Research Program
My research examines how human purposes, evidence, and institutional arrangements shape the authority we give to technological systems. Working across epistemology, ethics, and the philosophy of technology, I ask what warrants the claims we make about artificial systems and what institutions must provide when their outputs affect people’s lives. I distinguish a system’s computational contribution from the desire, authority, or answerability attributed to it. I also examine how people can justify reliance on that contribution while preserving the evidence and practical capacities required for challenge and correction. Across …
Research Program
My research examines how human purposes, evidence, and institutional arrangements shape the authority we give to technological systems. Working across epistemology, ethics, and the philosophy of technology, I ask what warrants the claims we make about artificial systems and what institutions must provide when their outputs affect people’s lives. I distinguish a system’s computational contribution from the desire, authority, or answerability attributed to it. I also examine how people can justify reliance on that contribution while preserving the evidence and practical capacities required for challenge and correction. Across this work, I argue that consequential judgments made through technology must remain open to informed examination and effective intervention by the people and institutions responsible for their use.
Attributed Agency and Institutional Authority
In “The Institutional Ideomotor Effect: Generative AI, Disclosure, and Attributed Agency,” submitted to Minds and Machines, I examine how a recipient can mistake a conversational interface for the party with authority to decide an institutional matter. A model may generate an answer without possessing the authority that the recipient attributes to it. I argue that an AI label does not itself explain the arrangement through which the answer was produced and authorized. A usable disclosure of the functional division identifies the rule being restated, the step left unchecked, the interface’s authority or lack of authority, and the office with its review procedure. This distinguishes explaining what generated an answer from identifying whether an institution has authorized it to count as a determination. The proposed effects of fuller disclosure on recipients’ beliefs remain empirical questions; my philosophical argument identifies what the disclosure supplies and what a recipient has grounds to infer from it.
Human Desire and Artificial Attribution
In “Human Desire, Artificial Attribution,” submitted to Global Philosophy, I examine claims that present language agents have genuine desires and wellbeing. Through a worked invitation case, I trace how human motivational conditions remain operative through memory, retrieval, planning, dialogue, and delivery. The model makes a causal contribution, including generating particulars that no person anticipated, while the completed act can remain part of an integrated human undertaking. I defend an actual capacity for felt attraction or aversion as a necessary condition of the desire and welfare attributions at issue. Representing a preference, acting on it, and satisfying it do not by themselves identify a further subject for whom the outcome can be better or worse. My argument permits an artificial realization of the relevant capacity and asks what evidence would warrant attributing it to the particular system. Even if artificial desire were warranted, the humans who select purposes, impose constraints, and authorize uses would remain answerable for their own conduct.
Corrective Capacity and Institutional Answerability
In “Corrective Capacity and Institutional Support: A Reply to Floridi,” published in Philosophy & Technology in 2026, I develop the operational conditions of that answerability. A hypothetical classroom assessment shows how an automated grader can read a response accurately while the authorized assessment requires additional evidence of the student’s understanding. The teacher remains responsible, while the institutional arrangement supplies access to the relevant record, legitimate permission to intervene, and the means of revising the result. I identify operational answerability with the connected capacity for a specified correction at a specified time. Correcting an individual assessment can require a different arrangement from repairing the platform that produced it. This account distinguishes responsibility from the capacity to act, and distinguishes that capacity from its exercise or success.
Pragmatism, Verification, and AI
Chapter 10 of my book Honest Deception, “AI at the Precipice: The Transmission Model Automated,” Under Contract with Bloomsbury Academic, places these questions within my broader examination of journalism, education, and institutional epistemic practices. I examine how situated selections and evaluative frameworks can be presented as neutral authority, and how generated outputs can acquire consequential force before the people affected can reach the relevant evidence, criteria, or decision record. I apply three principles developed across the book: Principled Transparency, Narrative Pluralism, and the Primacy of Verification. They require disclosure of operative frameworks and limits, preservation of justified alternatives where evidence is underdetermined, and checking procedures that can challenge the original verdict under independently governed standards. My account of Haptic Pragmatism describes structured, fallible engagement through intervention, testing, and correction. Felt concern can prompt inquiry, while evidence and checking determine whether a claim warrants reliance. Familiarity with a system’s behavior can help guide an investigation, but the claims produced through that interaction still require verification appropriate to their subject and use.
Warranted Judgment and Effective Correction
Across these works, I develop an account of warranted, revisable judgment and of institutions capable of acting on a justified challenge. Completing a procedure establishes a result within its supported scope; it does not, through completion alone, certify that the procedure captures every relevant condition. Identifying a responsible person likewise does not establish that correction is available. The practical requirements must be connected: accessible evidence, justified criteria, preserved records, legitimate authority, and usable means of intervention. My research therefore asks whether a technological result deserves the status attributed to it and whether the arrangement relying on it can investigate, suspend, revise, or remedy its consequential use. These questions apply to the standards themselves as well as to the outputs assessed under them. A school, publisher, employer, or public agency must be able to examine whether its criteria remain adequate, receive relevant evidence from affected people, and implement a warranted correction while that intervention can still help.