We argue that artificial networks are explainable and offer a novel theory of interpretability. Two sets of conceptual questions are prominent in theoretical engagements with artificial neural networks, especially in the context of medical artificial intelligence: Are networks explainable, and if so, what does it mean to explain the output of a network? And what does it mean for a network to be interpretable? We argue that accounts of “explanation” tailored specifically to neural networks have ineffectively reinvented the wheel. In (...) response to, we show how four familiar accounts of explanation apply to neural networks as they would to any scientific phenomenon. We diagnose the confusion about explaining neural networks within the machine learning literature as an equivocation on “explainability,” “understandability” and “interpretability.” To remedy this, we distinguish between these notions, and answer by offering a theory and typology of interpretation in machine learning. Interpretation is something one does to an explanation with the aim of producing another, more understandable, explanation. As with explanation, there are various concepts and methods involved in interpretation: Total or Partial, Global or Local, and Approximative or Isomorphic. Our account of “interpretability” is consistent with uses in the machine learning literature, in keeping with the philosophy of explanation and understanding, and pays special attention to medical artificial intelligence systems. (shrink)

We revisit the analogy suggested by Madelung between a non-relativistic time-dependent quantum particle, to a fluid system which is pseudo-barotropic, irrotational and inviscid. We first discuss the hydrodynamical properties of the Madelung description in general, and extract a pressure like term from the Bohm potential. We show that the existence of a pressure gradient force in the fluid description, does not violate Ehrenfest’s theorem since its expectation value is zero. We also point out that incompressibility of the fluid implies conservation (...) of density along a fluid parcel trajectory and in 1D this immediately results in the non-spreading property of wave packets, as the sum of Bohm potential and an exterior potential must be either constant or linear in space. Next we relate to the hydrodynamic description a thermodynamic counterpart, taking the classical behavior of an adiabatic barotopric flow as a reference. We show that while the Bohm potential is not a positive definite quantity, as is expected from internal energy, its expectation value is proportional to the Fisher information whose integrand is positive definite. Moreover, this integrand is exactly equal to half of the square of the imaginary part of the momentum, as the integrand of the kinetic energy is equal to half of the square of the real part of the momentum. This suggests a relation between the Fisher information and the thermodynamic like internal energy of the Madelung fluid. Furthermore, it provides a physical linkage between the inverse of the Fisher information and the measure of disorder in quantum systems—in spontaneous adiabatic gas expansion the amount of disorder increases while the internal energy decreases. (shrink)

Philosophical Issues, Volume 31, Issue 1, Page 85-103, October 2021.

The paper describes tableaux based proof methods for temporal logics of knowledge allowing non-trivial interaction axioms between the modal and temporal components, namely those of synchrony and no learning and synchrony and perfect recall. The interaction axioms allow the description of how knowledge evolves over time and makes reasoning in such logics theoretically more complex. Such logics can be used to specify systems that involve the knowledge of processes or agents and which change over time, for example agent based systems, (...) knowledge games or security protocols. Soundness, completeness and termination results for the tableaux are proved. (shrink)

After 50 years of being profoundly deaf, Patricia finds her world ‘transformed’—literally and metaphorically—when she receives a cochlear implant. Her sense of self and the taken-for-granted, comfortable world she knew before surgery disappear and she is thrown into an alien, surreal existence full of hyper-noise. Entry into this new world of sounds proves a mixed blessing as Pat struggles to come to terms with her changing relationships, not only with others but also with herself. On good days, she is exhilarated (...) by all her sensory gains and her feeling of being more connected with and to the world. On bad days she is distracted and overwhelmed by the intrusive noise and she is forcibly confronted with the painful reality of her own disability. The challenge she confronts is not simply the cognitive-perceptual one of learning to discriminate between sounds. Pat must also re-orientate herself and learn to cope with her transformed self and world. She must undertake a journey to come to terms with her past, present and future being. Pat shared her story with me and together we undertook collaborative existential phenomenological research, co-creating a narrative of her journey over the year and a half following her implant. This paper presents this narrative followed by an existential analysis of Pat’s disrupted, changing lifeworld. (shrink)

