Stronger AI Makes Deep Thinking More Crucial

机器之心 · wechat · 2026-07-14

This article discusses Fudan University's 2026 Blue Book on Intelligent Development in Humanities and Social Sciences, exploring why "deep thinking" becomes more critical once AI enters scientific research, governance, and knowledge production.

The article argues that while AI significantly boosts efficiency in literature review, data processing, paper writing, and tool usage, "being able to do" doesn't equal "understanding." In climate-social systems, academic research, and public governance, the real challenge lies in asking good questions, establishing true mechanisms, and verifying evidence chains, rather than merely processing more material.

The Blue Book specifically highlights the risks of AI in research and decision-making: automated model searches might amplify the "p-hacking" (testing until significant) problem; if governance systems merely stop at a formal "human-in-the-loop," manual reviews could devolve into accountability theater. It advocates integrating evidence chains, version histories, audits, and human judgments into the workflow, while distinguishing different risk scenarios like information retrieval, assisted judgment, and agentic decision-making.

The article concludes that the relationship between AI and the humanities/social sciences should shift from "one-way empowerment" to "two-way integration": technology answers "what can be done," while the humanities and social sciences continue to probe "why do it, how far should it go, and who bears the cost."

Original post →

More from AGI Musings

AGI Musings channel →