Popular Epistemology Digest — 26 August 2026
News and views on the sources and systems behind what we know and believe.
Consistently Good vs. Occasionally Great: A Rubric for Open-Ended Feedback Quality from Humans and Machines
2026-08-25 · arXiv (cs.CY)
Researchers developed a five-criteria rubric grounded in educational literature to evaluate the quality of feedback on open-ended short answer questions in introductory programming courses. The rubric is aimed at assessing feedback that guides students toward success on reattempts without revealing correct answers, and is applied to feedback from both human and AI sources.
Unfolding the Interdisciplinary Complexities of Climate Science: Fuxi-Climate Foundational Model
2026-08-25 · arXiv (cs.CY)
Researchers introduce the Fuxi-Climate Foundation Model (CFM), a climate-specialized large language model designed to support structured interdisciplinary reasoning across physical, socio-economic, and policy dimensions of climate science. The model addresses limitations of general-purpose LLMs in synthesizing climate knowledge for research and decision-making.
Evaluation in the Age of AI: Output as Evidence of Learning
2026-08-25 · arXiv (cs.CY)
The paper examines how the widespread adoption of large language models has disrupted traditional assessment methods in higher education, as tasks such as essay writing and coding can now be generated by AI with minimal human effort. The authors raise ethical questions about how educators should evaluate genuine student learning in this changed environment.
Large Language Models Simulate Intersectional Synthetic Identities with a Budget of One to Two Dimensions
2026-08-25 · arXiv (cs.CY)
A study tested standard demographic-persona prompting methods across eight large language models against 21 million simulated response distributions drawn from 15 waves of Pew's American Trends Panel. The findings indicate that while real respondents' opinions across intersectional subgroups are approximately additive across single-identity components, the models fail to capture this pattern accurately.
Expectations and Practices around AI Disclosure in CS Research
2026-08-25 · arXiv (cs.CY)
The paper investigates AI disclosure policies at top computer science publishing venues and examines whether current policies and practices align with their intended purpose. The study finds that despite the prevalence of such policies, significant gaps remain.
Enabling Organisational Change Through Ground-Up Initiatives: A Case Study from the STFC Scientific Computing Department
2026-08-25 · arXiv (cs.CY)
This paper presents a case study from the Science and Technology Facilities Council's Scientific Computing Department on translating high-level sustainability strategies into concrete actions toward UK Net Zero targets. The department, comprising over 200 staff, used ground-up initiatives to drive organisational change in its digital research infrastructure.
Whose Readiness Counts? Disagreement Within and Between Sectors in Perceived AI and Robotics Preparedness
2026-08-25 · arXiv (cs.CY)
Using a card-based survey in which 982 respondents provided 15,200 readiness evaluations across 17 named AI and robotics applications, the study examines how much information is lost when AI and Industry 4.0 preparedness is summarized as a single score for an organization or sector. The results reveal meaningful disagreement both within and between sectors that aggregate scores conceal.
