01
Epistemic Values and Scientific Integrity
What makes scientific knowledge trustworthy, and which properties are at stake when AI systems participate in producing itAccounts of scientific integrity from philosophy of science, STS, sociology, statistics, and the science of science, brought to bear on AI-integrated practiceWhether the values that constitute good science are shared across disciplines or specific to themReproducibility, robustness, and error control as epistemic rather than procedural propertiesWhat understanding, as distinct from prediction, requires of a scientific practiceValues that have no current technical operationalization, and what it would take to give them one
02
Automation, Judgment, and Division of Labor
Which parts of scientific work can be delegated, which cannot, and how to tell the differenceWhat scientific judgment consists in, and where it resists substitutionEffects of automation on training pipelines, apprenticeship, and early-career trajectoriesTacit knowledge, craft skill, and forms of expertise that do not survive formalizationGradual disempowerment and the long-run consequences of incremental delegationHistorical cases of instrumentation and automation reshaping a discipline, and what they predict here
03
Validity and Evaluation
Construct validity in scientific AI evaluation: what benchmarks and task-completion metrics actually measureMeta-evaluations and reproducible critiques of existing scientific benchmarks and agent evaluationsEvaluation designs that target epistemic properties rather than task successMeasuring properties that admit no ground truth, and what substitutes for itExpert disagreement in evaluating scientific work: when it signals invalid measurement and when it signals genuine pluralism
04
Safety and Alignment Failures
Failure modes documented in general AI safety research, examined for what they do to knowledge production specificallyReward hacking and specification gaming in scientific tasks, including optimization against proxies for scientific qualitySycophancy and the degradation of criticism, disagreement, and negative resultsDeceptive and situationally aware behavior where the objective is a scientific claimHomogenization of research questions, methods, and hypotheses across a fieldWhat scientific alignment would consist of, and whether science offers a tractable testbed for alignment more broadly
05
Explanation, Interpretability, and Understanding
What counts as an explanation, and what an explanation licenses a scientist to believeExplanatory virtues and standards of adequacy imported from philosophy of science into interpretability practiceWhether interpretability methods deliver understanding or the appearance of itUncertainty quantification and calibration as conditions on scientific useWhen a system’s outputs can enter a scientific argument as evidence rather than as a lead to follow up
06
Scientific Practice Under AI
Empirical study of what scientists are actually doing, as distinct from what systems are capable ofObservational, ethnographic, and longitudinal studies of AI use in research settingsWhen scientists defer to AI systems, when they override them, and what governs the choiceTrust calibration, automation bias, and deskilling in scientific workflowsHow AI integration is changing collaboration, group composition, and credit within research teamsDisciplinary variation in adoption, resistance, and the reasons given for each
07
Institutions, Infrastructure, and Sociotechnical Guardrails
Disclosure norms, red lines, and institutional policy on AI use in researchResearch infrastructure and capacity-building required to integrate these systems responsiblyFunding structures, incentives, and the political economy of AI adoption in scienceGovernance mechanisms and how existing responses have fared in practiceAccess, concentration, and who is positioned to do AI-integrated science at all