Assistive systems based on Artificial Intelligence (AI) are bound to reshape decision-making in all areas of society. One of the most intricate challenges arising from their implementation in high-stakes environments such as medicine concerns their frequently unsatisfying levels of explainability, especially in the guise of the so-called black-box problem: highly successful models based on deep learning seem to be inherently opaque, resisting comprehensive explanations. This may explain why some scholars claim that research should focus on rendering AI systems understandable, rather (...) than explainable. Yet, there is a grave lack of agreement concerning these terms in much of the literature on AI. We argue that the seminal distinction made by the philosopher and physician Karl Jaspers between different types of explaining and understanding in psychopathology can be used to promote greater conceptual clarity in the context of Machine Learning (ML). Following Jaspers, we claim that explaining and understanding constitute multi-faceted epistemic approaches that should not be seen as mutually exclusive, but rather as complementary ones as in and of themselves they are necessarily limited. Drawing on the famous example of Watson for Oncology we highlight how Jaspers’ methodology translates to the case of medical AI. Classical considerations from the philosophy of psychiatry can therefore inform a debate at the centre of current AI ethics, which in turn may be crucial for a successful implementation of ethically and legally sound AI in medicine. (shrink)

In this paper I argue that the search for explainable models and interpretable decisions in AI must be reformulated in terms of the broader project of offering a pragmatic and naturalistic account of understanding in AI. Intuitively, the purpose of providing an explanation of a model or a decision is to make it understandable to its stakeholders. But without a previous grasp of what it means to say that an agent understands a model or a decision, the explanatory strategies will (...) lack a well-defined goal. Aside from providing a clearer objective for XAI, focusing on understanding also allows us to relax the factivity condition on explanation, which is impossible to fulfill in many machine learning models, and to focus instead on the pragmatic conditions that determine the best fit between a model and the methods and devices deployed to understand it. After an examination of the different types of understanding discussed in the philosophical and psychological literature, I conclude that interpretative or approximation models not only provide the best way to achieve the objectual understanding of a machine learning model, but are also a necessary condition to achieve post hoc interpretability. This conclusion is partly based on the shortcomings of the purely functionalist approach to post hoc interpretability that seems to be predominant in most recent literature. (shrink)

The paper presents two objections against Putnam’s Twin Earth argument, which was intended to secure semantic externalism. I first claim that Putnam’s reasoning rests on two assumptions and then try to show why these assumptions are contentious. The first objection is that, given what we know about science, it is unlikely that there are any natural-kind terms whose extension is codetermined by a small set of microstructures required by Putnam’s indexical account of extension determination. The second objection is that there (...) may not be a plausible concept of a speech community whose adoption would classify Oscar and Twin Oscar as members of different speech communities and, at the same time, render Oscar and Twin Oscar as being in the same psychological state. I contend that Putnam’s argument fails because both objections are justified. (shrink)

There has been a probabilistic turn in contemporary cognitive science. Far and away, most of the work in this vein is Bayesian, at least in name. Coinciding with this development, philosophers have increasingly promoted Bayesianism as the best normative account of how humans ought to reason. In this paper, we make a push for exploring the probabilistic terrain outside of Bayesianism. Non-Bayesian, but still probabilistic, theories provide plausible competitors both to descriptive and normative Bayesian accounts. We argue for this general (...) idea via recent work on explanationist models of updating, which are fundamentally probabilistic but assign a substantial, non-Bayesian role to explanatory considerations. (shrink)

Many theories of kind representation suggest that people posit internal, essence-like factors that underlie kind membership and explain properties of category members. Across three studies (N = 281), we document the characteristics of an alternative form of construal according to which the properties of social kinds are seen as products of structural factors: stable, external constraints that obtain due to the kind’s social position. Internalist and structural construals are similar in that both support formal explanations (i.e., “category member has property (...) P due to category membership C”), generic claims (“Cs have P”), and the generalization of category properties to individual category members when kind membership and social position are confounded. Our findings thus challenge these phenomena as signatures of internalist thinking. However, once category membership and structural position are unconfounded, different patterns of generalization emerge across internalist and structural construals, as do different judgments concerning category definitions and the dispensability of properties for category membership. We discuss the broader implications of these findings for accounts of formal explanation, generic language, and kind representation. (shrink)

Most philosophers agree that abduction (in the sense of Inference to the Best Explanation) is a type of inference that is frequently employed, in some form or other, both in everyday and in scientific reasoning. However, the exact form as well as the normative status of abduction are still matters of controversy. This entry contrasts abduction with other types of inference; points at prominent uses of it, both in and outside philosophy; considers various more or less precise statements of it; (...) discusses its normative status; and highlights possible connections between abduction and Bayesian confirmation theory. (shrink)

