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. (shrink)

-/- 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. (shrink)

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. (shrink)

When engaging with magic, we are moved by seemingly impossible events that contradict what we believe to be possible in the real world. We are surprised, curious, and baffled when we cannot explain how the magic we are witnessing is possible. We generally understand the events to be illusions. But how is it possible to be moved by something we know to be unreal? This problem is related to the paradox of fiction in aesthetics. Here, we introduce the problem in (...) the domain of theatrical magic, discuss possible solutions, and present a tentative account that allows us to accommodate inconsistent, seemingly incompatible beliefs at different representational levels in the viewers’ mind. (shrink)

In recent years, scientists have increasingly taken to investigate the predictive nature of cognition. We argue that prediction relies on abstraction, and thus theories of predictive cognition need an explicit theory of abstract representation. We propose such a theory of the abstract representational capacities that allow humans to transcend the “here-and-now.” Consistent with the predictive cognition literature, we suggest that the representational substrates of the mind are built as ahierarchy, ranging from the concrete to the abstract; however, we argue that (...) there are qualitative differences between elements along this hierarchy, generating meaningful, often unacknowledged,diversity.Echoing views from philosophy, we suggest that the representational hierarchy can be parsed into:modality-specificrepresentations, instantiated on perceptual similarity;multimodalrepresentations, instantiated primarily on the discovery of spatiotemporal contiguity; andcategoricalrepresentations, instantiated primarily on social interaction. These elements serve as the building blocks ofcomplex structures discussed in cognitive psychology (e.g., episodes, scripts) and are the inputs for mental representations that behave like functions, typically discussed in linguistics (i.e.,predicators). We support our argument for representational diversity by explaining how the elements in our ontology are all required to account for humans’ predictive cognition (e.g., in subserving logic-based prediction; in optimizing the trade-off between accurate and detailed predictions) and by examining how the neuroscientific evidence coheres with our account. In doing so, we provide a testable model of the neural bases of conceptual cognition and highlight several important implications to research on self-projection, reinforcement learning, and predictive-processing models of psychopathology. (shrink)

The commentaries address our view of abstraction, our ontology of abstract entities, and our account of predictive cognition as relying on relatively concrete simulation or relatively abstract theory-based inference. These responses revisit classic questions concerning mental representation and abstraction in the context of current models of predictive cognition. The counter arguments to our article echo: constructivist theories of knowledge, “neat” approaches in artificial intelligence and decision theory, neo-empiricist models of concepts, and externalist views of cognition. We offer several empirical predictions (...) that address points of contention and that highlight the generative potential of our model. (shrink)

Predictive processing accounts of perception assume that perception does not work in a purely bottom-up fashion but also uses acquired knowledge to make top-down predictions about the incoming sensory signals. This provides a challenge for foundationalist accounts of perception according to which perceptual beliefs are epistemically basic, that is, epistemically independent from other beliefs. If prior beliefs rationally influence which perceptual beliefs we come to accept, then foundationalism about perception appears untenable. I review several ways in which foundationalism might be (...) reconciled with PP from both an internalist and externalist perspective, and argue that an externalist foundationalism provides the best match with PP. (shrink)

A flood of new multidisciplinary work on the causes of depersonalization disorder provides a new way to think about the feeling that experiences “belong” to the self. In this paper I argue that this feeling, baptized “mineness” or “subjective presence” : 565–573, 2013) emerges from a multilevel interaction between emotional, affective and cognitive processing. The “self” to which experience is attributed is a predictive model made by the mind to explain the modulation of affect as the organism progresses through the (...) world. When the world no longer produces predicted affect the organism needs to explain this unpredicted absence of feeling. It is important to this account that cognition and perception are otherwise intact. Consequently the mind’s representation of the world and its emotionally salient properties are unchanged, leading the mind to predict a characteristic affective response. When that prediction is not fulfilled the organisms feels as if she is no longer present in experience. This is reported at the as feeling of depersonalization. (shrink)