Both left libertarians, who support the redistribution of income and wealth through taxation, and right libertarians, who oppose redistributive taxation, share an important view: that, looming catastrophes aside, the state must never redistribute any part of our body or our person without our consent. Cécile Fabre rejects that view. For her, just as the undeservedly poor have a just claim to money from their fellow citizens in order to lead a minimally flourishing life, the undeservedly ‘medically poor’ have a just (...) claim to help from fellow citizens in order to lead such a life. Such obligatory help may in principle involve even the supply of body parts for transplantation. The state ought to exact such resources from the medically rich whenever doing so would secure the prospect of a minimally flourishing life to the medically poor without denying that prospect to anyone else. Fabre criticizes Ronald Dworkin's belief in ‘a prophylactic line that comes close to making the body inviolate, that is, making body parts not parts of social resources at all’. For her, ‘Duties to help. . . do not stop at material resources: they involve the body. . . in invasive ways’. (shrink)

I argue that the denial of speakers' knowledge of language rules is based on conceptual confusion and in particular on a misanalysis of what it is to know a rule. I shall turn to the task of establishing this point after first providing the background for this issue: the difficulty of conceptualizing verbal behavior both under the hypothesis that speakers do, and under the hypothesis that speakers do not know the rules of grammar. I shall argue that this difficulty and (...) the dispute it has engendered are based on overly simple and inadequate views of rule knowledge. Our ordinary concept of knowing a rule has been radically ignored by theorists who have written on this issue. In consequence, they have operated with conceptual schemes too impoverished to allow for the knowledge that speakers do have. A more adequate analysis or account of rule knowledge will thus be the main concern of this paper with the problem of speakers’ knowledge of grammar being the motivation to attempt such an analysis. (shrink)

Natural environments differ from artworks in two ways: (a) they are surroundings filled with objects, processes, and the observer, (b) they are natural, not intentionally created to be appreciated. I show that this serious problem for accounts of aesthetic appreciation of nature has led many thinkers in environmental aesthetics (e.g., Carlson and Rolston) to claim that appreciators should be actively engaged with a natural environment. But how? One suggestion has been that appreciators play the role of creative performers in the (...) arts. I explore this analogy, distinguishing three different kinds of performance. I argue that none is a good fit as a model of nature appreciation but that the analogy sheds considerable light on environmental art, especially the site-specific artworks of Andy Goldsworthy. (shrink)

The issues of difference, diversity, otherness, and the possibility of community have emerged as leading philosophical and socio-political questions in recent times. An increasing theoretical emphasis on issues of difference and alterity, with a corresponding convergence of topics like globalization, have brought into renewed focus not only the question of otherness, but more fundamentally, the question of how the encounter with the Other is to take place. My paper examines the treatment of issues of difference and otherness within a perspective (...) of philosophical hermeneutics. In recent writings, Gadamer has suggested that the principles of his philosophical hermeneutics can prove beneficial in reflections on contemporary social problems of diversity and cultural difference. In examining these claims, I question whether Gadamer acknowledges sufficiently the challenges and resistance of otherness. At the same time, I also question the claims of some theorists who maintain that it is never possible to bridge difference and otherness. In the end I suggest that while the resistance of difference might be underestimated, there are nevertheless promising approaches for thinking difference and diversity to be found in a hermeneutical perspective. (shrink)

