The no-free-lunch theorems of supervised learning

The no-free-lunch theorems promote a skeptical conclusion that all possible machine learning algorithms equally lack justification. But how could this leave room for a learning theory, that shows that some algorithms are better than others? Drawing parallels to the philosophy of induction, we point out that the no-free-lunch results presuppose a conception of learning algorithms as purely data-driven. On this conception, every algorithm must have an inherent inductive bias, that wants justification. We argue that many standard learning algorithms should rather be understood as model-dependent: in each application they also require for input a model, representing a bias. Generic algorithms themselves, they can be given a model-relative justification.

The Obligation Dilemma

This book argues for the prima facie plausibility of the surprising and paradoxical conclusion that there are no moral obligations regardless of whether determinism is true. In the form of a dilemma, the primary argument for this skeptical conclusion presupposes that obligation requires freedom. A minimal number of credible principles entail that this is the freedom both to do, and to refrain from doing, what is obligatory. On the deterministic horn of the dilemma, since determinism eliminates freedom to do otherwise, it imperils moral obligation. On the indeterministic horn, pertinent actions are too luck-infected to qualify as obligations. Hence, there are no moral obligations. The book’s principal goal is to develop the obligation dilemma as powerfully and clearly as possible to inspire sustained philosophical work to solve it (assuming that it can be solved). In many respects, the obligation dilemma mirrors the venerable responsibility dilemma: regardless of whether determinism is true, no one is morally responsible for anything. The book shows that various prevalent moves in favor of, or in response to, the responsibility dilemma are, when suitably amended, not promising as supportive of, or retorts to, the obligation dilemma. Exposing the obligation dilemma’s implications for responsibility, and its ramifications for forgiveness (something central to salutary interpersonal relationships), underscores its urgency.

GROUNDING PROCEDURAL RIGHTS

It is commonly held that we wrong someone if we punish them without first determining whether they are guilty through the process of a sufficiently fair and reliable procedure. This wrong is best explained by pre-institutional moral procedural rights. Recently, Christopher Heath Wellman has argued for the skeptical conclusion that there are no such rights, challenging a widely held orthodoxy. I propose two novel grounds for pre-institutional moral procedural rights and so answer Wellman's skepticism. First, we have rights not to be subject to punitive systems that do not include specific sorts of reliable procedures because otherwise we are subject to unreasonable risks of undeserved punishment. Second, not only do we have rights that others not harm us or unreasonably risk harming us, we have rights that they control for avoiding wrongfully harming us across relevant close possible worlds.

Skepticism and the Point of Knowledge

This chapter explores the relationship between philosophical skepticism and the concerns of daily life. The aim is to show that function-first epistemology can augment an argument against skepticism. The force of the skeptic’s argument, as well as our desire to reject the skeptical conclusion, is explained in the following way: our need to share information pushes us to accept stricter epistemic standards that might logically end at skepticism, but practical factors encourage us to formulate standards that stop short of skepticism. This tension creates an area of indeterminacy in which controversies about skepticism take place. This chapter explains the persuasive power of skepticism while also explaining why skeptical worries do not (and should not) threaten our everyday knowledge.

A Dilemma for Saulish Skepticism: Either Self-Defeating or Not Even Skepticism

Jennifer Saul argues that the evidence from the literature on implicit biases entails a form of skepticism. In this paper, I argue that Saul faces a dilemma: her argument is either self-defeating, or it does not yield a skeptical conclusion. For Saul, both results are unacceptable; thus, her argument fails.

How can Linguistic Meaning be Grounded – in a Deconstructionist Semiotics?

Deconstruction is one of the more (in)famous theories in recent times. In this paper, I argue that the theory of deconstruction, proposed by Derrida in particular, should be read as a systematic and rigorous examination of key philosophical and semiotic notions, such as sign and meaning. The relevance of taking deconstructive critique seriously is explored with the point of departure in Derrida’s argument that linguistic signs are characterized by repeatability. This view is situated against attempts to ground language in context, speaker intentions and truth conditions, showing how deconstruction challenges these attempts for not taking the repeatability of signs sufficiently into account. Instead, deconstructive semiotics radicalizes the idea that linguistic signs always involve differential structures that postpone the determination of meaning. While this might be read as a skeptical conclusion, I propose that it should be positively interpreted as a relevant contribution for the theoretical understanding of language, signs and meaning.

Incompatibilism and Prudential Obligation

Take determinism to be the thesis that for any instant, there is exactly one physically possible future (van Inwagen 1983, 3), and understand incompatibilism regarding responsibility to be the view that determinism is incompatible with moral responsibility. Of the many different arguments that have been advanced for this view, the crux of a relatively traditional one is this: If determinism is true, then we lack alternatives. If we lack alternatives, then we can't be morally responsible for any of our behavior. Therefore, if determinism is true, then we can't be morally responsible for any of our behavior. The second premise is a version of the principle of alternate possibilities (PAP): persons are morally responsible for what they have done only if they could have done otherwise. This principle, in conjunction with the assumptions that responsibility requires control, and that this control consists in the freedom to do otherwise, provides the vital bridge from the initial premise to the skeptical conclusion. Some incompatibilists, joining ranks with various compatibilists, have sought to reject this principle by invoking so-called ‘Frankfurt examples.’

An Internalist Rejoinder to Skepticism

In what follows, I will offer a rejoinder to one popular and influential version of external world skepticism which avoids (i) begging questions against the skeptical conclusion, (ii) arguing from prudential considerations to a dismissal of skeptical worries, and (iii) arguing for epistemological conclusions from semantic premises.

Thomas Hobbes and the Contractarian Theory of Law

The revival of interest in contractarian theories of morals and politics may encourage us to enquire into the prospects for a contractarian theory of law. Such a theory would be normative, aiming, not at the best explanation or justification of our actual legal practices and institutions, but rather at a rational reconstruction of those practices and institutions from the perspective of agreement among maximizing individuals. It would offer a secular and instrumental rationale for law that requires no dubious assumptions about the objectivity of values or the existence of moral order in the universe. Of course, such a reconstruction might fail, leading to a skeptical conclusion. Rational individuals, in a position to decide on their terms of interaction, might reject any structure that we should recognize as a legal framework. But this must seem unlikely. We may compare law with morality, recognizing that both have arisen in a framework of teleological and theological understandings that are themselves no longer plausible to us, without supposing that either must therefore be wanting from the perspective of rational agreement.