Psychological research on emotion perception anchors heavily on an object perception analogy. We present static “cues,” such as facial expressions, as objects for perceivers to categorize. Yet in the real world, emotions play out as dynamic multidimensional events. Current theoretical approaches and research methods are limited in their ability to capture this complexity. We draw on insights from a predictive coding account of neural activity and a grounded cognition account of concept representation to conceive of emotion perception as a stream (...) of synchronized conceptualizations between two individuals, which is supported and shaped by language. We articulate how this framework can illuminate the fundamental need to study culture, as well as other sources of conceptual variation, in unpacking conceptual synchrony in emotion. We close by suggesting that the conceptual system provides the necessary flexibility to overcome gaps in emotional synchrony. (shrink)

Depersonalization and derealization refer to states of dissociation in which one feels a sense of alienation in relation to one’s self and environment, respectively. Whilst transient episodes often diminish without treatment, chronic experiences of DP and DR may last for years, with common treatments lacking a strong evidence base for their efficacy. We propose a theoretical explanation of DP and DR based on interoceptive predictive coding, and discuss how transient experiences of DP and DR may be induced in the non-clinical (...) population using virtual reality. Further, we review the use of heartbeat evoked potentials in detecting the neural correlates of DP and DR allowing for an objective measure of these experiences in the non-clinical population. Finally, we discuss how the induction and detection of transient experiences of DP and DR in the non-clinical population could shed light on how the brain constructs one’s sense of self and reality. (shrink)

A number of perceptual (exteroceptive and proprioceptive) illusions present problems for predictive processing accounts. In this chapter we’ll review explanations of the Müller-Lyer Illusion (MLI), the Rubber Hand Illusion (RHI) and the Alien Hand Illusion (AHI) based on the idea of Prediction Error Minimization (PEM), and show why they fail. In spite of the relatively open communicative processes which, on many accounts, are posited between hierarchical levels of the cognitive system in order to facilitate the minimization of prediction errors, perceptual (...) illusions seemingly allow prediction errors to rule. Even if, at the top, we have reliable and secure knowledge that the lines in the MLI are equal, or that the rubber hand in the RHI is not our hand, the system seems unable to correct for sensory errors that form the illusion. We argue that the standard PEM explanation based on a short-circuiting principle doesn’t work. This is the idea that where there are general statistical regularities in the environment there is a kind of short circuiting such that relevant priors are relegated to lower-level processing so that information from higher levels is not exchanged (Ogilvie and Carruthers, Review of Philosophy and Psychology 7:721–742, 2016), or is not as precise as it should be (Hohwy, The Predictive Mind, Oxford University Press, Oxford, 2013). Such solutions (without convincing explanation) violate the idea of open communication and/or they over-discount the reliable and secure knowledge that is in the system. We propose an alternative, 4E (embodied, embedded, extended, enactive) solution. We argue that PEM fails to take into account the ‘structural resistance’ introduced by material and cultural factors in the broader cognitive system. (shrink)

The full scope of enactivist approaches to cognition includes not only a focus on sensory-motor contingencies and physical affordances for action, but also an emphasis on affective factors of embodiment and intersubjective affordances for social interaction. This strong conception of embodied cognition calls for a new way to think about the role of the brain in the larger system of brain-body-environment. We ask whether recent work on predictive coding offers a way to think about brain function in an enactive system, (...) and we suggest that a positive answer is possible if we interpret predictive coding in a more enactive way, i.e., as involved in the organism’s dynamic adjustments to its environment. (shrink)

We distinguish between three philosophical views on the neuroscience of predictive models: predictive coding, predictive processing and predictive engagement. We examine the concept of active inference under each model and then ask how this concept informs discussions of social cognition. In this context we consider Frith and Friston’s proposal for a neural hermeneutics, and we explore the alternative model of enactivist hermeneutics.

This paper centers around the notion that internal, mental representations are grounded in structural similarity, i.e., that they are so-called S-representations. We show how S-representations may be causally relevant and argue that they are distinct from mere detectors. First, using the neomechanist theory of explanation and the interventionist account of causal relevance, we provide a precise interpretation of the claim that in S-representations, structural similarity serves as a “fuel of success”, i.e., a relation that is exploitable for the representation using (...) system. Then, we discuss crucial differences between S-representations and indicators or detectors, showing that—contrary to claims made in the literature—there is an important theoretical distinction to be drawn between the two. (shrink)

