Inspired by recent breakthroughs in predictive modeling, practitioners in both industry and government have turned to machine learning with hopes of operationalizing predictions to drive automated decisions. Unfortunately, many social desiderata concerning consequential decisions, such as justice or fairness, have no natural formulation within a purely predictive framework. In efforts to mitigate these problems, researchers have proposed a variety of metrics for quantifying deviations from various statistical parities that we might expect to observe in a fair world and offered a (...) variety of algorithms in attempts to satisfy subsets of these parities or to trade o the degree to which they are satised against utility. In this paper, we connect this approach to fair machine learning to the literature on ideal and non-ideal methodological approaches in political philosophy. The ideal approach requires positing the principles according to which a just world would operate. In the most straightforward application of ideal theory, one supports a proposed policy by arguing that it closes a discrepancy between the real and the perfectly just world. However, by failing to account for the mechanisms by which our non-ideal world arose, the responsibilities of various decision-makers, and the impacts of proposed policies, naive applications of ideal thinking can lead to misguided interventions. In this paper, we demonstrate a connection between the fair machine learning literature and the ideal approach in political philosophy, and argue that the increasingly apparent shortcomings of proposed fair machine learning algorithms reflect broader troubles faced by the ideal approach. We conclude with a critical discussion of the harms of misguided solutions, a reinterpretation of impossibility results, and directions for future research. (shrink)

Typically, fair machine learning research focuses on a single decision maker and assumes that the underlying population is stationary. However, many of the critical domains motivating this work are characterized by competitive marketplaces with many decision makers. Realistically, we might expect only a subset of them to adopt any non-compulsory fairness-conscious policy, a situation that political philosophers call partial compliance. This possibility raises important questions: how does partial compliance and the consequent strategic behavior of decision subjects affect the allocation outcomes? (...) If k% of employers were to voluntarily adopt a fairness-promoting intervention, should we expect k% progress (in aggregate) towards the benefits of universal adoption, or will the dynamics of partial compliance wash out the hoped-for benefits? How might adopting a global (versus local) perspective impact the conclusions of an auditor? In this paper, we propose a simple model of an employment market, leveraging simulation as a tool to explore the impact of both interaction effects and incentive effects on outcomes and auditing metrics. Our key findings are that at equilibrium: (1) partial compliance by k% of employers can result in far less than proportional (k%) progress towards the full compliance outcomes; (2) the gap is more severe when fair employers match global (vs local) statistics; (3) choices of local vs global statistics can paint dramatically different pictures of the performance vis-a-vis fairness desiderata of compliant versus non-compliant employers; (4) partial compliance based on local parity measures can induce extreme segregation. Finally, we discuss implications for auditors and insights concerning the design of regulatory frameworks. (shrink)

The relationship between Peircean abduction and the modern notion of Inference to the Best Explanation is a matter of dispute. Some philosophers, such as Harman :88–95, 1965) and Lipton, claim that abduction and IBE are virtually the same. Others, however, hold that they are quite different :503, 1998; Minnameier in Erkenntnis 60:75–105, 2004) and there is no link between them :419–442, 2009). In this paper, I argue that neither of these views is correct. I show that abduction and IBE (...) have important similarities as well as differences. Moreover, by bringing a historical perspective to the study of the relationship between abduction and IBE—a perspective that is lacking in the literature—I show that their differences can be well understood in terms of two historic developments in the history of philosophy of science: first, Reichenbach’s distinction between the context of discovery and the context of justification—and the consequent jettisoning of the context of discovery from philosophy of science—and second, underdetermination of theory by data. (shrink)

Some naturalists feel an affinity with some religions, or with a particular religion. They may have previously belonged to it, and/or been raised in it, and/or be close to people who belong to it, and/or simply feel attracted to its practices, texts and traditions. This raises the question of whether and to what extent a naturalist can lead the life of a religious believer. The sparse literature on this topic focuses on religious fictionalism. I also frame the debate in these (...) terms. I ask what religious fictionalism might amount to, reject some possible versions of it and endorse a different one. I then examine the existing proposals, by Robin Le Poidevin, Peter Lipton, Andrew Eshleman and Howard Wettstein, and show that even on my version of religious fictionalism, much of what has been described by these authors is still possible. (shrink)

The commentaries provided by Haig; Derksen and Morawski; and Trafimow vary considerably in how they address critical realism and its implications for replication. Haig’s preference for Kaidesoja’s “naturalised” version of critical realism and Lipton’s inference to the best explanation is deeply problematic. While Derksen and Morawski concede that they deal only indirectly with critical realism, their endorsement of “performativity” negates it. In Trafimow’s case, ontology’s regulative role is untenably diminished and ultimately supplanted by classic methodologism. I conclude that replication (...) should be replaced by exploratory stratified contextualism. (shrink)

According to the evolutionary sceptic, the fact that our cognitive faculties evolved radically undermines their reliability. A number of evolutionary epistemologists have sought to refute this kind of scepticism. This paper accepts the success of these attempts, yet argues that refuting the evolutionary sceptic is not enough to put any particular domain of beliefs – notably scientific beliefs, which include belief in Darwinian evolution – on a firm footing. The paper thus sets out to contribute to this positive justificatory project, (...) underdeveloped in the literature. In contrast to a ‘wholesale’ approach, attempting to secure justification for all of our beliefs on the grounds that our belief-forming mechanisms evolved to track truth, we propose a ‘piecemeal’ approach of assessing the reliability of particular belief-forming mechanisms in particular domains. This stands in contrast to the more familiar attempt to transfer warrant obtained for one domain to another by showing how one is somehow an extension of the other. We offer a naturalist reply to the charge of circularity by appealing to reliabilist work on the problem of induction, notably Peter Lipton's distinction between self-certifying and non-self-certifying inductive arguments. We show how, for scientific beliefs, a non-self-certifying argument might be made for the reliability of our cognitive faculties in that domain. We call this strategy Humean Bootstrapping. (shrink)

