Probabilistic Worlds:   AI, Liberalism, and Experimental Philosophy

Probabilistic Worlds:  
AI, Liberalism, and Experimental Philosophy 

Denisa Reshef Kera| Science, Technology & Society studies 
278136-01 
  

Course Abstract (expand) 

This seminar investigates whether generative AI can function as a method of experimental philosophy by reconstructing early modern liberalism as a probabilistic conceptual system. Emerging in a context defined by epistemic uncertainty, expanding markets, and distributed authority, liberalism, especially in the work of John Locke, rests on a theory of knowledge grounded not in certainty but in probability. Liberal political institutions institutionalize uncertainty through consent, representation, markets, and constitutional constraint; moral reasoning shifts toward judgment under risk; and governance increasingly operates through statistical calculation and population management. 

Taking these structural parallels seriously, the course asks whether liberalism is compatible with probabilistic modeling, whether a historically bounded corpus can generate a coherent ideological world, and whether fine-tuning internalizes normative commitments. Students will construct a temporally bounded corpus (c. 1650–1750) and conduct comparative experiments using prompting, Retrieval-Augmented Generation (RAG), and fine-tuning to test whether liberalism can be instantiated as a probabilistic epistemic system. AI models are treated not merely as objects of critique, but as epistemic instruments for reconstructing conceptual worlds and advancing methodological innovation in philosophy and STS. 

Learning objectives 

Knowledge 

By the end of the semester, students will be able to explain the relationship between probabilistic epistemology and early modern liberal political theory, particularly in the work of John Locke; describe how liberal political structures (consent, representation, markets, and constitutional constraint) institutionalize uncertainty and distribute authority; and analyze how moral reasoning in liberal traditions shifts from teleological certainty to judgment under conditions of risk and expectation. They will also be able to explain the historical emergence of statistical governance, political arithmetic, and population-based administration in the eighteenth century, and to connect these developments to contemporary forms of technocratic and algorithmic authority. On the computational side, students will distinguish between prompt-based simulation, Retrieval-Augmented Generation (RAG), fine-tuning, and training-from-scratch as distinct epistemic research strategies; understand how large language models encode meaning probabilistically and how training data shapes conceptual distributions; and articulate major debates in STS concerning epistemic infrastructures, data bias, and the material conditions of knowledge production. 

Skills  

Students will be able to construct and curate a historically bounded corpus using explicit temporal and conceptual criteria, and design replication-style experiments comparing prompt-based simulation, Retrieval-Augmented Generation (RAG), and fine-tuned models. They will learn to operationalize philosophical concepts, such as property, consent, and liberty, into measurable experimental variables, and to develop methods for detecting anachronism, conceptual drift, and ideological coherence within model outputs. Students will critically evaluate whether generated results constitute simulation, historical reconstruction, or statistical artifact, while maintaining a structured research lab notebook documenting experimental design, dataset decisions, and interpretive challenges. The seminar culminates in a sustained research paper that advances a clear methodological claim about experimental epistemology and the use of AI as a philosophical instrument. 

Values

Students will cultivate methodological reflexivity regarding the epistemic and normative implications of training data and model design, recognizing that technical systems are never neutral infrastructures. They will develop intellectual humility in the face of uncertainty, reflecting the probabilistic foundations of liberal thought and the limits of epistemic certainty in both philosophy and computation. The course encourages critical awareness of how political ideologies can be encoded, reproduced, or transformed through statistical modeling, and fosters ethical responsibility in the construction and evaluation of AI systems as instruments of knowledge production. Students will be expected to approach both textual interpretation and experimental design with rigor, openness to interdisciplinary inquiry, and sustained attention to the normative stakes of reconstructing historical conceptual worlds. 

 Lessons plan (Including active learning):

* There may be changes in the syllabus depending on learning progress and effectiveness 

Final grade  

Course requirements 

Assignments 
Completion and timely submission of all assigned coursework is required. Coursework in this seminar includes replication-style experimental reports, corpus construction exercises, analytical reflections connecting primary texts to probabilistic modeling, one seminar leadership presentation, and a final research paper. 