According to the predictive coding theory of cognition , brains are predictive machines that use perception and action to minimize prediction error, i.e. the discrepancy between bottom–up, externally-generated sensory signals and top–down, internally-generated sensory predictions. Many consider PCT to have an explanatory scope that is unparalleled in contemporary cognitive science and see in it a framework that could potentially provide us with a unified account of cognition. It is also commonly assumed that PCT is a representational theory of sorts, in (...) the sense that it postulates that our cognitive contact with the world is mediated by internal representations. However, the exact sense in which PCT is representational remains unclear; neither is it clear that it deserves such status—that is, whether it really invokes structures that are truly and nontrivially representational in nature. In the present article, I argue that the representational pretensions of PCT are completely justified. This is because the theory postulates cognitive structures—namely action-guiding, detachable, structural models that afford representational error detection—that play genuinely representational functions within the cognitive system. (shrink)

In this paper, I use the predictive processing theory of perception to tackle the question of how perceptual states can be rationally involved in cognition by justifying other mental states. I put forward two claims regarding the epistemological implications of PP. First, perceptual states can confer justification on other mental states because the perceptual states are themselves rationally acquired. Second, despite being inferentially justified rather than epistemically basic, perceptual states can still be epistemically responsive to the mind-independent world. My main (...) goal is to elucidate the epistemology of perception already implicit in PP. But I also hope to show how it is possible to peacefully combine central tenets of foundationalist and coherentist accounts of the rational powers of perception while avoiding the well-recognized pitfalls of either. (shrink)

Despite the fact that the notion of internal representation has - at least according to some - a fundamental role to play in the sciences of the mind, not only has its explanatory utility been under attack for a while now, but it also remains unclear what criteria should an explanation of a given cognitive phenomenon meet to count as a representational explanation in the first place. The aim of this article is to propose a solution to this latter problem. (...) I will assume that representational explanations should be construed as a form of mechanistic explanations and proceed by proposing a general sketch of a functional architecture of a representational cognitive mechanism. According to the view on offer here, representational mechanisms are mechanisms that meet four conditions: the structural resemblance condition, the action-guidance condition, the decouplability condition, and the error-detection condition. (shrink)

I contrast two notions of cognition offered in Joseph LeDoux’s (2019) The Deep History of Ourselves toward arguing for the functional role of affect and consciousness in the evolution of matter. I argue that an emphasis on the cultural construction of emotions misrepresents the relationship between culture and biology. A more parsimonious story about the evolution of mind requires leaving behind some aspects of a cognitivist epistemology.

A key component of research on human sentence processing is to characterize the processing difficulty associated with the comprehension of words in context. Models that explain and predict this difficulty can be broadly divided into two kinds, expectation‐based and memory‐based. In this work, we present a new model of incremental sentence processing difficulty that unifies and extends key features of both kinds of models. Our model, lossy‐context surprisal, holds that the processing difficulty at a word in context is proportional to (...) the surprisal of the word given a lossy memory representation of the context—that is, a memory representation that does not contain complete information about previous words. We show that this model provides an intuitive explanation for an outstanding puzzle involving interactions of memory and expectations: language‐dependent structural forgetting, where the effects of memory on sentence processing appear to be moderated by language statistics. Furthermore, we demonstrate that dependency locality effects, a signature prediction of memory‐based theories, can be derived from lossy‐context surprisal as a special case of a novel, more general principle called information locality. (shrink)

Gilead et al. raise some fascinating issues about representational substrates and structures in the predictive brain. This commentary drills down on a core theme in their arguments; namely, the structure of models that generate predictions. In particular, it highlights their factorial nature – both in terms of deep hierarchies over levels of abstraction and, crucially, time – and how this underwrites agency.

There is a steadily growing literature on the role of the immune system in psychiatric disorders. So far, these advances have largely taken the form of correlations between specific aspects of inflammation (e.g. blood plasma levels of inflammatory markers, genetic mutations in immune pathways, viral or bacterial infection) with the development of neuropsychiatric conditions such as autism, bipolar disorder, schizophrenia and depression. A fundamental question remains open: why are psychiatric disorders and immune responses intertwined? To address this would require a (...) step back from a historical mind–body dualism that has created such a dichotomy. We propose three contributions of active inference when addressing this question: translation, unification, and simulation. To illustrate these contributions, we consider the following questions. Is there an immunological analogue of sensory attenuation? Is there a common generative model that the brain and immune system jointly optimise? Can the immune response and psychiatric illness both be explained in terms of self-organising systems responding to threatening stimuli in their external environment, whether those stimuli happen to be pathogens, predators, or people? Does false inference at an immunological level alter the message passing at a psychological level (or vice versa) through a principled exchange between the two systems? (shrink)

This comment argues that appraisals result from the almost incessant predictive activity of the awake brain, using its incoming and stored information. It explores how appraisal operates when it does. It argues that the analysis of the processes that lead to appraisals currently is still at an early stage.

