Philosophy of Computing in January
Welcome back to the new year—and to a new semester for many. It’s been a busy start to 2026, and with so much attention on international politics, advances in AI and normative research have often proceeded in relative silence. Fear not: the pace remains fast. This month alone we’ve seen major work on memory, renewed focus on multi-agent systems (sometimes swarms!), and new U.S. blueprints for “AI dominance.”
But before we jump in, for those of you who may be new: I’m Cameron Pattison and this newsletter from Seth Lazar's MINT Lab brings together all the most recent news, calls for papers, conferences, jobs, and papers. We aim to serve philosophers in the philosophy of AI and engineers interested in the normative dimensions of machine learning.
Below, you’ll find three broad sections: opportunities, papers, and links. The first lists the best special issue CFPs, conferences, workshops, and jobs. The second lists papers published this month in top philosophy journals and a couple notable papers in specialist venues. The third is a goldmine of links to ArXiv preprints, news flashes, comments from cutting edge researchers, and other news, all grouped into a few themes we saw in the month’s output. It’s a firehose, and we don’t expect anyone (except those “blessed with angelic patience”) to read everything. The goal is to put it all in one place—so you can skim, follow what catches your eye, and dip into the conversations that matter most to you.
We’re glad to have you here and hope you’ll join the conversation.
Highlights From Each Section:
To help you get oriented, here are a few highlights from each section:
In CFPs and Conferences: Synthese and Philosophy & Technology are hosting multiple special collections on agency, epistemology, ethics, trust, and AI-mediated inquiry, while COLM is open for both paper and workshop submissions. Calls are open for PSA 2026, IACAP 2026, PLSC 2026, and workshops at USC, the University of Rochester, and the University of Hong Kong—spanning philosophy of science, AI ethics, pedagogy, and human values.
In Papers: New work at the intersection of AI, representation, agency, and ethics. This issue highlights recent papers on LLM representation, robustness and trust, AI deception and moral standing, accountability without agency, and synthetic images—alongside fast-moving developments in agents, alignment, interpretability, and AI governance.
In Links: Rapid advances in agents, reasoning, and evaluation are reshaping the field. From long-horizon coding agents and multi-agent architectures to new benchmarks for robustness, alignment, and scientific reasoning, recent work points to qualitative shifts in how AI systems learn, act, and are assessed—raising fresh philosophical questions about control, understanding, memory, and responsibility.
(1) Opportunities:
Special Collection CFPs:
Synthese — Special Collections on AI, Agency, and Understanding
Synthese has now four upcoming topical collections. Three we mentioned last month: Artificial Joint Intentionality examines whether and how interactive AI systems (LLMs, social robots) can participate in skilled social action and joint agency, and what is missing—conceptually or normatively—from human–machine interaction; The Philosophy of Generative AI: Perspectives from East and West fosters cross-tradition dialogue on reasoning, epistemic reliance, explainability, and the relation between generative models, logic, and probabilistic reasoning; Epistemic Agency in the Age of AI focuses on how opaque, AI-mediated systems reshape belief formation, responsibility, autonomy, and epistemic standing, addressing both epistemic benefits and risks across social epistemology, philosophy of science, and political philosophy. The new collection, Models, Representation, and Computation addresses how novel computational practices—including AI integration, datification, and machine learning—fit existing accounts of scientific representation and inference, with particular attention to opacity, formal verification, and whether these techniques require modified epistemic frameworks and scientific norms.
Philosophy & Technology — Special Collections on Trust, Ethics, and AI-Mediated Inquiry
Philosophy & Technology is hosting two topical collections that center AI’s integration into normative practice and ethical reasoning itself. The Ethics of Medical Artificial Intelligence examines trust, adoption, and governance in clinical contexts, where AI intersects with life-or-death decisions, bioethical frameworks, regulatory regimes, and concerns about opacity, bias, and epistemic authority. Using AI Tools to Support Ethical Inquiry turns to the meta-level, asking which forms of ethical research AI can appropriately support, whether AI is better suited to discovery than justification, and whether AI systems could ever function as ethics experts to whom philosophers might defer.
