Towards Characterizing the Canonical Computations Generating Phenomenal Experience

Few people tackle the neural or computational basis of qualitative experience (Frith, 2019). Why? One major reason is that science and philosophy have both struggled to propose how we might even begin to start studying it. Here I propose that metacognitive computations, and the subjective feelings that go along with them, give us a solid starting point. Specifically, perceptual metacognition possesses unique properties that provide a powerful and unique opportunity for studying the neural and computational correlates of subjective experience, falling into three categories: (1) Metacognition is subjective: there is something it is like to feel ‘confident’; (2) Metacognitive processes are objectively characterizable: We can objectively observe metacognitive reports and define computational models to fit to empirical data; (3) Metacognition has multiple hierarchically-dependent “anchors”, presenting a unique computational opportunity for developing sensitive, specific models. I define this Metacognition as a Step Toward Explaining Phenomenology (M-STEP) approach to state that, given these properties, computational models of metacognition represent an empirically-tractable early step in identifying the generative process that constructs qualitative experience. By applying decades of developments in computational cognitive science and formal computational model comparisons to the specific properties of perceptual metacognition, we may reveal new and exciting insights about how the brain constructs subjective conscious experiences and the nature of those experiences themselves.

How hard is cognitive science?

Cognitive science is itself a cognitive activity. Yet, computational cognitive science tools are seldom used to study (limits of) cognitive scientists’ thinking. Here, we do so using computational-level modeling and complexity analysis. We present an idealized formal model of a core inference problem faced by cognitive scientists: Given observations of a system’s behaviors, infer cognitive processes that could plausibly produce the behavior. We consider variants of this problem at different levels of explanation and prove that at each level, the inference problem is intractable, or even uncomputable. We discuss the implications for cognitive science.

From Structuralism to Cognitive Science

This chapter traces the contribution of Lila R. Gleitman’s research over the course of 50 years. It situates her career relative to the structuralist and behaviorist paradigms of the 1950s and discusses how her work played a central role in replacing those paradigms with those of cognitive science and generative linguistics. It reviews her contributions to assessing children’s knowledge, the role of input in language acquisition, the role of innate structure in shaping language acquisition, and the relation between language and thought. Gleitman raised some of the most powerful arguments against the orthodoxies of her day and ushered in a computational cognitive science of language. Her work took the in-principle arguments of generative linguistics and showed how these could connect to the phenomenon of language acquisition in practice. She innovated new ways of probing children’s knowledge of language, new ways of thinking about that knowledge, and new ways of thinking about learning.

The role of communication in acquisition, curation, and transmission of culture

Veissière et al.'s proposal aims to explain how cognition enables cultural learning, but fails to acknowledge a distinctively human behavior critical to this process: communication. Recent advances in developmental and computational cognitive science suggest that the social-cognitive capacities central to TTOM also support sophisticated yet remarkably early-emerging inferences and communicative behaviors that allow us to learn and share abstract knowledge.

Advertisement for the Philosophy of the Computational Sciences

This chapter deals with those fields that study computing systems. Among these computational sciences are computer science, computational cognitive science, computational neuroscience, and artificial intelligence. In the first part of the chapter, it is shown that there are varieties of computation, such as human computation, algorithmic machine computation, and physical computation. There are even varieties of versions of the Church-Turing thesis. The conclusion is that different computational sciences are often about different kinds of computation. The second part of the chapter discusses three specific philosophical issues. One is whether computers are natural kinds. Another issue is the nature of computational theories and explanations. The last section of the chapter relates remarkable results in computational complexity theory to problems of verification and confirmation.

Exchanging Conflict Resolution in an Adaptable Implementation of ACT-R

In computational cognitive science, the cognitive architecture ACT-R is very popular. It describes a model of cognition that is amenable to computer implementation, paving the way for computational psychology. Its underlying psychological theory has been investigated in many psychological experiments, but ACT-R lacks a formal definition of its underlying concepts from a mathematical-computational point of view. Although the canonical implementation of ACT-R is now modularized, this production rule system is still hard to adapt and extend in central components like the conflict resolution mechanism (which decides which of the applicable rules to apply next).In this work, we present a concise implementation of ACT-R based on Constraint Handling Rules which has been derived from a formalization in prior work. To show the adaptability of our approach, we implement several different conflict resolution mechanisms discussed in the ACT-R literature. This results in the first implementation of one such mechanism. For the other mechanisms, we empirically evaluate if our implementation matches the results of reference implementations of ACT-R.