How are new Bayesian hypotheses generated within the framework of predictive processing? This explanatory framework purports to provide a unified, systematic explanation of cognition by appealing to Bayes rule and hierarchical Bayesian machinery alone. Given that the generation of new hypotheses is fundamental to Bayesian inference, the predictive processing framework faces an important challenge in this regard. By examining several cognitive‐level and neurobiological architecture‐inspired models of hypothesis generation, we argue that there is an essential difference between the two types of (...) models. Cognitive‐level models do not specify how they can be implemented in brains and include structures and assumptions that are external to the predictive processing framework. By contrast, neurobiological architecture‐inspired models, which aim to better resemble brain processes, fail to explain important capacities of cognition, such as categorization and few‐shot learning. The “scaling‐up” challenge for proponents of predictive processing is to explain the relationship between these two types of models using only the theoretical and conceptual machinery of Bayesian inference. (shrink)

The idea that the brain is an information processing system raises some challenging questions about whether information exists independently of brains. Answering these questions is relevant for clarifying the theoretical foundations of the sciences of mind and brain, but also for appropriately interpreting and evaluating the evidence about how brains—and other biological systems—work. This article claims that (1) informational descriptions in the sciences of mind and brain can be genuinely explanatory, despite assuming a mind-dependent notion of information; and (2) that (...) either Popperian objectivity or response-dependence in normal receivers is sufficient for grounding the explanatory role that information often plays in these sciences. Hierarchical predictive processing, which has become a central theoretical framework for neurocognitive research, is used as a case study for supporting these two claims. (shrink)

First, I discuss cross-cultural evidence showing that a good deal of enculturation takes place outside of thinking through other minds. Second, I review evidence challenging the claim that humans seek to minimize entropy. Finally, I argue that optimality claims should be avoided, and that descriptive Bayesianism offers a more promising avenue for the development of a Bayesian theory of culture.

Intuitively, the accuracy of initial word-referent mappings should be positively correlated with the outcome of learning. Yet recent evidence suggests an inverse effect of initial accuracy in adults, whereby greater accuracy of initial mappings is associated with poorer outcomes in a cross-situational learning task. Here, we examine the impact of initial accuracy on 4-year-olds, 10-year-olds, and adults. For half of the participants most word-referent mappings were initially correct and for the other half most mappings were initially incorrect. Initial accuracy was (...) positively related to learning outcomes in 4-year-olds, had no effect on 10-year-olds' learning, and was inversely related to learning outcomes in adults. Examination of item learning patterns revealed item interdependence for adults and 4-year-olds but not 10-year-olds. These findings point to a qualitative change in language learning processes over development. (shrink)

If there are two dependent positive real variables x1 and x2, and only x1 is known, what is the probability that x2 is larger versus smaller than x1? There is no uniquely correct answer according to “frequentist” and “subjective Bayesian” definitions of probability. Here we derive the answer given the “objective Bayesian” definition developed by Jeffreys, Cox, and Jaynes. We declare the standard distance metric in one dimension, d(A,B)≡|A−B|, and the uniform prior distribution, as axioms. If neither variable is known, (...) P(x2x1)=P(x2>x1). This appears obvious, since the state spaces x2x1 and x2>x1 have equal size. However, if x1 is known and x2 unknown, there are infinitely more numbers in the space x2>x1 than x2x1. Despite this asymmetry, we prove P(x2x1∣x1)=P(x2>x1∣x1), so that x1 is the median of p(x2|x1), and x1 is statistically independent of ratio x2/x1. We present three proofs that apply to all members of a set of distributions. Each member is distinguished by the form of dependence between variables implicit within a statistical model (gamma, Gaussian, etc.), but all exhibit two symmetries in the joint distribution p(x1,x2) that are required in the absence of prior information: exchangeability of variables, and non-informative priors over the marginal distributions p(x1) and p(x2). We relate our conclusion to physical models of prediction and intelligence, where the known ’sample’ could be the present internal energy within a sensor, and the unknown the energy in its external sensory cause or future motor effect. (shrink)

