By: Ludovic Noble, MSc Mind Language and Embodied Cognition 2015, University of Edinburgh
This essay will evaluate whether Integrated Information Theory, as put forward by Tononi, can be considered a plausible theory of consciousness. Before evaluating the theory’s strengths and weaknesses I will explain Integrated Information Theory according to Tononi’s account. This will entail elucidating Tononi’s definition of, firstly, information and, secondly, integrated information. I will then put forward two objections to the account and provide counter-arguments where appropriate. These objections will deal with the unfalsifiability of Integrated Information Theory and the theory’s attribution of consciousness to objects that would not ordinarily be considered conscious. While these objections will provide some challenges to the theory I will provide counter-arguments that soften their seriousness, to varying degrees. I will conclude that, because of these counter-arguments, the Integrated Information Theory is limited but nevertheless a plausible theory of consciousness.
Integrated Information Theory was developed by Tononi in an attempt to provide a theoretical framework within which empirical claims about consciousness can be made. Such a framework is necessary for consciousness to be studied scientifically. In order to provide this framework, Tononi equates the amount of consciousness with the amount of integrated information in a system or its Φ (phi) level. In doing so, Tononi equates consciousness with something that can be characterized “mathematically both in quantity and quality” (Oizumi, Albantakis & Tononi, 2014, p. 1) thus allowing empirically verifiable statements to be made about how much and what kind of consciousness a system possesses.
Tononi needs to provide definitions of both information and integrated information if the theory is to explain anything. He defines the amount of information as a “reduction of uncertainty” (Tononi, 2008, p. 217), meaning that there is less information in a system when there are fewer possible semantic alternatives. For example, the result of a coin toss contains less information than the result of the throwing of a die. This is because the result of the latter leaves five possible alternative states and is therefore more uncertain whereas the result of the former leaves only one possible alternative state and is therefore less uncertain. The former contains less information as a result because fewer possible alternatives can be derived meaning that the result says less about the possible state of affairs in the world. As Cerullo (2013, p. 47-48) explains, this definition of information works in the context of a theory of consciousness because it helps explain why a photodiode that can distinguish between two perceptual states has less consciousness than a human who can distinguish between a vast number of perceptual states. A photodiode can be set to distinguish between a black screen and a white screen, when presented with one of the two. It therefore operates according to only two possible states, black or white. While a human can perform the same task, the human can be said to be at a higher informational state than the photodiode because it is distinguishing between black, white and all the other possible visual alternatives (Tononi, 2008, p. 218), for example, black, white, red and all the frames of all films. This constitutes Tononi’s definition of information as the amount of possible alternatives for any given state; “the ability to discriminate among a large number of alternatives” (Tononi, 2008, p. 218). This leads us next to a definition of integrated information, as it is the integrated factor that, according to Tononi, produces consciousness.
Tononi defines integrated information as, essentially, unified information. This means that information in a system is integrated when it produces new information as a whole. This also means that the system is irreducible to its parts, meaning the function of the whole is lost when it is reduced to its components. When informational mechanisms within a system integrate to produce a structure that is “irreducible to non-interdependent components” (Oizumi, Albantakis & Tononi, 2014, p. 3) they are considered to be strongly integrated. It is this meshing of informational components that, according to Tononi, causes a system to have consciousness. This concept of integration is crucial because it explains why certain information systems can be complicated but not considered conscious. For example, consider the cerebellum. The cerebellum is a part of the brain that can be considered very high in informational content according to our earlier definition of information. While it is a very complicated organ in terms of it carrying a high quantity of information, Oizumi, Albantakis & Tononi (2014, p. 17) argue that it does not contribute to consciousness (2014, p. 17). Bower (2002, p. 150) conceives of the cerebellum as a device whose “output is governed strictly by the pattern of inputs it receives”. By contrast, the output of conscious systems is thought to be less strictly governed by input. Tononi explains this lack of consciousness by the fact that the cerebellum is comprised of many functionally distinct components that are not, in fact, integrated. Schmahmann (2008, p. 4) found that patients with damage to the anterior lobe of the cerebellum had neurological disorders relating to motor control but that their intellectual capabilities were intact. Patients with damage to their posterior lobe of the cerebellum had capable motor skills but limited intellectual processing. This suggests that the cerebellum is composed of functionally distinct parts that do not depend on one another. Conversely, the cerebral cortex, which is thought to be responsible for consciousness, contains functionally distinct modules that can also interact and inform one another (Oizumi et al., 2014, p. 17). This supports Tononi’s claim that informational systems are conscious when their informational components are integrated into a unified system. Tononi’s definition of a system of integrated information is then one where multiple informational components function as an informational super-entity when acting as a whole. The super-entity and therefore consciousness is lost if the integration is dissolved, even if the components can still perform their individual functions independently.