Confessions both hold a great promise and pose a grave danger. When the accused speaks against his interest and assumes responsibility for criminal actions this is viewed as a compelling sign of guilt. It is not, therefore, for naught that the confession has been crowned the "queen of evidence." Yet research conducted in the last few decades has shown that a substantial number of confessions are false, ranking the out of court confession high among the factors leading to the conviction (...) of innocent people. The acceptance of DNA testing, has further substantiated this finding, igniting a renewed interest in out-of-court confessions. It is hard to assess the magnitude of social harm caused by the widespread and persistent resort to obtaining confessions, but one thing is evident from the vast literature on the matter: this evidentiary mechanism is currently overused, much beyond the optimal level. The reasons for this over-use are varied, among them the fact that confessions are readily available for law enforcement officers, and that those in charge of trying the facts tend to overwhelmingly convict based on out-of-court confessions, while underestimating the associated dangers. As a result, confessions have come to play a paramount role in the criminal justice system. Many attempts have been made to deal with confession-based wrongful convictions. Examples include the Miranda rules or the requirements for varying degrees and types of corroborating evidence. To date, all the rules and proposals share a common characteristic: They all attempt to correct the evidentiary fallacies associated with out-of-court confessions by evidentiary means, whether admissibility oriented or weight oriented. In the article we argue that none of the proposed mechanisms is likely to solve the problem. The root of the problem lies in the fact that the entire criminal justice system is currently organized around confessions. Law enforcement officers focus on obtaining confessions and the prosecution uses it as its evidentiary centerpiece. Further restricting admissibility or requiring a higher degree of corroboration will not change this trend. Moreover, the confessional lure is too strong to resist. Like the Sirens' Song, the confession casts a spell on all those subjected to it and, no matter how strong a corroboration we require, we will eventually fall back on the confession. We propose an entirely different solution to the tendency to over-use confessions, one that utilizes penal means to change the incentive structure within the criminal justice system. We propose to incorporate into the sentencing guidelines a mandatory reduction of the criminal sanction whenever an out-of-court confession is introduced into evidence by the prosecution. In other words, our proposal is to elevate the cost of using out-of court confessions, as compared to other types of evidence. The article will demonstrate how placing such a "sentencing price tag" on the use of out-of-court confessions will correct the current bias in favor of using this evidence and induce law enforcement officials to seek extrinsic evidence, thus turning confessions into a residual evidentiary devise. The article will also show that the proposed regime will improve the court's ability to distinguish between true and false confessions. After discussing the proposed model's normative appeal, the article will proceed to deal with possible criticisms which can be leveled against it, whether retributive, utilitarian or expressive. (shrink)

We examine whether the "evidence of evidence is evidence" principle is true. We distinguish several different versions of the principle and evaluate recent attacks on some of those versions. We argue that, whatever the merits of those attacks, they leave the more important rendition of the principle untouched. That version is, however, also subject to new kinds of counterexamples. We end by suggesting how to formulate a better version of the principle that takes into account those new counterexamples.

We present a new theoretical construct labeled motivational readiness. It is defined as the inclination, whether or not ultimately implemented, to satisfy a desire. A general model of readiness is described which builds on the work of prior theories, including animal learning models and personality approaches, and which aims to integrate a variety of research findings across different domains of motivational research. Components of this model include the Want state, and the Expectancy of being able to satisfy that Want. We (...) maintain that the Want concept is the critical ingredient in motivational readiness: without it, readiness cannot exist. In contrast, some motivational readiness can exist without Expectancy. We also discuss the role of incentive in motivational readiness. Incentive is presently conceived of in terms of a Match between a Want and a Perceived Situational Affordance. Whereas in classic models incentive was portrayed as a first order determinant of motivational readiness, here we describe it as a second order factor which affects readiness by influencing Want, Expectancy, or both. The new model’s relation to its theoretical predecessors, and its implications for future research, also are discussed. (shrink)

Two experiments demonstrated letter-context effects that cannot easily be accounted for by postperceptual theories based on structural redundancy, iigural goodness, or memory advantage. In Experiment 1, subjects identified the color of a letter fragment more accurately in letter than in nonletter contexts. In Experiment 2, subjects identified the feature presented in a precued color more accurately in letters than in nonletters. We argue that these effects result from topdown perceptual processing.

It is argued here that a critical prediction of the E-Z Reader model is that experimental manipulations that disrupt early encoding of visual and orthographic features of the fixated word without affecting subsequent lexical processing should influence the processing difficulty of the fixated word without producing any processing effect on the next word. This prediction is explained and illustrated.

This article deploys sadomasochism as a framework for understanding medical practice on an institutional level. By examining the case of the factitious illness Munchausen syndrome, this article analyzes the operations of power in the doctor-patient relationship through the trope of role-playing. Because Munchausen syndrome causes a disruption to the dyadic relationship between physicians and patients, a lens of sadomasochism highlights dynamics of power in medical practice that are often obscured in everyday practice. Specifically, this article illustrates how classification and diagnosis (...) are concrete manifestations of the mobilization of medical power. (shrink)