This paper argues that the still-emerging paradigm of situated cognition requires a more systematic perspective on media to capture the enculturation of the human mind. By virtue of being media, cultural artifacts present central experiential models of the world for our embodied minds to latch onto. The paper identifies references to external media within embodied, extended, enactive, and predictive approaches to cognition, which remain underdeveloped in terms of the profound impact that media have on our mind. To grasp this impact, (...) I propose an enactive account of media that is based on expansive habits as media-structured, embodied ways of bringing forth meaning and new domains of values. We apply such habits, for instance, when seeing a picture or perceiving a movie. They become established through a process of reciprocal adaptation between media artifacts and organisms and define the range of viable actions within such a media ecology. Within an artifactual habit, we then become attuned to a specific media work that engages us. Both the plurality of habits and the dynamical adjustments within a habit require a more flexible neural architecture than is addressed by classical cognitive neuroscience. To detail how neural and media processes interlock, I will introduce the concept of neuromediality and discuss radical predictive processing accounts that could contribute to the externalization of the mind by treating media themselves as generative models of the world. After a short primer on general media theory, I discuss media examples in three domains: pictures and moving images; digital media; architecture and the built environment. This discussion demonstrates the need for a new cognitive media theory based on enactive artifactual habits—one that will help us gain perspective on the continuous re-mediation of our mind. (shrink)

What is the content of a mental state? This question poses the problem of intentionality: to explain how mental states can be about other things, where being about them is understood as representing them. A framework that integrates predictive coding and signaling systems theories of cognitive processing offers a new perspective on intentionality. On this view, at least some mental states are evaluations, which differ in function, operation, and normativity from representations. A complete naturalistic theory of intentionality must account for (...) both types of intentional state. (shrink)

When the history of life on earth is viewed as a history of cell division, all of life becomes a single cell lineage. The growth and differentiation of this lineage in reciprocal interaction with its environment can be viewed as a developmental process; hence the evolution of life on earth can also be seen as the development of life on earth. Here, in reviewing this field, some potentially fruitful research directions suggested by this change in perspective are highlighted. Variation and (...) selection become, for example, bidirectional information flows between scales, while the notions of “cooperation” and “competition” become scale relative. The language of communication, inference, and information processing becomes more useful than the language of causation to describe the interactions of both homogeneous and heterogeneous living systems at any scale. Emerging scale‐free theoretical frameworks such as predictive coding and active inference provide conceptual tools for reconceptualizing biology as the study of a unified, multiscale dynamical system. (shrink)

Meaning has traditionally been regarded as a problem for philosophers and psychologists. Advances in cognitive science since the early 1960s, however, broadened discussions of meaning, or more technically, the semantics of perceptions, representations, and/or actions, into biology and computer science. Here, we review the notion of “meaning” as it applies to living systems, and argue that the question of how living systems create meaning unifies the biological and cognitive sciences across both organizational and temporal scales.

How do human infants learn the causal dependencies between events? Evidence suggests that this remarkable feat can be achieved by observation of only a handful of examples. Many computational models have been produced to explain how infants perform causal inference without explicit teaching about statistics or the scientific method. Here, we propose a spiking neuronal network implementation that can be entrained to form a dynamical model of the temporal and causal relationships between events that it observes. The network uses spike-time (...) dependent plasticity, long-term depression, and heterosynaptic competition rules to implement Rescorla–Wagner-like learning. Transmission delays between neurons allow the network to learn a forward model of the temporal relationships between events. Within this framework, biologically realistic synaptic plasticity rules account for well-known behavioral data regarding cognitive causal assumptions such as backwards blocking and screening-off. These models can then be run as emulators for state inference. Furthermore, this mechanism is capable of copying synaptic connectivity patterns between neuronal networks by observing the spontaneous spike activity from the neuronal circuit that is to be copied, and it thereby provides a powerful method for transmission of circuit functionality between brain regions. (shrink)