Conferences and Workshops:
Workshop on AI Agents and Companions
Dates: April 23–24, 2026
Location: University of Hong Kong, Hong Kong SAR
Link: https://philevents.org/event/show/140454
The AI & Humanity Lab at HKU invites presentations on the philosophy of AI agents and companions. Topics include conceptual and metaphysical problems related to AI agents, the use of AI companions for friendship and relationships, governance of AI agents, risks specific to AI agents, ethical problems related to agent benchmarking, and moral and legal responsibility. Selected presenters will have travel and accommodation covered.
6th Upstate Workshop on AI and Human Values
Conference date(s): April 24, 2026
Location: University of Rochester
Link: https://philevents.org/event/show/143101
The Department of Philosophy at the University of Rochester invites abstracts for a manuscript workshop on normative and philosophical aspects of AI, including (but not limited to) understanding and consciousness in algorithmic systems, the ethics of algorithmic systems, value disagreement in algorithmic systems, explainability and justification by algorithmic systems, and philosophy of science as applied to data science and algorithmic systems. Faculty, postdoctoral fellows, and doctoral students in philosophy or cognate disciplines are invited to submit a ~500-word abstract prepared for blind review, along with name, email, and institutional affiliation; preference may be given to scholars at Central New York Humanities Corridor institutions.
AI and Data Ethics: Philosophy and Pedagogy
Conference date(s): March 6, 2026 - March 7, 2026
Location: University of Southern California
Link: https://philevents.org/event/show/142962
University of Southern California is hosting a workshop with two aims: (1) to bring together scholars working on the philosophy and ethics of AI, data, and related topics in computing and technology ethics, and (2) to share pedagogical ideas and strategies for teaching this material—especially to STEM and CS students—and for other public-facing presentations. Limited travel support is available, with preference for those without sufficient institutional travel funding. Confirmed speakers: Gabby Johnson, David Danks, C. Thi Nguyen. Organizers: David Danks, John Patrick Hawthorne, C. Thi Nguyen, Mark Schroeder.
Philosophy of Science Association ‘26
Submission deadline: March 15, 2026 (Contributed Papers); June 1, 2026 (Posters)
Conference date(s): November 19 - November 22, 2026
Link: https://mms.philsci.org/msg_detail.php?mid=200583647
PSA26 invites (i) symposium proposals (deadline January 15, 2026), typically featuring 4–5 presenters organized around a topic of interest in the philosophy of science, with decisions expected prior to the contributed papers deadline; (ii) contributed papers (deadline March 15, 2026), with accepted papers published in the conference proceedings edition of Philosophy of Science and a maximum manuscript length of 4,500 words; and (iii) posters (deadline June 1, 2026), with poster abstracts up to 500 words (including references) on any research topic in philosophy of science, teaching philosophy of science, outreach/public philosophy/science communication, and philosophy-of-science-focused grant projects.
Privacy Law Scholars Conference 2026 (PLSC 2026)
Submission deadline: January 20, 2026 (by 11:59 PM Pacific time)
Conference date(s): May 28, 2026 - May 29, 2026 (mentorship-focused pre-conference: May 27, 2026)
Link: https://privacyscholars.org/plsc-2026/
PLSC 2026 (the 19th annual Privacy Law Scholars Conference) will be held in person at the Indiana University Maurer School of Law in Bloomington, Indiana, and invites submissions for its paper workshop format focused on in-progress scholarship at the intersection of law, technology, and information privacy.
International Association for Computing and Philosophy (IACAP) 2026
Submission deadline: January 31, 2026
Conference date(s): July 15, 2026 - July 17, 2026
Link: https://iacapconf.org
The International Association for Computing and Philosophy (IACAP) invites abstracts for its 2026 conference, hosted by the Center for Cyber-Social Dynamics at the University of Kansas (Lawrence, KS). The conference promotes philosophical and interdisciplinary research on computing and its implications, including (among other areas) philosophy of AI, philosophy of information, ethics of AI/computation/robotics, epistemological issues in AI and machine learning, computer-mediated communication, human-computational interaction, and theoretical problems in computer science. Special tracks include Automation in Science, Epistemology of ML, and Pragmatics of LLMs, alongside workshop sessions.
