Saturday, June 30, 2012

Cogito, Ergo Sum

People willingly concede that when it comes to nuclear physics or kidney dialysis, specialized knowledge is essential. But let the conversation turn to consciousness, and everybody chimes in, on the assumption that they are all entitled to their own pet theory in the absence of pertinent facts. Nothing could be further from the truth.
One of the reasons for such naive theorizing is that the enigmatic nature of consciousness baffles us all (scientist, non-scientist alike), and we try to analyze it within the realms of our individual conscious experiences. For those truly curious, by Christof Koch will be an intellectual adventure that attempts to provide several such "pertinent facts" and directs us toward the right questions that need to addressed to resolve this longstanding enigma.

 Book review: Consciousness by Christof Koch

First things first. What is consciousness? In contrast to a commonsense definition, that equates consciousness with our inner, mental life, a behavioral definition would focus on actions or behaviors that can be deemed as signatures of conscious experience, a neuronal definition would seek to identify the minimal physiological mechanisms required for any one conscious sensation, while a philosopher's definition would describe consciousness in terms of qualia (the plural of quale), or what it feels like to have a particular experience.

Here are some thought-provoking questions that Koch poses to elucidate the abstruse nature of conscious experience: "Are qualia an elementary feature of matter itself, or do they come about only in exceedingly organized systems? Put differently, do elementary particles have qualia, or do only brains have them? Does a single-celled bacterium experience some form of proto-consciousness? What about a worm or a fly? Is there a minimum number of neurons needed for a quale to occur? Or is it the way neurons are wired together that matters most? Can a computer with silicon transistors and copper wire be conscious? Do androids dream of electric sheep, as Philip Dick rhetorically asked? Does my Mac enjoy its intrinsic elegance, whereas my accountant's slab of non-Mac machinery suffers because of its squat gray exterior and clunky software? Is the Internet, with its billion nodes, sentient?"

Behaviorally speaking, wakefulness and REM sleep represent conscious states, while deep sleep, anesthesia, fainting, concussion, and coma do not. Koch mentions that, when we are conscious, we are conscious of something specific the only exception being during meditative exercises when one can be aware without being aware of anything in particular. This latter bit particularly resonated in my mind given my earlier attempts at meditative practice and my curiosity regarding the neuroscience of the meditative experience.

At the neuronal level, it is essential to identify what Koch refers to as the neuronal correlates of consciousness (NCC). These are the minimal neural mechanisms jointly sufficient for any one specific conscious percept. "Minimal" is key here, because, without it, the full brain can count as an NCC! Koch postulates that neurons in higher regions of neocortex are closely allied to consciousness. Most of the evidence provided in the book in support of this stems out of experiments on visual processing. These experiments point to the fact that not everything our eyes see leads to a visual experience. Once information is transferred from the retina to different parts of the primary visual cortex, substantial unconscious preprocessing takes place before it is passed on to the neurons higher up. Thus, not all kinds of cortical activity are capable of generating conscious experiences. This specific chapter as well as the one that follows (and continues the discussion on our unconscious self) reminded me of two other books I recently read. One of them is The Age of Insight by Nobel laureate Erik Kandel and the other Harnessed by Mark Changizi. While Kandel delves deep into the secret territories of unconscious visual processing, Changizi offers an auditory parallel and illustrates how low-level structures of sound (such as Doppler pitch shift) are detected by our ears and processed without our conscious knowledge.

To answer some of the questions I shared early on, what is needed is a theory for consciousness, that, as Koch says, must not only be "descriptive (i.e., consciousness involves this part of the brain and those connections)" but also "prescriptive (i.e., it must give necessary and sufficient conditions for consciousness to occur)." One such attempt is a mathematical framework proposed by Koch and Francis Crick based on the Giulio Tononi's theory of integrated information. This theory defines each conscious experience being one of many highly differentiated states, with the full repertoire of possible states highly integrated with each other. Further, consciousness can be quantified in terms of the reduction in the entropy of the integrated system relative to the sum of the entropies of its individual states, a measure Koch denotes $\phi$. Koch points out, "if all of the brain'­s neurons were to fire in a synchronized manner, as in a grand mal epileptic seizure, integration would be maximal, but differentiation would be minimal. Maximizing $\phi$ is about finding the sweet spot between these two opposing tendencies." This would explain why pyramidal neurons in the cortex with their small-world network structure may yield a larger $\phi$ and contribute to conscious experience, while the cerebellum with 4 times as many neurons as the entire cortex but a "crystalline-like synaptic organization" does not. Small-world networks are characterized by dense local connectivity and fewer long distance connections. In comparison, cerebellar neurons have a modular structure with minimal interaction between distant modules and are, therefore, far less integrated. Based on this model, Giulio defined a qualia space, a multidimensional space with as many dimensions as the number of states in the system where each point is a probability distribution on the possible states. Each experience or quale can then be represented as a set of interactions between states, or, in other words, by a particular set of arrows linking points in Q-space [link to paper]. Now, as per this framework, any system (living or non-living) with $\phi>0$ has some conscious experience, however faint! The theory of integrated information therefore supports panpsychism, the notion in philosophy that all matter has conscious existence.

Eerie as it may seem to imagine androids, electric sheep, and the Internet echoing "Cogito, ergo sum" or "I think, therefore I am" in the footsteps of Descartes, it may not be a complete impossibility after all! Panpsychism is an elegant concept and has a deeply aesthetic appeal. But eventually everything rests on experimental validation. Macroscopic imaging methods are severely limited by their spatial and temporal resolution. Functional MRI, for example, tracks the BOLD response that occurs on a scale of seconds compared to the millisecond scale of neuronal activity. A typical fMRI voxel contains about a million neurons. But, with emerging optogenetic tools, the picture may change. Optogenetics allows selective optical activation or suppression of a specific class of neurons in a living organism. Koch proposes that, by using optogenetics to switch off cortico-cortical feedback throughout the brain, we can create mice incapable of phenomenal experience. The feedback can, of course, be switched back on and consciousness restoredall with the trigger of a light switch! Welcome to the world of thinking androids and zombie mice!