Zum Inhalt springenZur Suche springen

The Cognitive and Ontological Dimensions of Naturalness

Debates about naturalness go far back into the history of philosophy – just think of the famous (yet somewhat morbid) metaphor of cutting nature at its joints, harking back to Plato's Phaedrus. The same is true of a whole range of related topics, with the problem of projectibility in the philosophy of science standing out in particular. Whereas, according to a famous line of argument by Nelson Goodman (Fact, Fiction, and Forecast, 1955), we use some predicates in our arguments (e.g., being green), we do not do so with other predicates, sometimes called "gerrymandered" (e.g., being grue). We may refer to the former as projectible predicates, while the latter are non-projectible ones. As there is no obvious criterion in sight for distinguishing these kinds of predicates, however, this seems to undermine the justifiability of our reasoning practices. Now, it seems plausible to assume that projectible predicates are precisely the ones that refer to natural entities. This, however, raises the question of the criteria of naturalness – and this is where the workshop takes off.

An interesting proposal comes from recent cognitive research. It was developed by Peter Gärdenfors and is based on geometric properties of cognitive representations. In a nutshell, his idea is that natural properties and concepts are convex regions in conceptual spaces. It has proved fruitful, not only in explaining the behavior of cognitive agents, but also in constructing artificial agents. It explores the issue of naturalness from a pragmatist and instrumentalist design perspective. However, the philosophical foundations of the approach provide room for controversy. According to Gärdenfors, natural properties and concepts are internal, cognitive entities. But current analytic metaphysicians such as David Lewis, who has been a key figure in shaping the debate over naturalness, are typically more concerned with the structure of external reality – Lewis would probably have dismissed Gärdenfors’ account as psychologistic. There thus seem to be two distinct dimensions of naturalness, namely a cognitive and an ontological one – which raises the question of the relation between the two. One central aim of the workshop is to bring together proponents of both camps in order to get to the nature of this relation.

Abstracts (In alphabetical order)

  • Co-authors: Lina Bendifallah , Igor Douven
  • Title: Naturalness and social concepts
  • Abstract: A central aim of the conceptual spaces theory is to establish criteria of what makes a concept natural and thus to distinguish concepts that play, or can play, a role in our thinking and theorizing from those that do not. The best known criterion is the so-called convexity principle, according to which natural concepts are representable as a convex region in a conceptual space. A more recent and more general approach has proposed a number of optimal design principles as criteria, which subsume the convexity criterion. However, these principles were in first instance only proposed with perceptual concepts in mind. Given that the conceptual spaces framework purports to apply to a broad range of concepts and that social concepts play a central role in our thinking, theorizing and decision-making, the question arises of whether (and if so, how) the optimality proposal should be extended to this kind of concepts. We propose to answer this question in two steps. First, we will present a principled method, based on the use of the conceptual spaces theory, that has been empirically validated to represent systems of social concepts. While some concepts seem highly subjective, this new way of representing social concepts will help us put them on solid grounds, notably by showing that they correspond to the naturalness constraints as implied by the optimality proposal. The second step we need to take to include social concepts will require an extension of our definition of naturalness which builds on Lewis’ Best System Analysis.

Author: Dr. Verónica Gómez Sánchez

Title: No thought without magnets

Abstract: According to the Lewisian doctrine of reference magnetism, representation/reference relations do not only depend on how representations are deployed by their users, but also on the degree of naturalness of the represented entities. The main thesis of this talk is that any viable reductive account of reference will have to presuppose a version of reference magnetism. My first goal is to make this thesis precise, and to refine and generalize a line of argument for it sketched in Sider (2011). My second goal is to work toward a positive characterization of the connection between reference and naturalness. I argue that, while one familiar way of drawing this connection is problematic, there is a promising alternative which avoids a number of standard objections in the literature. This alternative allows for a unified account of reference, while vindicating the idea that naturalness plays a more important role in determining reference for natural kind concepts than it does for other concepts.