There is already a clear strength in equating consciousness with integrated information. If the quantity of consciousness can be equated with the quantity of integrated information in a system then claims about the quantity of consciousness in a system can be empirically verified. In order to do so one would need to calculate the Φ level in any given system according to Tononi’s mathematical formulas. If Tononi’s theory is correct, then the levels of consciousness in animals, in patients in vegetative states and in artificial intelligence machines can be determined by calculating how much integrated information exists in the system. Tononi (2008, p. 221) points to computer simulations of networks with a high Φ level to demonstrate the similarity between such systems and mammalian brain areas thought to be responsible for consciousness, suggesting that the correlation demonstrates a link between integrated information and consciousness.
However, we may be too hasty to identify integrated information with consciousness. It could be that integrated information in a system merely correlates with its level of consciousness but that the two are not fundamentally linked, as Marcus (2013, para. 9) recognises. Given that claims about consciousness are unfalsifiable, the link between consciousness and any objectively measured phenomena will be tenuous for the lack of ability to objectively prove it. Even if we take cases such as that of the cerebellum to demonstrate a link between consciousness and integrated information this link could be taken instead to be a “barometer of intelligence rather than of consciousness” (Marcus, 2013, para. 9). Tononi needs to provide a solid reason for identifying integrated information with consciousness if the theory is to provide an empirical basis for investigating claims about consciousness.
Whilst this criticism is fair it may not be sufficient to defeat the thesis that integrated information theory is consciousness. This is because all theories that attempt to equate consciousness with objective phenomena are vulnerable to the same criticism. Consciousness is a subjectively experienced phenomenon and, due to its nature, cannot be measured objectively. This is why, in order to study it objectively, it must be equated with some other objective phenomena, in this case, integrated information. However, a link between the subjective phenomena of consciousness and any objective phenomena can never be objectively demonstrated no matter how much the data correlates. If there is to be an empirically verifiable theory of consciousness at all then the link between consciousness and whatever is to be empirically tested will always be one that is correlational. In theory, if technology allows, one could determine the Φ level of a human brain and all of its components. If the Φ level corresponded in the correct way to all components of the brain responsible for producing conscious experience then this would provide a strong reason to suppose that integrated information is responsible for producing consciousness in informational systems. It is therefore not a problem for the integrated information theory of consciousness that it does not provide an objectively verifiable link between consciousness and integrated information because, due to the subjective nature of consciousness, there can be no such link that does not go beyond correlation.
Another objection to Integrated Information Theory accuses it of ascribing consciousness to things that we would not normally count as conscious, such as the before mentioned photodiode. Depending on how the term ‘information’ is interpreted, an atom, under this view, could also be considered conscious. If the state of any proton, neutron or electron in an atom can be considered functionally distinct informational components, then their integration in one atom could be considered a system of integrated information. This fits the before mentioned criteria for integrated information but on a much smaller scale. The state of any given electron can be considered an informational state: there are conceivable alternatives as to its position; it could occupy an alternative shell. Similarly, there could be an alternative number of protons or neurons within the nucleus, giving them an informational value that can be calculated according to Tononi’s definition. These functionally distinct components are integrated to produce a super-entity, the atom. The function of the super-entity is lost if it is reduced to its functionally distinct components. If one accepts that atoms are integrated informational systems then Integrated Information Theory succumbs to a sort of panpsychism. Panpsychism is the view that “consciousness is an essential property of all matter along with energy and mass” and is dismissed as the dualist belief “that objects are inhabited with spirits” (Freeman, 1999, p. 124).
Panpsychism, however, need not be characterized in dualist terms. I will here argue that it is not problematic for Integrated Information Theory to be associated with panpsychism. This will be because the nature of consciousness is such that panpsychism does not necessarily entail dualism but can still un-problematically allow for non-human entities to be conscious, since Tononi rejects the argument from analogy that determines what can intuitively thought to be conscious.
