ASSC 17 Tutorials
Tutorials - July 12
M1: Using Bayes to interpret non-significant results
Zoltan Dienes (School of Psychology, University of Sussex, Brighton, U.K.)
The purpose of the tutorial is to present simple tools for dealing with non-significant results. In particular, people will be taught how to apply Bayes Factors to draw meaningful inferences from non-significant data, using free easy-to-use on-line software: Software which allows one to determine whether there is strong evidence for the null and against one’s theory, or if the data are just insensitive, a distinction p-values cannot make. These tools have greater flexibility than power calculations and allow null results to be interpreted over a wider range of situations. Such tools should allow the publication of null results to become easier.
While the tools will be of interest to all scientists, they are especially relevant to researchers interested in the conscious/unconscious distinction, because inferring a mental state is unconscious often rests on affirming a null result. For example, for perception to be below an objective threshold, discrimination about stimulus properties must be at chance. Similarly, for perception to be below a subjective threshold by the zero correla- tion criterion, ability to discriminate one’s own accuracy must be at chance. To interpret a non-significant result, what is needed is a non-arbitrary specification of the distribution of discrimination abilities given conscious knowledge. Conventional statistics cannot solve this problem, but Bayes Factors provide an easy simple solution. The solution is vital for progress in the field, as so many conclusions of unconscious mental states rely on null results with no indication of whether the non-significant result is purely due to data insensitivity.
M2: Integrated information, predictive coding, and qualia
Anil. K. Seth & Ryota Kanai (Sackler Centre for Consciousness Science and Dep’t of Informatics, University of Sussex, Brighton, U.K.)
Current research in consciousness science must better integrate theory and experiment in developing our understanding of qualia . Two classes of brain theory are now emerging as leading candidates. Integrated information theory (IIT, ) proposes that consciousness has to do with the amount of information generated by a neural system as a whole, compared to the sum of its parts. Integrated information (‘phi’) can be operational- ized as a variant of dynamical complexity and compared with similar measures [3,4]. IIT thus highlights informa- tion theory and complexity as key tools for naturalizing consciousness and qualia. Predictive coding (PC) proposes that perception emerges via Bayesian inference: Perceptual content is determined top-down predictive signals arising from generative models of external causes, which are continually modified by bottom-up prediction-error signals . PC thus highlights re-entrant processing and probabilistic inference as key concepts. While both frameworks are powerfully explanatory, IIT is underconstrained by current cognitive neuroscience and difficult to test, while for PC the relationship between conscious and unconscious perception is poorly specified. In this tutorial, we will first provide basic introduction to IIT and PC with special emphasis on their relationship to understanding qualia. To facilitate interdisciplinary discussion, the tutorial does not assume any mathematical background and we will focus on conceptual understanding of the theories rather than math- ematical details. In a later part of the tutorial, we will discuss how these different frameworks might be synthesized into a coherent computational framework.
M3: First-person methods: Philosophers' dreams or researchers' nightmares? Perspectives from philosophy and the study of dreaming
Jennifer M. Windt (Dep’t of Philosophy, University of Mainz, DE) & Sascha B. Fink (Institute of Cognitive Science, University of Osnabrück, DE)
The best way to learn about the specific character of conscious experience is to study what people say about it. Fundamental features of consciousness (e.g. holistic integration, phenomenal embodiment, specious presence, etc.) were only established as targets for consciousness research through first-person methods. Most agree that such methods are indispensable for consciousness studies. At the same time, phenomena such as change blindness suggest that first-person access to phenomenality is not perfect. First-person reports can be confabulated, contradictory, or incomplete. This calls the validity of first-person approaches into question. Apparently, we need first-person reports – but how far can we trust them?
Dreams illustrate the problem of first-person reports in a compelling manner. Because dreams are largely decoupled from sensory input and behavioral output, dream researchers rely almost exclusively on dream reports. At the same time, dream reports have often been taken to be particularly unreliable, occasionally leading to outright skepticism regarding the experiential character of dreaming. Despite these theoretical disagreements, however, dream research is a thriving field and can provide a fresh perspective on problems
The tutorial has four goals: (1) Provide an introduction to the basic problems raised by first-person reports using the example of dreaming, (2) suggest specific consequences from the philosophical debate on dream- ing for the use of first-person reports in consciousness research, (3) discuss philosophical positions on the validity of first-person reports and the reliability of introspection, and (4) discuss the role and value of the researcher’s own experience.
M4: Measuring (un)awareness
David Carmel (Dep’t of Psychology, University of Edinburgh) & Steve Fleming (Center for Neural Science, New York University)
Most research on perceptual awareness attempts to understand consciousness by investigating the twin themes of conscious and unconscious perception – i.e., what perceptual processes are associated with conscious experience and what can be accomplished in the absence of awareness. There is, however, a great deal of confusion regarding how to assess and measure each of these modes of perceptual processing. This tutorial will offer researchers at all levels an overview of pertinent methodological and conceptual issues, leaving participants with an understanding of the questions they need to consider when designing studies, and how the answers to these questions constrain the conclusions that can be drawn from research findings. For unconscious perception, the questions that will be discussed include how to ensure suppression of percep- tual stimuli from awareness, how to decide which suppression technique is most appropriate for a specific research question, and whether different kinds of unconscious processing indicate similar neural underpin- nings.
For conscious processing, the questions that will be addressed are how to measure the level and extent of subjective conscious experience, whether different ways of assessing reports of awareness (confidence, appearance, wagering) address equivalent constructs, and how detection and identification of perceptual stimuli differ.
