Towards a computational (neuro)phenomenology of mental action: modelling meta-awareness and attentional control with deep-parametric active inference

Abstract

Metacognition refers to the capacity to access, monitor, and control aspects of one’s mental operations and is central to the human condition and experience. Disorders of metacognition are a hallmark of many psychiatric conditions and the training of metacognitive skills is central in education and in many psychotherapies. This paper provides first steps towards the development of a formal neurophenomenology of metacognition. To do so, we leverage the tools of the active inference framework, extending a previous computational model of implicit metacognition by adding a hierarchical level to model explicit (conscious) meta-awareness and the voluntary control of attention through covert action. Using the example of mind-wandering and its regulation in focused attention, we provide a computational proof of principle for an inferential architecture apt to enable the emergence of central components of metacognition: namely, the ability to access, monitor, and control cognitive states.

Predictive processing as a systematic basis for identifying the neural correlates of consciousness

“…There has been a significant amount of research mapping a large set of properties of PP to many different aspects of conscious phenomenology (Clark, 2015;Hohwy, 2013Hohwy, , 2020aSeth, 2019). Existing theories of consciousness are being interpreted in the light of PP, such as Heterophenomenology (Dołęga & Dewhurst, 2020), Global neuronal workspace theory (Hohwy, 2013;Whyte, 2019;Whyte & Smith, 2020), versions of Higher-order thought theory and Metacognitive theories (Hohwy, 2015;Sandved Smith et al, 2020), Attention-based theories (Marchi & Hohwy, 2020), and Integrated information theory (for related discussion, see Kanai et al, 2019).…”

“…That is, the PP system is constantly 'looking down' at its own representations seeking to learn how their precisions (and more generally their sufficient statistics) change over time. Many questions are still open about how this general PP element can speak to the role of different kinds of metacognition in consciousness (for deployment of active inference to metacognition, see Sandved Smith et al, 2020). An outstanding challenge here will be to formally integrate the framework of signal detection theory, in which most theories of metacognition are articulated (e.g., Barrett et al, 2013) with PP (see, e.g., FitzGerald et al, 2014Fleming, 2020).…”

Distinguishing absence of awareness from awareness of absence

“…1, panel c). In more elaborate Bayesian models, meta-inference about current attentional states not only affects beliefs about world states, but also drives the active deployment of endogeneous attention (Sandved Smith et al, 2020). Common to these higher-order models of awareness is a hierarchical structure of beliefs, with higher-level beliefs about content-invariant aspects of awareness (for example, beliefs about attentional state or expected firing rate following a lesion) informing lower-level inference about external world states (for example, the presence or absence of a stimulus).…”

From generative models to generative passages: A computational approach to (neuro)phenomenology

“…The priors that one can add to the architecture of the generative model or process may offer the possibility of mapping a variety of claims coming from different traditions. For instance, one might argue that the generative process (that generates the data to be explained, in our case hyletic data) might be the focus of a formalized Merleau-Pontian phenomenology, since the latter emphasizes the 'invisible' underbelly of a situated, worldbound consciousness and its roots in the material body, which together lend structure to conscious experience (Merleau-Ponty 1968); or might argue that Husserlian phenomenology's emphasis on self-grasping in phenomenological self-experience lends itself well to hierarchical modelling (Husserl 1991(Husserl , [b] 2012Sandved-Smith et al 2020;Hesp et al 2021).…”

“…In this paper, we pursue the naturalization of phenomenology using modelling techniques developed in computational neuroscience and biology. Specifically, we propose a new version of neurophenomenology as 'generative passages' (Lutz 2002;Roy et al 1999;Varela 1997), based on methodological innovations that have emerged recently under the rubric of generative modelling (Friston 2019;Hesp et al 2021;Sandved-Smith et al 2020). We call this approach computational phenomenology because it applies methods from computational modelling to phenomenology.…”