher rotation policies could ironically lead to a "fairness rebound" effect over time. Another analyzes 12,000 dialog logs to unpack deep learning mechanisms in long-term interactions. A third employs an adversarial AI-Delphi method to explore how AI might reshape the research process itself. And an analysis of 86,000 students found that AI-customized instruction has no significant advantage over high-quality generic teaching.

Based on these findings, the research team introduced several new concepts: a "human guarantor system," the risk of "AI hegemony," and a "Collaboration Quotient (CQ)" to measure human-AI synergy.