Several demonstrations will clarify the issues that will be discussed, and generous provision for discussion will be made to allow consideration of specific problems or issues arising in participants’ own research.
A1: Investigating animal pain and consciousness
Paula Droege (Dep't of Philosophy, Pennsylvania State University)
Neuroscientists have been making remarkable progress in identifying candidates for the neural correlates of consciousness (NCC) in humans. Through careful investigation of conscious and unconscious processes, the role of thalamocortical circuits and information integration in the production of consciousness is becoming clearer. But what about non-human animals? How can we apply the advances in consciousness research to animals (e.g. fish, cephalopods) that share few if any human physical and functional structures? The capacity to represent the current environment in contrast to the past and future is essential for consciousness and marks an important development in cognitive skill. We suggest this capacity for temporal representation can bring together physiological and behavioral evidence to help determine which animals are conscious and which are not.
The tutorial will begin with a discussion of the problems and prospects for research on animal consciousness, considering such questions as: Is nociception sufficient for pain? What is the relation between consciousness and self-consciousness? Then we will review several research paradigms designed to assess consciousness in animals. One important question here is: When is an explanation in terms of higher-level processes such as consciousness and cognition simpler than an explanation in terms of associative conditioning?
A2: Representational theories of consciousness
Rocco J. Gennaro (Dep’t of Philosophy, University of Southern Indiana)
The notion of ‘representation’ is central to many philosophical theories of consciousness and also figures importantly in psychology and neuroscience. Some questions raised by the role of representation in these fields are: What does it mean to say that a mental state is ‘representational’? What is the difference between a first-order representation and a higher-order (or meta-) representation? This tutorial will begin with a discus- sion of how the concept ‘representation’ is used in the philosophical literature on consciousness. In addition, various senses of ‘conscious’ are distinguished and explained. The key question then becomes: What makes a mental state a conscious mental state? We shall survey a number of leading representational theories of consciousness found in the current literature: First-Order Representational Theory of Consciousness (Tye), Higher-Order Perception (HOP) Theory (Lycan), Higher-Order Thought (HOT) Theory (Rosenthal), Dual Content Theory (Carruthers), and Self-Representational Theory (Kriegel). After the main tenets of each approach are presented, we shall discuss the arguments for and against the theory in question. Significant attention will be paid to well-known objections to each theory, for example, the problem of misrepresentation, the question of animal consciousness, and how these theories might address the “hard problem” of consciousness. Finally, there will be some discussion of how these models might be realized in the brain. Also important is the reduc- tionist motive of most representational theorists: Can any of these theories offer a viable reductionist account of consciousness?
A3: The Integrated Information theory of consciousness
Giulio Tononi (Dep’t of Psychiatry, University of Wisconsin), Christof Koch (Cognitive and Behavioral Biology, Cal Tech; CSO, Allen Institute for Brain Science), Nao Tsuchiya (Monash University, Melbourne, AU) & Masafumi Oizumi (Dep’t of Psychiatry, University of Wisconsin)
The Integrated Information theory of consciousness (IIT) has recently attracted attention among conscious- ness researchers. IIT stems from thought experiments that lead to phenomenological axioms and onto- logical postulates (information, integration, exclusion, and compositionality). According to IIT, an experi- ence is an integrated information structure, which in principle can be completely characterized, both in quantity and quality, by determining to what extent a system of causal mechanisms is irreducible to its parts. Many observations concerning the neural substrate of consciousness fall naturally into place within the IIT framework. Among them are the association of consciousness with certain neural systems rather than with others; the fact that neural processes underlying consciousness can influence or be influenced by neural processes that remain unconscious; the reduction of consciousness during dreamless sleep and generalized epileptic seizures; and the distinct role of different cortical architectures in affecting the quality of experience. The tutorial will i) introduce the basic notions of IIT to a broad audience without requiring a mathematical background, and provide hands-on examples in which integrated information can be computed rigorously; ii) introduce measures of integrated information that can be applied to empirical data and discuss how they can be applied to evaluate the level of consciousness in wake, sleep, anesthesia, and disorders of consciousness; iii) demonstrate how integrated information grows in animats adapting to a complex environment, thereby shedding light on the evolution of consciousness; iv) consider theoretical and practical aspects of measures of integrated information, potential problems, and future developments.
Our intended audience is broad. We do not assume any prior knowledge of integrated information theory or information theory in general. In the first part of the tutorial, we start from the basics of the probability theory and information theory, which are key to understanding the theory. After the introduction of the basics, the contents of the tutorial will be at the level of a master class.
A4: Deciphering the information contained in patterns of human brain activity
Frank Tong (Dep’t of Psychology, Vanderbilt University)
Surprisingly detailed information about visual and mental states can be decoded from non-invasive mea- sures of human brain activity. Brain decoding approaches have successfully revealed what a person is seeing, perceiving, attending to, or remembering. Multidimensional models can further be used to investi- gate how the brain encodes complex visual scenes or abstract semantic information, and to reconstruct the stimulus that was viewed. Such feats of “brain reading” or “mind reading”, though impressive, raise impor- tant conceptual, methodological, as well as ethical issues. What does successful decoding reveal about the sensory or cognitive functions performed by a brain region? How should brain signals be spatially selected and mathematically combined, to ensure that decoding reflects inherent computations of the brain rather than those performed by the decoder? What ethical considerations might emerge with the advancement of these methodologies? The tutorial will cover the fundamentals of “brain reading”, and should be suitable for people from a broad range of backgrounds, with one component emphasizing the more technical and mathematical aspects of pattern classification. Questions and interactive discussion will be emphasized, especially when considering the strengths and limitations of fMRI pattern analysis methods.