Hajo Greif

The aim of this paper is to grasp the relevant distinctions between various ways in which models and simulations in Artificial Intelligence (AI) relate to cognitive phenomena. In order to get a systematic picture, a taxonomy is developed that is based on the coordinates of formal versus material analogies and theory-guided versus pre-theoretic models in science. These distinctions have parallels in the computational versus mimetic aspects and in analytic versus exploratory types of computer simulation. The proposed taxonomy cuts across the traditional dichotomies between symbolic and embodied AI, general intelligence and symbol and intelligence and cognitive simulation and human/non-human-like AI. According to the taxonomy proposed here, one can distinguish between four distinct general approaches that figured prominently in early and classical AI, and that have partly developed into distinct research programs: first, phenomenal simulations (e.g., Turing’s “imitation game”); second, simulations that explore general-level formal isomorphisms in pursuit of a general theory of intelligence (e.g., logic-based AI); third, simulations as exploratory material models that serve to develop theoretical accounts of cognitive processes (e.g., Marr’s stages of visual processing and classical connectionism); and fourth, simulations as strictly formal models of a theory of computation that postulates cognitive processes to be isomorphic with computational processes (strong symbolic AI). In continuation of pragmatic views of the modes of modeling and simulating world affairs, this taxonomy of approaches to modeling in AI helps to elucidate how available computational concepts and simulational resources contribute to the modes of representation and theory development in AI research—and what made that research program uniquely dependent on them.

Alan M. Turing’s last published work and some posthumously published manuscripts were dedicated to the development of his theory of organic pattern formation. In “The Chemical Basis of Morphogenesis” (1952), he provided an elaborated mathematical formulation of the theory of the origins of biological form that had been first proposed by Sir D’Arcy Wendworth Thompson in On Growth and Form (1917/1942). While arguably his most mathematically detailed and his systematically most ambitious effort, Turing’s morphogenetical writings also form the most thematically self-contained and least philosophically explored part of his work. We dedicate our inquiry to the reasons and the implications of Turing’s choice of biological topic and viewpoint. We will probe for possible factors in Turing’s choice that go beyond availability and acquaintance with On Growth and Form. On these grounds, we will explore how and to what extent his theory of morphogenesis actually ties in with his concept of mechanistic computation. Notably, Thompson’s pioneering work in biological ‘structuralism’ was organicist in outlook and explicitly critical of the Darwinian approaches that were popular with Turing’s cyberneticist contemporaries—and partly used by Turing himself in his proto-connectionist models of learning. Resolving this apparent dichotomy, we demonstrate how Turing’s quest for mechanistic explanations of how organisation emerges in nature leaves room for a non-mechanist view of nature.

Concerns over epistemic opacity abound in contemporary debates on Artificial Intelligence (AI). However, it is not always clear to what extent these concerns refer to the same set of problems. We can observe, first, that the terms 'transparency' and 'opacity' are used either in reference to the computational elements of an AI model or to the models to which they pertain. Second, opacity and transparency might either be understood to refer to the properties of AI systems or to the epistemic situation of human agents with respect to these systems. While these diagnoses are independently discussed in the literature, juxtaposing them and exploring possible interrelations will help to get a view of the relevant distinctions between conceptions of opacity and their empirical bearing. In pursuit of this aim, two pertinent conditions affecting computer models in general and contemporary AI in particular are outlined and discussed: opacity as a problem of computational tractability and opacity as a problem of the universality of the computational method.

The problem of epistemic opacity in Artificial Intelligence is often characterised as a problem of intransparent algorithms that give rise to intransparent models. However, the degrees of transparency of an AI model should not be taken as an absolute measure of the properties of its algorithms but of the model’s degree of intelligibility to human users. Its epistemically relevant elements are to be specified on various levels above and beyond the computational one. In order to elucidate this claim, I first contrast computer models and their claims to algorithm-based universality with cybernetics-style analogue models and their claims to structural isomorphism between elements of model and target system. While analogue models aim at perceptually or conceptually accessible model-target relations, computer models give rise to a specific kind of underdetermination in these relations that needs to be addressed in specific ways. I then undertake a comparison between two contemporary AI approaches that, although related, distinctly align with the above modelling paradigms and represent distinct strategies towards model intelligibility: Deep Neural Networks and Predictive Processing. I conclude that their respective degrees of epistemic transparency primarily depend on the underlying purposes of modelling, not on their computational properties.