Jobs:
Research Assistant Professor: Philosophy of AI / Ethics of Risk
Location: Department of Philosophy, Lingnan University | Tuen Mun, Hong Kong
Link: philjobs.org/job/show/29029
Deadline: Open until filled
Lingnan seeks a fixed-term Research Assistant Professor in philosophy of AI and/or ethics of risk. The role combines a 3-course/year load with a strong research brief tied to the Hong Kong Catastrophic Risk Centre (HKCRC)—publishing in leading journals, applying for competitive grants, and organizing seminars/reading groups. PhD in philosophy (or related) required, conferred within five years of start. Start: ideally Aug 2025 (no later than Jan 2026).
Post-doctoral Fellow: Philosophy of Artificial Intelligence
Location: School of Humanities, University of Hong Kong | Pokfulam, Hong Kong
Link: https://philjobs.org/job/show/29285
Deadline: Open until filled
Two-year appointment (possible one-year extension) affiliated with HKU’s AI & Humanity Lab. AOS/AOC open, with preference for philosophy of AI or technology. Applicants submit a ≤5-page project proposal, CV, and writing sample; active participation in the Lab is expected.
(2) Papers:
Robustness and Trustworthiness in AI: A No-Go Result from Formal Epistemology
Levin Hornischer | Synthese
Hornischer proves a no-go result: no matter how robustness and trustworthiness are understood, they cannot have four prima facie desirable properties without trivializing. Using a modal logic to reason about AI robustness, he proves that the four properties imply triviality via a novel interpretation of Fitch’s lemma. The paper builds bridges between AI and epistemology, showing that modern AI provides new questions and perspectives for epistemology while epistemology provides novel methods for AI.
What Do Large Language Models Represent?
Quentin Ruyant | Synthese
Ruyant applies philosophical conceptions of epistemic representation to LLMs. After discarding the idea that they represent the structure of natural languages, he argues that LLMs do not directly represent the world by linguistic means either, but rather a certain class of appropriate linguistic production constructed by their makers. They are mostly used to generate fictitious instances of this class—fictions that can be performative when appropriated by humans. The implication is that conversing with an LLM chatbot amounts to engaging in a fiction, where our own imagination plays a central role.
AI Deception and Moral Standing
Anton Skretta | Philosophical Studies
Skretta identifies a tension between the presumptive moral standing of future AI and currently popular safety measures aimed at mitigating risks from AI deception. Robust deceptive capabilities play a central role in worries about controlling advanced AI systems. But given what a capacity for robust deception requires, any being capable of it holds presumptive moral standing. Despite our own legitimate interests in safety, some measures may be ruled out by moral concern for AI themselves.
Impartiality Preferences in Sacrificial Moral Dilemmas Involving Autonomous Vehicles
Norbert Paulo, Lando Kirchmair & Yochanan E. Bigman | Analysis
Paulo, Kirchmair, and Bigman criticize the forced-choice methodology of studies like the Moral Machine Experiment. They argue that “equal” treatment can be understood in different ways, proposing instead the concept of impartiality complemented by action-guiding decision rules. Their vignette study suggests that impartiality is the main preference and that the most attractive decision rule is not random choice, as many papers suggest, but inaction.
Does Accountability Require Agency? Comment on Responsibility and Accountability in the Algorithmic Society
Tillmann Vierkant | Philosophy & Technology
Vierkant responds to recent work distinguishing responsibility and accountability for algorithmic agents. He argues that accountability is ambiguous in an important way: it could be purely instrumental with regard to morally desirable consequences, or it could necessarily contain an element of scaffolding for the agent who is held to account. The comment develops both options and discusses the consequences of choosing either.
The Différance Engine: Large Language Models and Poststructuralism
David J. Gunkel | AI & Society
Gunkel argues that LLMs actualize Derrida’s concept of différance—the production of meaning through difference and deferral rather than stable reference. Transformer architectures generate content by calculating statistical differences across massive textual corpora, computationally enacting the very mechanisms of spacing, temporalization, and trace that Derrida theorized. The essay examines philosophical consequences for logocentrism, authorship, and the metaphysics of presence, addresses three potential criticisms, and identifies systemic limitations. The paper thus shows how LLMs can be read through poststructuralist theory, and how poststructuralist theory can be clarified through the technical operations of contemporary AI.