A comprehensive collection of the writings of Mark Fisher (1968-2017), whose work defined critical writing for a generation. This comprehensive collection brings together the work of acclaimed blogger, writer, political activist and lecturer Mark Fisher (aka k-punk). Covering the period 2004 - 2016, the collection will include some of the best writings from his seminal blog k-punk; a selection of his brilliantly insightful film, television and music reviews; his key writings on politics, activism, precarity, hauntology, mental health and (...) popular modernism for numerous websites and magazines; his final unfinished introduction to his planned work on "Acid Communism"; and a number of important interviews from the last decade. Edited by Darren Ambrose and with a foreword by Simon Reynolds. (shrink)

Suppose we learn that we have a poor track record in forming beliefs rationally, or that a brilliant colleague thinks that we believe P irrationally. Does such input require us to revise those beliefs whose rationality is in question? When we gain information suggesting that our beliefs are irrational, we are in one of two general cases. In the first case we made no error, and our beliefs are rational. In that case the input to the contrary is misleading. In (...) the second case we indeed believe irrationally, and our original evidence already requires us to fix our mistake. In that case the input to that effect is normatively superfluous. Thus, we know that information suggesting that our beliefs are irrational is either misleading or superfluous. This, I submit, renders the input incapable of justifying belief revision, despite our not knowing which of the two kinds it is. (shrink)

Law, Economics, and Morality examines the possibility of combining economic methodology and deontological morality through explicit and direct incorporation of moral constraints into economic models. Economic analysis of law is a powerful analytical methodology. However, as a purely consequentialist approach, which determines the desirability of acts and rules solely by assessing the goodness of their outcomes, standard cost-benefit analysis is normatively objectionable. Moderate deontology prioritizes such values as autonomy, basic liberties, truth-telling, and promise-keeping over the promotion of good outcomes. It (...) holds that there are constraints on promoting the good. Such constraints may be overridden only if enough good is at stake. While moderate deontology conforms to prevailing moral intuitions and legal doctrines, it is arguably lacking in methodological rigor and precision. Eyal Zamir and Barak Medina argue that the normative flaws of economic analysis can be rectified without relinquishing its methodological advantages and that moral constraints can be formalized so as to make their analysis more rigorous. They discuss various substantive and methodological choices involved in modeling deontological constraints. Zamir and Medina propose to determine the permissibility of any act or rule infringing a deontological constraint by means of mathematical threshold functions. Law, Economics, and Morality presents the general structure of threshold functions, analyzes their elements and addresses possible objections to this proposal. It then illustrates the implementation of constrained CBA in several legal fields, including contract law, freedom of speech, antidiscrimination law, the fight against terrorism, and legal paternalism. (shrink)

Advances in artificial intelligence (AI) raise important questions about whether people view moral evaluations by AI systems similarly to human-generated moral evaluations. We conducted a modified Moral Turing Test (m-MTT), inspired by Allen et al. (Exp Theor Artif Intell 352:24–28, 2004) proposal, by asking people to distinguish real human moral evaluations from those made by a popular advanced AI language model: GPT-4. A representative sample of 299 U.S. adults first rated the quality of moral evaluations when blinded to their source. (...) Remarkably, they rated the AI’s moral reasoning as superior in quality to humans’ along almost all dimensions, including virtuousness, intelligence, and trustworthiness, consistent with passing what Allen and colleagues call the comparative MTT. Next, when tasked with identifying the source of each evaluation (human or computer), people performed significantly above chance levels. Although the AI did not pass this test, this was not because of its inferior moral reasoning but, potentially, its perceived superiority, among other possible explanations. The emergence of language models capable of producing moral responses perceived as superior in quality to humans’ raises concerns that people may uncritically accept potentially harmful moral guidance from AI. This possibility highlights the need for safeguards around generative language models in matters of morality. (shrink)

The article begins by reconstructing the just distribution of the social bases of self-respect, a principle of justice that is covert in Rawls’s writing. I argue that, for Rawls, justice mandates that each social basis for self-respect be equalized. Curiously, for Rawls, that principle ranks higher than Rawls’s two more famous principles of justice - equal liberty and the difference principle. I then recall Rawls’s well-known confusion between self-respect and another form of self-appraisal, namely, confidence in one’s determinate plans and (...) capacities. Correcting that confusion forces Rawls to accept objectionable and illiberal politics. Surprisingly, a consistent Rawls must endorse absolute economic equality, deny liberty any priority whatsoever, or sponsor still other illiberal political views - evidence of a flaw in the ethical basis of Rawls’s politics. (shrink)