“Teleosemantic” or “biosemantic” theories form a strong naturalistic programme in the philosophy of mind and language. They seek to explain the nature of mind and language by recourse to a natural history of “proper functions” as selected-for effects of language- and thought-producing mechanisms. However, they remain vague with respect to the nature of the proposed analogy between selected-for effects on the biological level and phenomena that are not strictly biological, such as reproducible linguistic and cultural forms. This essay critically explores various interpretations of this analogy. It suggests that these interpretations can be explicated by contrasting adaptationist with pluralist readings of the evolutionary concept of adaptation. Among the possible interpretations of the relations between biological adaptations and their analogues in language and culture, the two most relevant are a linear, hierarchical, signalling-based model that takes its cues from the evolution of co-operation and joint intentionality and a mutualistic, pluralist model that takes its cues from mimesis and symbolism in the evolution of human communication. Arguing for the merits of the mutualistic model, the present analysis indicates a path towards an evolutionary pluralist version of biosemantics that will align with theories of cognition as being environmentally “scaffolded”. Language and other cultural forms are partly independent reproducible structures that acquire proper functions of their own while being integrated with organism-based cognitive traits in co-evolutionary fashion.

Paleontological evidence suggests that human artefacts with intentional markings might have originated already in the Lower Paleolithic, up to 500.000 years ago and well before the advent of ‘behavioural modernity’. These markings apparently did not serve instrumental, tool-like functions, nor do they appear to be forms of figurative art. Instead, they display abstract geometric patterns that potentially testify to an emerging ability of symbol use. In a variation on Ian Hacking’s speculative account of the possible role of “likeness-making” in the evolution of human cognition and language, this essay explores the central role that the embodied processes of making and the collective practices of using such artefacts might have played in early human cognitive evolution. Two paradigmatic findings of Lower Paleolithic artefacts are discussed as tentative evidence of likenesses acting as material scaffolds in the emergence of symbolic reference-making. They might provide the link between basic abilities of mimesis and imitation and the development of modern language and thought.

The aim of this paper is to grasp the relevant distinctions between various ways in which models and simulations in Artificial Intelligence (AI) relate to cognitive phenomena. In order to get a systematic picture, a taxonomy is developed that is based on the coordinates of formal versus material analogies and theory-guided versus pre-theoretic models in science. These distinctions have parallels in the computational versus mimetic aspects and in analytic versus exploratory types of computer simulation. The proposed taxonomy cuts across the traditional dichotomies between symbolic and embodied AI, general intelligence and symbol and intelligence and cognitive simulation and human/non-human-like AI. According to the taxonomy proposed here, one can distinguish between four distinct general approaches that figured prominently in early and classical AI, and that have partly developed into distinct research programs: first, phenomenal simulations (e.g., Turing’s “imitation game”); second, simulations that explore general-level formal isomorphisms in pursuit of a general theory of intelligence (e.g., logic-based AI); third, simulations as exploratory material models that serve to develop theoretical accounts of cognitive processes (e.g., Marr’s stages of visual processing and classical connectionism); and fourth, simulations as strictly formal models of a theory of computation that postulates cognitive processes to be isomorphic with computational processes (strong symbolic AI). In continuation of pragmatic views of the modes of modeling and simulating world affairs, this taxonomy of approaches to modeling in AI helps to elucidate how available computational concepts and simulational resources contribute to the modes of representation and theory development in AI research—and what made that research program uniquely dependent on them.

Given the personal acquaintance between Alan M. Turing and W. Ross Ashby and the partial proximity of their research fields, a comparative view of Turing’s and Ashby’s work on modelling “the action of the brain” (letter from Turing to Ashby, 1946) will help to shed light on the seemingly strict symbolic/embodied dichotomy: While it is clear that Turing was committed to formal, computational and Ashby to material, analogue methods of modelling, there is no straightforward mapping of these approaches onto symbol-based AI and embodiment-centered views respectively. Instead, it will be demonstrated that both approaches, starting from a formal core, were at least partly concerned with biological and embodied phenomena, albeit in revealingly distinct ways.

There are two related points at which J.J. Gibson’s ecological theory of visual perception remains remarkably underspecified: Firstly, the notion of information for perception is not explicated in much detail beyond the claim that it “specifies” the environment for perception, and, thus being an objective affair, enables an organism to perceive action possibilities or “affordances.” Secondly, misperceptions of affordances and perceptual illusions are not clearly distinguished from each other. Although the first claim seems to suggest that any perceptual illusion amounts to the misperception of affordances, there might be some relevant differences between various ways of getting things wrong. In this essay, Gibson’s notion of “specifying” information shall be reconstructed along the lines of Fred Dretske’s relational theory of information. This refined notion of information for perception will then be used to carve out the distinction between perceptual illusions and the misperception of affordances, with some help from the “Empirical Strategy” (developed by Purves et al.). It will be maintained that there are cases where perceptual illusions actually help an organism to correctly perceive an affordance. In such cases, the prima facie misrendered informational relations involved are kept intact by a set of appropriate transformation rules. Two of Gibson’s intuitions shall thus be preserved: the objectivity of informational relations and the empowerment of the organism as an active perceiver who uses those objective relations to his specific ends.

