AI Research
AI deception: a survey of examples, risks, and potential solutions
Coauthored with Peter Park, Aidan O’Gara, Michael Chen, and Dan Hendrycks
Synopsis: A range of AI systems have learned how to deceive humans.
Language agents reduce the risk of existential catastrophe
Coauthored with Cameron Domenico Kirk-Giannini
AI & Society forthcoming
Synopsis: AI agents with folk psychology built on top of large language models are the safest path to AGI.
Generative agents are unethical
Coauthored with Cameron Domenico Kirk-Giannini
op-ed in ABC
Synopsis: Building generative AI agents is potentially unethical, because they may have wellbeing.
AI wellbeing
Coauthored with Cameron Domenico Kirk-Giannini
under review
Synopsis: Some AIs today have wellbeing. This raises serious ethical concerns.
Aggregating utilities for corrigible AI
under review
Synopsis: corrigible AIs allow us to change their goals; one path to corrigible AI is by designing them to maximize ‘aggregated’ utility instead of ordinary utility.
Shutdown-seeking AI
Coauthored with Pamela Robinson
Philosophical Studies forthcoming
Synopsis: one promising strategy for building safe AI is to give AIs the goal of being shut down.
The Polarity Problem
Coauthored with Cameron Domenico Kirk-Giannini
under review
Synopsis: a model of whether there will be one vs many extremely powerful AIs in the future.
Other Research
Iterated knowledge
Oxford University Press forthcoming
Synopsis: omega knowledge, or infinitely iterated knowledge, plays an important role in philosophy. It is neither scarce nor identical to knowledge.
Safety, closure, and extended methods
Coauthored with John Hawthorne
Journal of Philosophy forthcoming
Synopsis: if knowledge requires belief that is safe from error, then knowledge is not preserved by deduction.
Attitude verbs’ local context
Coauthored with Kyle Blumberg
Linguistics and Philosophy forthcoming
Synopsis: 'Ann wants Bill to stop smoking’ presupposes that Ann thinks Bill used to smoke. Recent theories of presupposition do not predict this.
A question-sensitive theory of intention
Coauthored with Bob Beddor
Philosophical Quarterly forthcoming
Synopsis: intention is question-sensitive.
Contextology
Coauthored with Cameron Domenico Kirk-Giannini
Philosophical Studies forthcoming
Synopsis: all of the research that has relied on Stalnaker’s theory of context is conceptually incoherent.
Sly Pete in dynamic semantics
Journal of Philosophical Logic forthcoming
Synopsis: dynamic semantics offers an elegant model of the puzzling conversational dynamics of Sly Pete cases.
Getting accurate about knowledge
Coauthored with Sam Carter
Mind forthcoming
Synopsis: a theory of knowledge in terms of facts about the accuracy of evidence.
Omega knowledge matters
Oxford Studies in Epistemology 2022
Honorable Mention for the Sanders Prize in Epistemology
Synopsis: why infinitely iterated knowledge matters, and how to get it.
Knowledge from multiple experiences
Coauthored with John Hawthorne
Philosophical Studies 2021
Synopsis: knowledge is reducible to evidential probability. This reduction explains striking facts about perceptual knowledge.
Fragile knowledge
Mind 2021
Synopsis: if you know p, then for all you know: you know that you know p.
Probability for epistemic modalities
Coauthored with Paolo Santorio
Philosophers’ Imprint 2021
Synopsis: we capture the interaction between probability and modality, including Stalnaker's Thesis, in a fairly standard dynamic/informational semantics.
Counterfactual contamination
Coauthored with John Hawthorne
Australasian Journal of Philosophy 2021
Synopsis: knowledge requires belief to be true in normal cases, not in similar cases.
Mighty knowledge
Coauthored with Bob Beddor
Journal of Philosophy 2021
Synopsis: a theory of what it takes to safely believe and know epistemic modal claims.
The normality of error
Coauthored with Sam Carter
Philosophical Studies 2020
Synopsis: justification is possible knowledge, and so justification fails to agglomerate.
Losing confidence in luminosity
Coauthored with Dan Waxman
Noûs 2020
Synopsis: a theory of confidence as distinct from credence, and a new anti-luminosity argument based upon it.
Epistemic modal credence
Philosophers’ Imprint 2020
Synopsis: how to assign credence to epistemic modal claims.
Free choice and homogeneity
Semantics and Pragmatics 2019
Synopsis: Free Choice is a homogeneity effect.
The counterfactual direct argument
Linguistics and Philosophy 2019
Synopsis: a new principle about counterfactuals constrains their logic and meaning.
Free choice impossibility results
Journal of Philosophical Logic 2019
Synopsis: you can't validate Free Choice in a classical semantics.
A theory of conditional assertion
Journal of Philosophy 2019
Synopsis: a formal definition of conditional assertion.
Generalized update semantics
Mind 2019
Synopsis: dynamic semantics without tests.
Conditional heresies
Coauthored with Fabrizio Cariani
Philosophy and Phenomenological Research 2018
Synopsis: CEM and SDA are incompatible, or are they?
A stronger doctrine of double effect
Coauthored with Ben Bronner
Australasian Journal of Philosophy 2017
Synopsis: the traditional doctrine of double effect is too weak.
Triviality results for probabilistic modals
Philosophy and Phenomenological Research 2017
Synopsis: a battery of impossibility results about epistemic modal belief.
Believing epistemic contradictions
Coauthored with Bob Beddor
The Review of Symbolic Logic 2017
Synopsis: a puzzle about epistemic modal belief, and a Lockean solution.
A preface paradox for intention
Philosophers' Imprint 2016
Synopsis: intentions come in degrees.