PHRAIL: Philosophy Research for AI Literacy

PHRAIL (“frail”) is a research group at UC Merced organized by Prof. D. Hicks.

Membership

In Spring 2026, PHRAIL is closely linked to Prof. Hicks’ Philosophy of Science course. Membership is currently open only to students currently enrolled in that course.

Some core assumptions

Normal technologies
Generative systems are normal technologies (Narayanan and Kapoor 2025). Widespread adoption and diffusion of a normal technology requires sociotechnical infrastructure that takes years or decades to develop and construct.
Social construction of technology
Technology develops through a process of “alternati[ng] … variation and selection” (Pinch and Bijker 1984). Different social groups respond to novel technologies by developing variations more-or-less fit to their concerns and interests. Some of these variations can survive selection in the broader social context. As a normal technology, generative systems can be expected to develop through this same process.
Neither hype nor skepticism
The capabilities of generative systems have been incredibly overhyped, potentially leading to a massive speculative bubble. At the same time, generative systems are likely to have significant, legitimately valuable uses in knowledge work. In line with the social construction of technology, these uses must be developed rather than discovered.
Affordance mismatch
These uses are more obscured than revealed by the standard chatbot interface. This interface was inspired by a crude misreading of Turing (1950), and we’ve known for decades that this interface affords delusional beliefs about the system’s nature and capabilities (Weizenbaum 1966; Anthony 2025).
Need for sociotechnical expertise
Identifying the actual capabilities of generative systems — and projecting their potential ethical, legal, and social implications, both beneficial and harmful — requires a combination of technical expertise — understanding how the systems work — and socio-ethical expertise — understanding how people work.
Role of humanities and social science students
Many students in natural science, math, and engineering programs have little or no exposure to the relevant fields of humanities and social sciences. Humanities and social science students are well-prepared on the socio-ethical side, but may feel intimidated by the technical details of generative systems. While some math background is important for certain aspects of generative systems — for example, understanding the architecture of generative models requires linear algebra — many aspects are broadly accessible. Therefore, one important task of PHRAIL is to cultivate the intellectual curiosity and confidence necessary for humanities and social science students to engage with the technical details of generative systems.
Excess of objectivity in environmental debates
Power and water use are two major sites of contention in the public controversy surrounding generative models. These environmental debates suffer from an “excess of objectivity” (Sarewitz 2004): because lifecycle analyses are highly dependent on a wide range of reasonable assumptions, different sides in the debate can easily come up with reasonable estimates that are orders of magnitude apart (Green 2025). So these debates are proxies (Hicks 2017) for the basic emotivist positions in the broader controversy: “I disapprove of AI; do so as well” and “I approve of AI; do so as well.”

Activities

Phil Sci RAG

In Spring 2026, we’ll be starting off by experimenting with Phil Sci RAG, a retrieval-argumented generation system using Google’s NotebookLM. The RAG contains “suggested readings” for Philosophy of Science: about 15 books or articles for each assigned reading in the course.

We’ll first take some time to understand what a RAG is and how it’s different from the typical chatbot interface. Our next goal will be to develop rubrics for evaluating generated summaries of academic articles, for use in one of the course assignments.

Process notes

  • process notes
    • fortnightly memo to group
    • what you’ve been doing, reading, thinking about wrt generative systems
    • incorporate screenshots, chat logs,
    • what you noticed, esp. if it conflicted with things you or others might expect
    • questions for future thought, reading, or experimentation

References

Anthony, Sam. 2025. “‘AI’ Is Bad UX.” Apperceptive. December 22, 2025. https://buttondown.com/apperceptive/archive/ai-is-bad-ux/.
Green, Hank, dir. 2025. Why Is Everyone So Wrong About AI Water Use?? https://www.youtube.com/watch?v=H_c6MWk7PQc.
Hicks, Daniel J. 2017. “Scientific Controversies as Proxy Politics.” Issues in Science and Technology, January 2017. https://www.jstor.org/stable/24891967.
Narayanan, Arvind, and Sayash Kapoor. 2025. “AI as Normal Technology.” http://knightcolumbia.org/content/ai-as-normal-technology.
Pinch, Trevor J., and Wiebe Bijker. 1984. “The Social Construction of Facts and Artefacts: Or How the Sociology of Science and the Sociology of Technology Might Benefit Each Other.” Social Studies of Science 14 (3): 399–441. http://www.jstor.org/stable/285355.
Sarewitz, Daniel. 2004. “How Science Makes Environmental Controversies Worse.” Environmental Science and Policy 7 (5): 385–403. https://doi.org/10.1016/j.envsci.2004.06.001.
Turing, A. M. 1950. “Computing Machinery and Intelligence.” Mind 59 (236): 433–60. http://www.jstor.org.proxy1.lib.uwo.ca/stable/2251299.
Weizenbaum, Joseph. 1966. “ELIZA—a Computer Program for the Study of Natural Language Communication Between Man and Machine.” Communications of the ACM 9 (1): 36–45. https://doi.org/10.1145/365153.365168.

Reuse