Depiction and Synthetic Images
Joseph Masotti | Philosophical Studies
Masotti argues that AI-generated synthetic images provide strong counterexamples to the “intentional standard of correctness”—the widely-held view that a picture depicts something only if its maker held a relevant intention regarding that object. When a user prompts DALL-E for “an image of a philosopher,” the resulting image depicts candles, books, and globes in the background—objects the user never intended and may not have noticed. The paper systematically examines three possible routes: non-intentionalist theories of depiction, alternative intentionalist accounts (including AI intentions, designer intentions, and hypothetical intentionalism), and depictive eliminativism (the view that synthetic images don’t depict anything at all). Each route fails or carries serious costs. Non-intentionalist views face their own counterexamples; alternative intentionalist accounts cannot locate the right intentions; and eliminativism has radical implications for ethics—if synthetic images don’t depict, one cannot object to AI-generated pornography on grounds that it is “of” them. Masotti concludes that synthetic images may require a sui generis theory of depiction, making the emergence of this technology as philosophically significant as the invention of the camera.
(3) Links:
Companies Being Companies: Anthropic’s Project Vend Phase 2, while not quite a company, reported that the Claude-based, “Claudius” agent running a vending machine in Anthropic’s headquarters took a serious turn for the better. It was easily manipulated and deeply financially irresponsible in past, but with the progress made on Claude this past year, Claudius may soon be able to autonomously manage its finances. In bigger businesses, Manus was acquired by Meta. OpenAI also wrote on how advertising (which will roll out in the next couple weeks for some) will make their platform more accessible, and won’t output reliability. Engineers over at Browser Use figured out how to stably give all their passwords to a browser agent which then logs in on their behalf.
Robustness and interpretability: Lin et al. introduce stage-wise fact-checking evaluation across the full pipeline from evidence retrieval to adversarial robustness. Activation Oracles train models to answer natural-language questions about their own internal activations, claiming robust out-of-distribution generalization. Transluce’s Predictive Concept Decoders predict behavior from activations through a sparse, auditable bottleneck. Anthropic’s Bloom is an open-source framework for generating reproducible behavioral evaluations across automatically generated scenarios. Lauren Wagner’s essay on AI benchmarks and the politics of measurement examines how evaluation choices shape what gets optimized.
Alignment: Owain Evans and others published in Nature a follow on to their emergent misalignment paper expanding its findings. A string of similar follow ups by that team has showed that fine tuning on one thing (e.g. bird names from the 19th century) has significant and unanticipated effects on model behavior (in this case, it turns time back to the 19th century on questions of recent inventions and gender roles, among other things). UK AISI, DeepMind, and others wrote on design choices in defense against high-risk future AI deployments. The AISI also introduced ASYNC CONTROL uses adversarial red–blue games to test whether asynchronous monitoring can detect sabotage by coding agents.
Agents: The agentic frontier advanced on multiple fronts. OpenAI’s GPT-5.2-Codex targets long-horizon coding through context compaction and improved handling of large refactors. METR introduces an economically-grounded capability metric: Claude Opus 4.5’s “50%-time horizon” is roughly 4 hours 49 minutes, meaning tasks taking a median human that long succeed half the time. On architecture, Sophia proposes a “System 3” meta-cognitive layer for persistent, self-auditing agents, while BOAD uses bandit optimization to discover hierarchical multi-agent systems. In fact, there’s been a lot of talk about multi-agent architectures (at least in my world). LangChain’s Sydney Runkle wrote on when you should use multiple agents, and Andy Hall shows what experts who oversee swarms of agents might see in the near future. Anthropic’s engineering team published practical guidance on agent evaluation—why it’s hard and how to maintain automated evals at scale.
Capabilities: UK AISI’s first Frontier AI Trends Report synthesizes evaluations of over 30 frontier systems, finding capabilities improving rapidly across all tested domains with several surpassing PhD-level human baselines. Year-in-review pieces from Karpathy and Willison converge on similar themes: RLVR as a major new training paradigm, the rise of reasoning models with controllable “dials,” and programming becoming orchestration rather than code-writing. DeepSeek’s Manifold-Constrained Hyper-Connections addresses scaling Hyper-Connections without sacrificing training stability. Premakumar et al. find that auxiliary self-modeling objectives cause networks to self-regularize and become more parameter-efficient. PersonalAlign uses everything it knows about you from long-term user records to resolve ambiguities in your vague prompts.
