David W. Hogg on why we do astrophysics (in the face of LLMs and the lack of clinical value)

2026-05-07 · Source: Statistical Modeling, Causal Inference, and Social Science · Field: Science & Research — Space Science & Astronomy, Artificial Intelligence & Machine Learning, Research Methodology & Innovation · Depth: Intermediate, medium

Summary

David W. Hogg's February 2026 arXiv rumination, "Why we do astrophysics?", critically examines the human element in scientific research, the role of Large Language Models (LLMs), and academic practices within astrophysics. Hogg asserts that LLMs "show no signs of intelligence" despite their conversational abilities, and questions traditional academic hierarchies, suggesting a data scientist with astronomy knowledge might be more effective than an astronomer with data science training. The paper delves into the nature of scientific literature, arguing that "astrophysics is the astrophysics literature," and challenges the notion that papers and citations are not "coin" for career advancement. Hogg also explores the implicit knowledge in scientific practice, advocating for its explicit documentation, and discusses the economic motivations behind scientific careers. He proposes two extreme policy recommendations for LLMs in astrophysics, "let-them-cook" and "ban-and-punish," arguing against both, and ultimately reframes the core inquiry from "how we do astrophysics" to "why we do astrophysics."

Key takeaway

For AI Scientists and Research Scientists evaluating the integration of LLMs into scientific workflows, Hogg's critique suggests caution against conflating LLM capabilities with scientific intelligence. You should focus on developing nuanced, moderate policies for LLM use, avoiding extreme "let-them-cook" or "ban-and-punish" approaches. Additionally, consider the value of interdisciplinary training, where domain expertise is augmented by strong data science skills, rather than the reverse.

Key insights

Hogg's paper challenges conventional views on LLM intelligence, academic publication, and the core motivations behind scientific research.

Principles

• Astrophysics is fundamentally its literature.
• Implicit knowledge should be made explicit.

In practice

• Consider publishing software in dedicated venues.
• Document implicit project knowledge explicitly.

Topics

• Astrophysics
• Large Language Models
• Scientific Practice
• Academic Publishing
• Data Science

Best for: AI Scientist, Research Scientist, AI Ethicist

Related on AIssential

• The uncritical adoption of AI in science is alarming — we urgently need guard rails — Uncritical AI adoption in science risks reduced research quality, narrowed scope, and deskilling future scientists.
• How to stay relevant as a software developer — Human developers can gain an edge over LLMs by staying current with pre-print research and diversifying their knowledge.
• Attributing and situating knowledge cannot be left to language models — LLMs cannot reliably handle scholarly attribution, risking miscitation and undermining research standards.
• Mar 23, 2026ScienceVibe physics: The AI grad student — AI can significantly accelerate expert-guided theoretical physics research, but requires rigorous human oversight.
• It's not the Language Model, it's the Tool: Deterministic Mediation for Scientific Workflows — Typed mediation uses LLMs to orchestrate deterministic, researcher-defined tools for reproducible scientific analysis, ensuring consistent results.
• The Fed Chair Just Said What AI Leaders Won't: The Models Don't Work — LLMs are not sufficient for complex system prediction, prescription, and diagnosis due to inherent architectural limitations.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Statistical Modeling, Causal Inference, and Social Science.