  • Co-authors: Siegfried Jaag & Christian Löw
  • Title: Natural Properties: Does the Humean Best Systems Account need them?
  • Abstract: David Lewis famously argued that for the Humean best systems account (BSA) of laws to work, a distinguished vocabulary for best systems needs to be provided. Lewis proposes a metaphysical solution to this problem that involves positing perfectly natural properties. Pragmatic Humeans, instead, propose an pragmatic/epistemic solution to the same problem: They argue that the language of best system is determined by its practical usefulness for prediction and explanation. In this talk we explain, what exactly the epistemic solution amounts to and examine its broader ramifications for the Humean metaphysics of laws.
  • Author: Matías Osta Vélez
  • Title: From feature correlations to conceptual coherence
  • Abstract: Our conceptual repertoire is highly diverse. We have concepts for natural kinds, such as bird or fruit; for artificial kinds and artifacts, such as bachelor and hammer; and for things that we group in the service of certain goals, such as winter clothing or kitchen utensils. Many psychologists claim that these types of categories differ according to their degree of coherence. However, what they mean by coherence is not always clear. In this talk I will argue that the fundamental notion for understanding conceptual coherence is that of feature-feature correlation. More precisely, I will claim that what makes a concept more or less coherent is the degree to which its features are correlated with each other. I will present a measure of coherence using the theory of conceptual spaces and I will discuss the relationship between conceptual coherence and concept utility.
  • Author: Corina Strößner
  • Title: Conceptual spaces and natural concepts: normative and descriptive aspects
  • Abstract: Conceptual spaces are a frequently applied framework of representing conceptual content. One of the central aims of using them is to determine what makes a concept natural. The notion of natural concepts seems to combine normative rational aspects (their relation to induction, optimal payoff between frugality and informativity, ontological appropriateness etc.) as well as rather descriptive ones (learned easily, lexicalized in natural language etc.). How do these normative and descriptive aspects relate to each other? To which extend is the notion of natural concepts as suggested by the conceptual spaces framework a normative one or a descriptive one? During the talk, I will first address the relation of normative and descriptive aspects in cognitive science and present Bayesianism as a paradigmatic example of a mixed normative-descriptive approach, that is an approach in which there is an interplay between empirical claims and normative ones (joint work with U. Hahn).  Second, I will compare the framework of conceptual spaces and in particular their potential criteria of natural concepts to the Bayesian approach. While there are clear differences in the subject (beliefs, inferences versus concepts) as well as in the normative justification, there are also commonalities, for instance the application of optimality arguments. This indicates that the framework of conceptual spaces and its way to capture conceptual naturalness is best viewed as a mixed normative-descriptive approach.
  • Author: Gottfried Vosgerau
  • Title: Giving the Mind a Place in Nature: Naturalness as Arising from Interaction
  • Abstract: The underlying idea of naturalness seems to be that natural properties are those not influenced by our mental makeup. For such a view, a fundamental dualism is required: that “nature” and “mind” are categorically distinct. In a monist/physicalist/naturalized framework, however, the mind is conceived of as part of nature. Consequently, a natural property – or at least a natural concept – should not rule out minds. In this talk, I want to explore the idea that a concept is natural if it contains enough of both the “world” (the non-mind parts of nature) and minds. A property that lacks the mind-part cannot be apprehended at all, so there cannot be any concept of it neither (trivially, since concepts are mind-things). A concept that lacks the world-part, however, is “arbitrary” or “gerrymandered”. How can the concepts in between the two poles be characterized? My proposal is: Natural concepts are those that reveal themselves in the interaction between minds and the world. To spell out this basic idea, I will use a naturalized notion of affordances (Tillas et al. 2016) to distinguish different kinds of concepts: Basic affordances are based on unintentional movements and provide the basis for interacting intentionally with the world. Complex affordances are based on reliable correlations between intentional actions and directly perceptible features. Within the complex affordances, the kind of source of the correlations gives rise to further distinctions: Social affordances are based on correlations that are reliable because of our social praxis (e.g. because of conventions). Physical affordances are based on causal relations between directly perceptible features and intentional actions; they basically constitute the class of properties that can be (indirectly) measured, since measurement is spelled out as an intentional action that results in directly perceptible features causally linked to the property that is being measured (e.g. placing a thermometer such that the height of the mercury shows the temperature).  The more reliable the correlation between directly perceptible features and actions is, the more natural a corresponding concept is. From a philosophy of mind perspective, a lot of concepts are natural enough to qualify as “natural for thinking”, including non-physical concepts. From a metaphysical perspective, “perfectly natural” properties are the most interesting ones. In the account presented here, these are the ones that are based on perfect correlations, which arguably can only arise from basic, direct causal relations.