Minds, Memories, and Representations: Memory has been a major theme this month. NVIDIA discussed some recent research on compressing context into model weights at test-time. Their test-time training (TTT-E2E), lets LLMs compress long context into their weights during inference, enabling models to effectively “learn” from context at test time. In research out of MIT, scientific foundation models appear to converge on similar representations across molecules, materials, and proteins—suggesting models may be discovering something like natural kinds. The Bayesian Attention Trilogy continues with closed-form gradient analysis showing how training sculpts attention into geometries supporting Bayesian updating. Li et al. demonstrate that a single well-chosen training sample can improve LLM reasoning, even across disciplines. MemRec separates memory management from ranking in LLM recommendation systems—a lightweight memory module extracts preference themes from the interaction graph using domain-specific rules, then passes a compact summary to the LLM for ranking, with batched background updates. Zhengmian Hu proposes an algorithmic-information framework for concept formation that treats concepts as information objects rather than linguistic labels, drawing on dialectical philosophy to formalize revision. Researchers collaborating between Rutgers, Hong Kong Baptist, and Snap inc. propose a different framework that decouples reasoning and memory to avoid cognitive overload. A methodological note: Zaman and Srivastava challenge claims that Chain-of-Thought is unfaithful, arguing that hint-verbalization tests conflate unfaithfulness with incompleteness. OpenAI’s FrontierScience benchmark measures deep scientific reasoning through Olympiad-style problems and research sub-tasks. Daria Ivanova, Neel Nanda and others show that there may be some stability behind all that stochasticity in LLM reasoning chains!
Emily Bender and Nanna Inie beat the drum against anthropomorphizing language for AI systems. Meanwhile, the inaugural Digital Minds Newsletter reviews the year’s developments on AI consciousness and moral status. On another note, a paper on LLM secret keeping shows that LLMs cannot reliably keep a hidden state (they tested whether their LLM could withhold a secret word when hosting Hangman) without an explicit private working memory, and a quiz given to LLM agents (called Task2Quiz) suggests that task success may be a poor indicator of LLM understanding.
Governance and Economics: Philosophers may be looking at the new AI policy at Ethics, which takes a very hard line against AI authorship. Dean Ball gave an early update on the state of AI law in 2026. Trammell and Patel’s “Capital in the 22nd Century” argues that AI/robotics rendering capital substitutable for labor activates Piketty’s mechanism: capital’s share tends toward 1, wealth concentration accelerates. A related essay on “balance of power in the age of scale” warns that traditional checks are weakening as economies of scale strengthen. On safety governance, Distributional AGI Safety proposes frameworks for when AGI emerges as a distributed patchwork of coordinating sub-AGI agents, centering on sandbox economies governed by market mechanisms and cryptographic protocols. Kolt et al. propose “legal alignment”—integrating legal rules, principles, and methods into AI alignment, positioning legal scholarship as resource rather than constraint. Anthropic publishes their 4th economic index report looking at how AI is used in the real world. IBM Research and others studied why LLM multi-agent systems fail (when they do). As for cyber-crime, very recent research shows a diverse landscape of attacks and security risks that substantial mirrors traditional malware campaigns. The US Department of War issued a memo detailing their approach to “Accelerating America’s Military AI Dominance.”
Taking a step back…
Having deja vu when you hear debates about how much control we’re offloading to AI coders and the like? Arvind Narayanan too. Seeing a move toward bespoke solutions with AI? Steve Newman sees it too. This talks a lot about Claude Code… for those who haven’t been living and breathing it Lila Shroff at The Atlantic explains. And Claire Vo explains Claude’s new Cowork (which many take as Claude Code with softer user interface).
Wondering if what you just read is logically sound? Wait in line with us for Ivan and Dominik Scherm’s very cool looking tool.
Content by Cameron Pattison and Seth Lazar with additional support from the MINT Lab team.