This paper merges the non-expected utility approach (Tversky and Kahneman, J Risk Uncertain 5:297–323, 1992 and Quiggin, J Econ Behav Organ 3:323–343, 1982) into Akerlof’s (Quart J Econ 84:488–500, 1970) model of Market for Lemons. We derive the results for different probability weighting functions and analyze the phenomenon of market failure in light of non-expected utility maximization. Our main finding suggests that when the proportion of traded lemons is high (low), the problem of market failure is mitigated (enhanced). In addition, (...) for the case of Cumulative Prospect Theory, we show that (a) the higher the loss aversion is, the more pronounced is the market failure; (b) gain-domain elevation is negatively related to the extent of market failure; and (c) the value function is (i) negatively monotonic in the gain-domain diminishing sensitivity parameter when the market is characterized by a high proportion of “peaches,” and (ii) positively monotonic in the loss-domain diminishing sensitivity parameter when the market is characterized by a high proportion of “lemons.”. (shrink)

The Self-Intimation thesis has it that whatever justificatory status a proposition has, i.e., whether or not we are justified in believing it, we are justified in believing that it has that status. The Infallibility thesis has it that whatever justificatory status we are justified in believing that a proposition has, the proposition in fact has that status. Jointly, Self-Intimation and Infallibility imply that the justificatory status of a proposition closely aligns with the justification we have about that justificatory status. Self-Intimation (...) has two noteworthy implications. First, assuming that we never have sufficient justification for a proposition and for its negation, we can derive Infallibility from Self-Intimation. Interestingly, there seems to be no equivalently simple way to derive Self-Intimation from Infallibility. This asymmetry provides reason for thinking that bottom-level justification rather than top-level justification drives the explanation for why the levels of justification align. Second, Self-Intimation suggests a counterintuitive treatment of information concerning what justificatory status a proposition has. It follows from Self-Intimation that we always have justification for the truth about whether a proposition is justified for us, and therefore, that higher-order evidence could change what we should believe on this matter only by misleading us. This permits forming beliefs about whether a proposition is justified for us without regard to higher-order evidence, and thus reveals a reason for thinking that top-level justification is evidentially inert. (shrink)

Is it ever rational to suspend judgment about whether a particular doxastic attitude of ours is rational? An agent who suspends about whether her attitude is rational has serious doubts that it is. These doubts place a special burden on the agent, namely, to justify maintaining her chosen attitude over others. A dilemma arises. Providing justification for maintaining the chosen attitude would commit the agent to considering the attitude rational—contrary to her suspension on the matter. Alternatively, in the absence of (...) such justification, the attitude would be arbitrary by the agent's own lights, and therefore irrational from the agent's own perspective. So, suspending about whether an attitude of ours is rational does not cohere with considering it rationally preferable to other attitudes, and leads to a more familiar form of epistemic akrasia otherwise. (shrink)

In political philosophy and in economics, unfair inequality is usually assessed between individuals, nowadays often on luck-egalitarian grounds. You have more than I do and that's unfair. By contrast, in epidemiology and sociology, unfair inequality is traditionally assessed between groups. More is concentrated among people of your class or race than among people of mine, and that's unfair. I shall call this difference the egalitarian ‘divorce’. Epidemiologists, and their ‘divorce lawyers’ Paula Braveman, Norman Daniels, and Iris Marion Young, explain that (...) not every inequality between individuals is an inequity. Only inequality between social status groups is unfair. Only such inequality stems from problems like partiality, discrimination, oppression, inequality-related population health problems, and unfair distribution of prospects. And it alone is actionable. By contrast, inequality between individuals, e.g. in longevity, is natural, inevitable, less important, or otherwise less informative — not unjust. In ‘divorce trial’ mode, I respond to epidemiologists that group inequalities lack the intrinsic disvalue that they ascribe them. They may be instrumentally or contingently bad or wrong, not essentially so. I then shift to ‘remarriage’: although group inequalities lack intrinsic badness or wrongness, for multiple reasons they remain useful to measure. (shrink)