  • Author: Peter Gärdenfors
  • Title: Naturalness of concepts and the economics of cognition and communication
  • Abstract: Douven and Gärdenfors (Mind and Language 2020) propose a number of criteria for which concepts can be considered to be natural. Starting from these criteria, I will discuss how they can be grounded in economic or information-theoretic principles of cognition and communication. My focus will be on the thesis that concepts are convex regions in conceptual spaces. In particular, I analyze the consequences of constraints on memory and learnability for how natural concepts should look like. I also show that a requirement of a ‘meeting of minds’ for communicative success leads to similar consequences.
  • Author: James Hampton
  • Title: Varieties of Natural Concepts
  • Abstract: Natural concepts are those that occur in human thought and language. While many appear to be constituted by similarity relations, which make them suitable for modelling in conceptual spaces, others may be differently constituted. These concepts include abstract concepts, essentialist natural kinds and logical or mathematical concepts. I discuss the different sources, uses and epistemological bases of these different forms of concept.
  • Author: Prof. Catherine Kendig
  • Title: Naturalness in the making
  • Abstract: Both empirically based investigations and analytic metaphysics provide windows through which the structure of the world can become known. What I ask, and attempt to answer, is, how does what we know depend on what we think is out there in the world and how do these conceptual commitments inform and shape our research endeavors? Many argue that the world constrains the concepts and categories scientists can use in their investigations of it. This constraint is often voiced in the sentiment that the world pushes back. I provide a frame within which to consider what it means to be in a world that pushes back, but also one that we push on, and how that exchange informs our understanding of naturalness. If we are interested in understanding natural kinds or natural partitionings, we need to attend to the activities involved in their making—the partitioning of parts and the kinding of kinds. To do this, I examine two cases in botany where conceptual commitments shaped the ontological dimensions of the world: 1) Goethe's LEAF, ROOT, STEM model; and 2) Simon Schwendener's dual theory of lichens. Lastly, I suggest, one desideratum of any metaphysics would be that it provides a way to flesh out how the world pushing back is felt by the investigator with respect to the ontology they employ.
  • Author: Sebastian Scholz
  • Title: Conceptual Spaces: A Solution to Goodman’s New Riddle of Induction?
  • Abstract: Peter Gärdenfors treats Goodman’s New Riddle of Induction as a problem of knowledge representation. He maintains natural concepts – those we employ in (inductive) reasoning – are regions in Conceptual Spaces that have certain geometric properties. While the account has its merits, I will argue that it does not offer a convincing solution to Goodman's Riddle. My main argument takes the form of a dilemma: If Gärdenfors takes his instrumentalism seriously, then his answer to Goodman's epistemological question (what is the justification of our inferential practice?) amounts to a rationalist conception that takes naturality to reside exclusively in the cognitive domain (in our conceptual spaces), such that the justification cannot reach out into the world. If, by contrast, he relaxes his instrumentalism by talking about the structure of external reality, then the urgent need for an ontological criterion of naturalness arises. Since Goodman requires a way to distinguish projectible from non-projectible predicates, it is not sufficient to posit natural entities as ontological primitives. The geometric criterion that Gärdenfors offers, however, applies to cognitive representations and prima facie cannot be applied to external entities. I will conclude by discussing the possibility and utility of adapting other criteria of naturalness.
  • Co-authors: Paul Thorn  & Gerhard Schurz
  • Title: Induction with and without Natural Predicates
  • Abstract: In 1955, Nelson Goodman formulated so called ‘gruesome’ predicates in order to present a previously unrecognized problem concerning the cogency of inductive inference. Goodman's gruesome properties are analytically dependent on the property of being included in the sample on which the inductive inference is based. Typical approaches to the problem take the form of criteria aimed at blocking intuitively undesirable inductive inferences. Rather than proposing general criteria aimed at blocking inductive inferences with gruesome predicates, this paper is based on computer simulations. We make the following observation: In the kinds of circumstances in which one reasonably takes oneself to be when one makes a cogent inductive inference, induction with non-gruesome predicates is reliable, and induction with gruesome predicates is unreliable. Our approach to Goodman’s problem is quite permissive regarding the range of predicates that may be employed in making cogent inductive inferences, allowing induction with predicates that undoubtedly count as ‘unnatural’. As a postscript to our discussion of gruesome predicates, we consider whether there is any reason to prefer some species of ‘natural’ predicate as a basis for induction. We answer in the affirmative, and explain why certain kinds of natural predicates tend to support more reliable inductive inferences in cases where the sample size is small.

The contributions will be published in a special issue of the journal Philosophia. The deadline for the first manuscript is August 2023, for the revised manuscripts end of 2023.