The argument from analogy, as Tononi describes it, tries to determine the level of consciousness in any system simply by comparing its similarity with systems that are certainly consciousness, namely, human systems. As Tononi recognises, this analogy falls apart when we consider man-made artifacts or fruit flies (Tononi, 2008, p. 237). This is because there is no obvious reason why entities that are structurally dissimilar to humans should lack consciousness, if they have a similarly complex cognitive functionality. While a structural similarity with human cognitive systems may be an indicator of consciousness, there is no reason to consider it a necessary condition for consciousness. The only other accepted method (other than employing the argument from analogy) for neurobiologists to “know when other people are conscious, is to ask: ‘what did you feel, and when did you feel it?’” (Freeman, 1999, p. 130). This method of determining consciousness through reporting cannot be employed to discover whether consciousness exists in systems that are not human. So, if structural similarity to human cognitive systems is not a necessary condition for consciousness and the only other way of determining consciousness is through reportage then it is within the bounds of reason to suggest that consciousness can exist in non-human kinds of systems of integrated information. Therefore the association with panpsychism does not so readily threaten Integrated Information Theory.
Many scientists and philosophers will be uncomfortable with the idea that atoms or other non-human entities are conscious. However, this discomfort can be alleviated if we are careful with the manner in which we use the word ‘conscious’ and draw from Heidegger’s idea of ‘Dasein’ instead. ‘Dasien’ (literally, ‘there-being’) was Heidegger’s attempt to introduce a concept for the human way of being in the world, featured in his ontology. ‘Dasein’ describes the “distinctive kind of entity that human beings as such are” (Wheeler, 2014, Modes of Encounter section, para. 1). It describes the continuous unfolding of activity from the point of view of a human, which is considered by Heidegger to be the fundamental ontological entity of reality. Humans, as systems of integrated information, represent a state of being in that there is something that it is like to be a human. This is what consciousness is: the state of being experienced from within (Tononi, 2008, p. 234).
Of course, an atom is not conscious in the way a human is conscious and it would unreasonable to suggest that we could imagine it to be conscious in any way that we usually use the term. However, the atom, as a system of integrated information, can be described as a state of being in the world that is a fundamental ontological entity, similarly to how a human ‘Dasein’ can be described as a state of being in the world that is a fundamental ontological entity. There can be considered a state of being that, experienced from within, is what it is like to ‘be’ an atom. This sounds less controversial than proclaiming an atom to be conscious because we are not indicating any similarity between the conscious human state and the experience of an atom. Instead we are imagining the atom’s existence as a system of integrated information that conforms to a way of being that is experienced from within, which is unintelligible to humans. This state of being is a fundamental ontological entity in the same way that a human state of being, which is a more complex system of integrated information, is a fundamental ontological entity. Tononi (2008, p. 234) alludes to this point when he says that it is not enough to describe a conscious state in order to fully understand it, one has to be the integrated information system and to experience it. Panpsychism has then not been found to be a philosophically unsound position and so proponents of the Integrated Information Theory need not fear it.
Integrated Information Theory is a plausible theory of consciousness. While it is open to worries about the validity of equating consciousness with integrated information, it provides an account which is, in principle, empirically verifiable. The worry regarding the connection between consciousness and integrated information has been shown to be negligible, given that all theories of consciousness that attempt to equate consciousness with an objective property will be open to the same problem. The theory’s issue with panpsychism may be overcome if panpsychism is given philosophical support. In conclusion, while the success of Integrated Information Theory relies on the advancement of technology to test its empirical claims, it provides a plausible theory of consciousness that creates the empirical framework needed to establish a scientific understanding consciousness.
Bower, J., M. (2002). The Organisation of Cerebellar Cortical Circuitry Revisted: Implications for Function [Electronic Version]. Acad. Sci., 978, 135-155.
Cerullo, M., A. (2013). Integrated Information Theory: A Promising but Ultimately Incomplete Theory of Consciousness [Electronic Version]. Journal of Consciousness Studies, 18, 45-58.
Freeman, W., J. (1999). How Brains Make Up Their Minds. London: Weidenfield & Nicolson.
Marcus, G. (2013). How Much Consciousness Does an iPhone Have? Retrieved from http://www.newyorker.com/tech/elements/how-much-consciousness-does-an-iphone-have
Oizumi, M., Albantakis, L., & Tononi, G. (2014). From the Phenomenology to the Mechanisms of Consciousness: Integrated Information Theory 3.0 [Electronic Version]. PLOS Computational Biology, 10, 1-25.
Schmahmann, J., D. (2008). The Cerebellum in Behavioural Neurology and Neuropsychiatry [Electronic Version]. The Official Publication of the National Ataxia Foundation, 36, 3-6.
Tononi, G. (2008). Consciousness as Integrated Information: a Provisional Manifesto [Electronic Version]. Biol. Bull, 215, 216-242.
Wheeler, M. (2014). Martin Heidegger. Retrieved from http://plato.stanford.edu/archives/fall2014/entries/heidegger/