What is the role of the environment, and of the information it provides, in cognition? More specifically, may there be a role for certain artefacts to play in this context? These are questions that motivate "4E" theories of cognition (as being embodied, embedded, extended, enactive). In his take on that family of views, Hajo Greif first defends and refines a concept of information as primarily natural, environmentally embedded in character, which had been eclipsed by information-processing views of cognition. He continues with an inquiry into the cognitive bearing of some artefacts that are sometimes referred to as 'intelligent environments'. Without necessarily having much to do with Artificial Intelligence, such artefacts may ultimately modify our informational environments. With respect to human cognition, the most notable effect of digital computers is not that they might be able, or become able, to think but that they alter the way we perceive, think and act.

Two aspects of cognitive coupling, as brought forward in the Extended Mind Hypothesis, are discussed in this paper: how shall the functional coupling between the organism and some entity in his environment be spelled out in detail? What are the paradigmatic external entities to enter into that coupling? These two related questions are best answered in the light of an aetiological variety of functionalist argument that adds historical depth to the “active externalism” promoted by Clark and Chalmers and helps to counter some of the core criticisms levelled against this view. Under additional reference to conceptual parallels between the Extended Mind Hypothesis and a set of heterodox theories in biology—environmental constructivism, niche construction, developmental systems theory—an argument for the grounding of environmentally extended cognitive traits in evolved biological functions is developed. In a spirit that seeks to integrate extended functionalism with views from cognitive integration and complementarity, it is argued that instances of environmental coupling should be understood as being constitutive to cognitive functions in either of two distinct ways. It is further argued that the historically and systematically prior environmental counterparts in that coupling are features of the natural environment. Language and linguistically imbued artefacts are likely to have descended from more basic relations that have an extension over the environment

This paper commences from the critical observation that the Turing Test (TT) might not be best read as providing a definition or a genuine test of intelligence by proxy of a simulation of conversational behaviour. Firstly, the idea of a machine producing likenesses of this kind served a different purpose in Turing, namely providing a demonstrative simulation to elucidate the force and scope of his computational method, whose primary theoretical import lies within the realm of mathematics rather than cognitive modelling. Secondly, it is argued that a certain bias in Turing’s computational reasoning towards formalism and methodological individualism contributed to systematically unwarranted interpretations of the role of the TT as a simulation of cognitive processes. On the basis of the conceptual distinction in biology between structural homology vs. functional analogy, a view towards alternate versions of the TT is presented that could function as investigative simulations into the emergence of communicative patterns oriented towards shared goals. Unlike the original TT, the purpose of these alternate versions would be co-ordinative rather than deceptive. On this level, genuine functional analogies between human and machine behaviour could arise in quasi-evolutionary fashion.

There have been attempts to subsume Charles Darwin's theory of evolution under either one of two distinct intellectual traditions: early Victorian natural science and its descendants in political economy (as exemplified by Herschel, Lyell, or Malthus) and the romantic approach to art and science emanating from Germany (as exemplified by Humboldt and Goethe). In this paper, it will be shown how these traditions may have jointly contributed to the design of Darwin's theory. The hypothesis is that their encounter created a particular tension in the conception of his theory which first opened up its characteristic field and mode of explanation. On the one hand, the domain of the explanandum was conceived of under a holistic and aesthetic view of nature that, in its combination with refined techniques of observation, was deeply indebted to Humboldt in particular. On the other hand, Darwin fashioned explanations for natural phenomena, so conceived, in order to identify their proper causes in a Herschelian spirit. The particular interaction between these two traditions in Darwin, it is concluded, paved the way for a transfer of the idea of causal laws to animate nature while salvaging the romantic idea of a complex, teleological and harmonious order of nature. Copyright © 2015 Elsevier Ltd. All rights reserved.

Two popular, yet highly controversial concepts of non-human agency from two different fields of knowledge are compared in this essay: the theory of the Selfish Gene, introduced into neo-Darwinian evolutionary biology by Richard Dawkins, and Actor-Network Theory, as brought forward in Science & Technology Studies by Bruno Latour. It is argued that the two theories, despite all apparent differences, share key motifs and motivations when they try to forward knowledge in their respective fields by adopting a vocabulary that aims at metaphorical redescriptions of the origins of intentionality – in both senses: as purposefulness of action and as meaning.