Colloquia, Fall 19-Spring 20

Equivalence from a Metaphysical Point of View

Abstract: Equivalent' theories represent the very same state of the world; any differences are merely conventional or notational. According to one view, equivalent theories are those that say the same thing about fundamental reality, understood in a particularly fine-grained way. According to a second view, which I call "quotienting" (short for "quotienting-out conventional content by hand"), theories may be equivalent even when we cannot state, in an intrinsic or "artifact-free" way, the content that the theories have in common. In a sense these are the extreme positions on the metaphysics of equivalence. The first view (which I accept) leads to the uncomfortable conclusion that we must recognize questions such as whether negation and conjunction, as opposed to negation and disjunction, form the metaphysically correct basis for propositional logic as being genuine questions. The second is dizzying, but sheds light on various otherwise perplexing viewpoints in metaphysics, philosophy of physics, and philosophy of mathematics.

Data as Collectively Generated Patterns: Making Sense of Data Ownership

Why Should We Care About Explainable Systems? Perspectives from Moral and Political Philosophy

"A Non-Ideal Theory of Sexual Consent"

Most of the literature on sexual consent presupposes that consent requires autonomy, and proceeds as though autonomy is a roughly digital property: you either have it or you don’t, and if you don’t have it then you can’t legitimately consent to sexual activity. But in fact, autonomy is a continuum, and our autonomy is almost always partial. Our autonomy can be compromised by limitations in our capacities, or by the power relationships within which we are embedded. If we insist that real consent requires full autonomy, then virtually no sex will turn out to be consensual. Accepting this consequent would either make the notion of consent useless, or it would turn most sex into rape, and neither is a helpful outcome. I argue that under conditions of compromised autonomy, consent must be socially and interpersonally scaffolded. To understand consent as an ethically crucial but non-ideal concept, we need to think about how it is related to other requirements for ethical sex, such as the ability to exit a situation, trust, safety, broader social support, epistemic standing in the community, and more.

"Contextual Integrity Up and Down the Data Food Chain"

Abstract: According to the theory of contextual integrity (CI), privacy is appropriate flow of information, not control or secrecy. Appropriateness is based on legitimate norms, or rules, which prescribe data flows with reference to five parameters — sender, recipient, subject, information type, and transmission principle. The values of these parameters range over contextually meaningful ontologies — of actors (acting in particular capacities) and information types. A recent challenge to CI is posed by an instrumented world of networked, sensor-enabled devices and data ravenous machine learning systems, which generates data whose semantics, at best, is unclear. Nissenbaum’s talk assesses the gravity of this challenge to CI. She introduces the construct of a data food chain to express the challenge and reveal its weak spots.

"Phenomenal Powers: A New Response to Hume."

abstract:

"Hume argued that there are no causal powers, at least not as far as we can know or positively conceive of, because all causes are conceivable without their effects. But a seeming exception to this claim can be found in the realm of phenomenal properties: it is difficult to conceive of the feeling of pain making a subject who experiences it pursue it (or do anything else than avoid it), or the feeling of pleasure making a subject avoid it (or do anything else than pursue it)—at least in the absence of interfering motives. These connections are standardly explained away as merely psychological, as analytic or constitutive (as per analytic functionalism), or as merely normative. I will argue that they should rather be taken at face value: as indicating that phenomenal pain and pleasure truly necessitate their effects in a properly causal way, and thereby constitute real, irreducible causal powers. I will then suggest that this view supports a kind of panpsychism, and that this is no reason to reject it."

Title: The Parent Trap: How Doxastic Wronging Starts at Home

Abstract: When Mulan reflects on how she'll never be the perfect bride or the perfect daughter, we recognize her pain. Our parents exert a kind of doxastic influence over our lives and it can hurt when we fail to live up to their expectations. In previous work I've focused on racist beliefs to illuminate the ways beliefs can wrong, however, the paradigmatic example of doxastic wronging likely lies somewhere closer to home. To both the racist and to our parents we may find ourselves saying, "You don't know me." With respect to the racist this response makes sense, but when it comes to our parents there are very few people who know us better. If there were anyone justified in forming expectations about us surely it'd be our parents. Although at some level we can all relate to Mulan, it is puzzling how the expectations our parents place upon us wrong us and it is unclear whether we are justified in these feelings of hurt and resentment. This paper is an attempt to answer these questions by providing a taxonomy of the ways our parents can wrong us in order to better understand how our relationships shape our attitudinal and doxastic obligations more generally.

Ethics and (Artificially) Intelligent Health Systems

Abstract:

Whether health care systems function effectively, efficiently, and equitably depends on their ability to gather and learn from reliable medical evidence across the lifecycle of intervention development and service delivery.  Orthodox research ethics has been reluctant to embrace the claim that there is a social imperative to promote learning in health systems because of the widespread perception that learning activities are often inconsistent with equal respect for all patients.  These worries are exacerbated by features of one of the most promising approaches to continuous learning in medicine, adaptive platform trials.  In this talk I lay out the source of this perceived dilemma and consider whether AI systems offer a way to learn while avoiding randomized clinical trials.  After detailing some sources of residual uncertainty surrounding AI systems in medicine, I argue that AI systems don’t avoid this dilemma.  I then argue that the underlying dilemma rests on a failure to understand the social nature of medical uncertainty and a tendency to represent a social epistemic and ethical problem as a decision problem for a single agent.  I argue that a properly social conception of medical uncertainty and the goals of research can reconcile a commitment to equal regard for all patients with a strong social imperative to learn.  

title to be announced

title to be announced

title to be announced