This paper aims at elucidating the connection between explanatory considerations and epistemic justification stipulated by explanationism which take epistemic justification to be definable in terms of best explanations. By relying on the notion of truthlikeness, this paper argues that it is rational for a subject to expect the best explanation she has for her evidence to be more truthlike than any of the other potential explanations available to her by virtue of containing a class of propositions that, given her evidence, (...) she is justified in believing. Based on this elucidation of the connection between explanatory considerations and epistemic justification, an explanationist account of the evidential support relation is then offered. (shrink)

Walter Bock was committed to developing a framework for evolutionary biology. Bock repeatedly discussed how evolutionary explanations should be considered within the realm of Hempel’s deductive-nomological model of scientific explanations. Explanation in evolution would then consist of functional and evolutionary explanations, and within the latter, an explanation can be of nomological-deductive and historical narrative explanations. Thus, a complete evolutionary explanation should include, first, a deductive functional analysis, and then proceed through nomological and historical evolutionary explanations. However, I will argue that (...) his views on the deductive proprieties of functional analysis and the deductive-nomological parts of evolution fail because of the nature of evolution, which contains a historical element that the logic of deduction and Hempel’s converting law model do not compass. Conversely, Bock’s historical approach gives a critical consideration of the historical narrative element of evolutionary explanation, which is fundamental to the methodology of the historical nature of evolutionary theory. Herein, I will expand and discuss a modern view of evolutionary explanations of traits that includes the currentacknowledgement of the differences between experimental and the historical sciences, including the token and type event dichotomy, that mutually illuminate each other in order to give us a well confirmed and coherent hypothesis for evolutionary explanations. Within this framework, I will argue that the duality of evolutionary explanations is related to two components of character evolution: origin, with its evolutionary pathways along with the history, and maintenance, the function (mainly a current function) for the character being selected. (shrink)

In this paper I argue that the search for explainable models and interpretable decisions in AI must be reformulated in terms of the broader project of offering a pragmatic and naturalistic account of understanding in AI. Intuitively, the purpose of providing an explanation of a model or a decision is to make it understandable to its stakeholders. But without a previous grasp of what it means to say that an agent understands a model or a decision, the explanatory strategies will (...) lack a well-defined goal. Aside from providing a clearer objective for XAI, focusing on understanding also allows us to relax the factivity condition on explanation, which is impossible to fulfill in many machine learning models, and to focus instead on the pragmatic conditions that determine the best fit between a model and the methods and devices deployed to understand it. After an examination of the different types of understanding discussed in the philosophical and psychological literature, I conclude that interpretative or approximation models not only provide the best way to achieve the objectual understanding of a machine learning model, but are also a necessary condition to achieve post hoc interpretability. This conclusion is partly based on the shortcomings of the purely functionalist approach to post hoc interpretability that seems to be predominant in most recent literature. (shrink)

Cognitive Science, Volume 46, Issue 2, February 2022.

At the core of anthropomorphism lies a false-positive cognitive bias to over-attribute the pattern of the human body and/or mind. Anthropomorphism is independently discussed in various disciplines, is presumed to have deep biological roots, but its cognitive bases are rarely explored in an integrative way. I present an inclusive, multifaceted interdisciplinary approach to refine the psychological bases of mental anthropomorphism. I have integrated 13 conceptual dissections of folk finalistic reasoning into four psychological inference systems (physical, design, basic-goal and belief stances); (...) the latter three are truly teleological and thus prone to anthropomorphisms. I then have empirically integrated the cross-disciplinary genetic, neural, cognitive, psychiatric, developmental, comparative and evolutionary/adaptive evidence that converge supporting the nature of the distinct stances. The over-reactive calibration of the three teleological systems is framed as an evolved design feature to avoid harmful ancestral contexts. Nowadays, these stances easily engage with scientific reasoning about bio-evolutionary matters with negative and positive consequences. Design, basic-goal, and belief stances benefit biology by providing cognitive foundations, expressing a high-powered explanatory system, promoting functional generalization, fostering new research questions and discoveries, enabling metaphorical/analogical thinking and explaining didactically with brevity. Hence, it is neither feasible nor advantageous to completely eliminate teleology from biology. Instead, we should engage with the eight classes of problems in bio-philosophy and bio-education that relate to the three stances: types of anthropomorphism, variety of misunderstandings, misleading appeal, legitimacy controversy, gateway to mysticism, total prohibition and its backfire effect. Recognizing the distinction among design, basic-goal, and belief stances helps to elucidate much of the logic underlying these issues, so that it enables a much more detailed taxonomy of anthropomorphisms, and organizes the various misunderstandings about evolution by natural selection. It also offers a solid psychological grounding for anchoring definitions and terminology. This tripartite framework also shed some light on how to better deal with the over-reactive stances in bio-education, by organizing previous pedagogical strategies and by suggesting new possibilities to be tested. Therefore, this framework constitutes a promising approach to advance the debate regarding the psychological underpinnings of anthropomorphisms and to further support regulating and clarifying teleology and anthropomorphism in biology. (shrink)