The present paper proposes a route to modal claims that allows us to infer to certain possibilities even if they are sensorily unimaginable and beyond the evidential capacity of stipulative imagining. After a brief introduction, Sect. 2 discusses imaginative resistance to help carve a niche for the kinds of inferences about which this essay is chiefly concerned. Section 3 provides three classic examples, along with a discussion of their similarities and differences. Section 4 recasts the notion of potential explanation in (...) Lipton’s (Inference to the best explanation, Routledge, Abingdon, 2004) in order to accommodate inferences to possibility claims; Sect. 5 then attempts to characterise a principle underlying such inferences. Section 6 concludes by discussing how the proposal relates to other modal epistemologies, with emphasis on the potential of such inferences to produce genuinely new ideas. (shrink)

Many scientists and philosophers of science think that beauty should play a role in theory selection. Physicists like Paul Dirac and Steven Weinberg explicitly claim that the ultimate explanations of the physical world must be beautiful. And philosophers of science like Peter Lipton say that we should expect the loveliest theory to also be the most likely. In this paper, I contend that these arguments from loveliness bear a striking similarity to Thomas Aquinas’ arguments from fittingness; both seem to (...) presume that the most beautiful theory is also the most probable. To do this, I first explain the explanatory virtues that are commonly thought to be constitutive of a lovely theory. Second, I elucidate Aquinas’ arguments from fittingness and show how they work in light of his account of beauty. Lastly, I connect the two kinds of aesthetic arguments and show the extent to which they overlap. (shrink)

IBE ('Inference to the best explanation' or abduction) is a popular and highly plausible theory of how we should judge the evidence for claims of past events based on present evidence. It has been notably developed and supported recently by Meyer following Lipton. I believe this theory is essentially correct. This paper supports IBE from a probability perspective, and argues that the retrodictive probabilities involved in such inferences should be analysed in terms of predictive probabilities and a priori probability (...) ratios of initial events. The key point is to separate these two features. Disagreements over evidence can be traced to disagreements over either the a priori probability ratios or predictive conditional ratios. In many cases, in real science, judgements of the former are necessarily subjective. The principles of iterated evidence are also discussed. The Sceptic's position is criticised as ignoring iteration of evidence, and characteristically failing to adjust a priori probability ratios in response to empirical evidence. (shrink)

Dans cet article, nous examinerons le thème de la découverte en science. Nous soutiendrons qu’il est possible de définir une stratégie rationnelle soutenant la découverte au moyen du principe de l’abduction. Afin de démontrer cette thèse, il s’agira d’abord de diagnostiquer le problème de la complexité et de l’irrationalité de la découverte scientifique et de considérer la position néo-positiviste endossée par Karl Popper et Hans Reichenbach, selon laquelle la découverte ne peut faire l’objet d’une étude proprement logique ou épistémologique. Par (...) la suite, l’analyse se concentrera sur la position de Norwood R. Hanson qui, répondant et contredisant celle de Popper et de Reichenbach, affirme qu’il existe une « logique de la découverte », fondée sur le principe de l’abduction. Après avoir déployé l’argument de Hanson et explicité sa conception du principe de l’abduction, nous montrerons, en nous appuyant sur la vision d’Atocha Aliseda, comment l’abduction est liée à la découverte en science. Finalement, nous examinerons la notion d’inférence à la meilleure explication et le concept « d’hypothèse explicative » de Peter Lipton, ce qui permettra d’expliquer dans quelle mesure cette variante de l’abduction nous permet de comprendre les éléments et conditions qui favorisent la génération de découvertes en science. (shrink)

This paper evaluates the claim that it is possible to use nature’s variation in conjunction with retention and selection on the one hand, and the absence of ultimate groundedness of hypotheses generated by the human mind as it knows on the other hand, to discard the ascription of ultimate certainty to the rationality of human conjectures in the cognitive realm. This leads to an evaluation of the further assumption that successful hypotheses with specific applications, in other words heuristics, seem to (...) have a firm footing because they were useful in another context. I argue that usefulness evaluated through adaptation misconstrues the search for truth, and that it is possible to generate talk of randomness by neglecting aspects of a system’s insertion into a larger situation. The framing of the problem in terms of the elimination of unfit hypotheses is found to be unsatisfying. It is suggested that theories exist in a dimension where they can be kept alive rather than dying as phenotypes do. The proposal that the subconscious could suggest random variations is found to be a category mistake. A final appeal to phenomenology shows that this proposal is orphan in the history of epistemology, not in virtue of its being a remarkable find, but rather because it is ill-conceived. (shrink)

In this paper art history and visual studies, the disciplines that study visual culture, are presented as a field whose conjectural paradigm can be used to understand the epistemic problems associated with abduction. In order to do so, significant statements, concepts and arguments from the work of several specialists in this field have been highlighted. Their analysis shows the fruitfulness and potential for understanding the study of visual culture as a field that is interwoven with the assumptions of abductive cognition.