Because the course is cumulative and methodologically structured, assignments build on prior experimental work. Students are expected to engage closely with primary texts, document their experimental procedures clearly, and provide evidence for all methodological claims. All submissions must be completed by the stated deadlines. Failure to submit required components, particularly the final research paper, may result in a failing grade. 

Attendance 
Attendance is not formally mandatory, and sessions will be recorded. However, this seminar functions as an experimental laboratory as well as a discussion-based course. Significant portions of corpus construction, model testing, and adversarial evaluation occur during in-class workshops. Students are therefore strongly encouraged to attend regularly. Those who are absent remain fully responsible for all course materials, announcements, collaborative dataset decisions, and assignment instructions provided during class sessions. Absence does not exempt students from experimental components or submission deadlines. 

Bibliography:

Primary texts 

Hobbes, Thomas. Leviathan, selected chapters on sovereignty and knowledge. (Recommended: Richard Tuck, ed., Cambridge Univ. Press.) 

Hume, David. An Enquiry Concerning Human Understanding, sections on probability and causation. (Standard edition: T. L. Beauchamp, ed., Oxford Univ. Press.) 

Locke, John. An Essay Concerning Human Understanding, Book IV: “Of Knowledge and Probability.” (Any scholarly edition; e.g., Peter H. Nidditch, ed., Oxford: Clarendon Press.) 

Locke, John. Two Treatises of Government, esp. chapters on property, consent, and legislative power. (Standard edition: Peter Laslett, ed., Cambridge Univ. Press.) 

Mill, John Stuart. Utilitarianism. (Any critical edition, e.g., George Sher, ed., Hackett Publishing.) 

Petty, William. A Treatise of Taxes and Contributions (1662), selected chapters. 

Woolf, Virginia. Orlando: A Biography. (Any standard edition.) 

Probability, Risk, and Statistical Thought 

Hacking, Ian. The Emergence of Probability: A Philosophical Study of Early Ideas about Probability, Induction and Statistical Inference. Cambridge: Cambridge University Press, 1975.  

Hacking, Ian. The Taming of Chance. Cambridge: Cambridge University Press, 1990. 

Gigerenzer, Gerd, Zeno Swijtink, Theodore Porter, Lorraine Daston, John Beatty, and Lorenz Krüger. The Empire of Chance: How Probability Changed Science and Everyday Life. Cambridge: Cambridge University Press, 1989. 

Porter, Theodore M. Trust in Numbers: The Pursuit of Objectivity in Science and Public Life. Princeton, NJ: Princeton University Press, 1995.  

Daston, Lorraine. Classical Probability in the Enlightenment. Princeton, NJ: Princeton University Press, 1988. 

Stigler, Stephen M. The History of Statistics: The Measurement of Uncertainty before 1900. Cambridge, MA: Belknap Press of Harvard University Press, 1986. 

Experimental & Computational Humanities 

Underwood, Ted. Distant Horizons: Digital Evidence and Literary Change. Chicago: University of Chicago Press, 2019.  

Guldi, Jo & Armitage, David. The History Manifesto. Cambridge: Cambridge University Press, 2014.  

Moretti, Franco. Distant Reading. London/New York: Verso, 2013 

AI Foundations 

Murphy, Kevin P. Probabilistic Machine Learning: An Introduction. Cambridge, MA: MIT Press, 2022. 

Jurafsky, Daniel & Martin, James H. Speech and Language Processing: An Introduction to Natural Language Processing, Computational Linguistics, and Speech Recognition. 3rd ed. Draft online manuscript (Stanford University, ongoing). 

Tunstall, Lewis, von Werra, Leandro & Wolf, Thomas. Natural Language Processing with Transformers: Building Language Applications with Hugging Face. Sebastopol, CA: O’Reilly Media, 2022. 

Alammar, Jay & Grootendorst, Maarten. Hands-On Large Language Models: Language Understanding and Generation. Sebastopol, CA: O’Reilly Media, 2024.