Information Flow Theory

Information Flow Theory of Consciousness (IFT) is a materialist theory which defines consciousness as form of self-awareness which may arise within any system capable of processing information. IFT draws a distinction between computational complexity or intelligence from conscious self-awareness (CSA) as independent features of a system based on direction of information flow. IFT is distinguished from other theories of consciousness in that it provides a bottom-up explanation of the emergence of the conscious percept which scales through evolution, makes discreet predictions regarding both biologic and artificial consciousness, and is experimentally falsifiable.

Information processing units

IFT defines any information processing unit (IPU) as a discreet system which interacts with its environment through three fundamental components, an input (I), a computation or processing node (N₀), and an output (O). Within the context of the IPU, “awareness” is operationally defined as its ability to both detect and process information received through its input. The IPU, by definition, cannot be aware of any information that may exist in its environment but is undetectable by its input stream. By extension, due to the strictly unidirectional information flow, the IPU cannot be aware of itself as its only informational input comes from information external to its processing node.

Computation within the processing node (N₀) can be envisioned as scaling from simple to complex through the progressive linking of multiple subordinate IPUs together in any number of configurations. Regardless of the degree of processing complexity, this composite node remains “aware” only of the external information provided by the input stream. This description provides the rationale for separating the concepts of intelligence (e.g. processing complexity) from the concept of self-awareness as distinct phenomena (see Figure 1).

Internal model of external reality (IMER)

N₀ computation can be mapped onto biologic structures ranging in complexity from bacteria to formal nervous systems. Within this context, the evolutionary purpose of increasing N₀ computational complexity may be seen as an adaptive improvement in translating environmental sensory input into actionable behavior thereby increasing overall fitness. Organisms generally possess an array of sensory apparatus capable of translating a range of environmental stimuli into the input stream. This poses a practical issue as different aspects of information regarding a single event may arrive over a range of rates and times (e.g. visual versus auditory stimuli). In order to deal with this, nervous systems evolved mechanisms to integrate these temporal and spatially distinct information streams into a unified internal model of external reality (IMER) in a process sometimes referred to as “binding”. The IMER arising out of information binding is critical in order to encode for an appropriate output behavior to deal with a particular environmental stimulus (e.g. capture prey, defend against attack, mate, etc…). However, regardless the apparent computational complexity of the IMER and the array of behaviors it produces, the unidirectional flow of information precludes self-awareness at the level of the composite or C₀ IPU.

Self-awareness

Self-awareness within the operational definition of IFT indicates that a system is capable of both receiving and computing information regarding itself. This mandates interruption of unidirectional information flow through the addition of a minimum of two features to the C₀ IPU: a second computational node (N₁) and a recursive information circuit (R₁) capable of transmitting the output of N₀ into the input of N₁ [Figure 1]. Within this arrangement, the new aggregate system (C₁ IPU) may be described as “self-aware”. Specifically, the N₁ node is “aware” of the computational output of the N₀ node, but critically has no independent access to environmental informational input nor may it directly influence the environment through behavioral output.

Conscious self-awareness (CSA)

A comparison of the C₀ and C₁ IPU architecture demonstrate a divergence in the concepts of computational complexity (e.g. intelligence) and self-awareness as neither necessitate the other. The term consciousness within IFT emerges, not as a discreet phenomenon, but rather as a description of how self-awareness is experienced through the integrated computational output of the N₀ node. This definition addresses why human consciousness is mediated through our disparate sensory experiences and yet is perceived as unitary. These same features are directly mirrored in the N₀ computation of the human IMER which binds distinct sensory input into a unified mode. This  N₀/N₁ symmetry suggests a shared basic computational architecture between the two nodes. “Conscious” self-awareness (CSA) therefore may be seen as the byproduct of algorithms which evolved to compute behavioral output in response to the environment now redeployed to compute consciousness in response to the IMER.

Human CSA and language

Human CSA (CSA-H) is characterized by an additional property which is not allowed within the C₁ IPU architecture, namely the ability to communicate externally. This ability requires an additional link between the N₁ node and the IPU output stream. In order to avoid collapse into unidirectional information flow, this unique recursive circuit (R₂) must be mediated through the N₀ node. An IPU containing a complete R₁/R₂ circuit (C₂IPU) gains the capacity for the N₁ node to both influence the construction of the N₀ IMER as well as externally communicate its computational state to the external environment. In humans this is observed in the process of learning and incorporation of scientific concepts in the way we organize and process the physical world (see Table).

Predictions of Information Flow Theory

IMER complementarity

From the perspective of the C_1/2 IPU, there is no external reality beyond that which it can sense and integrate into its IMER. As the IMER can never model the totality of independent objective reality (IOR), the degree to which the IMER approximates IOR may be expressed as the IMER/IOR ratio with a lower boundary of 0 and an upper boundary <1. In order for two independent C_1/2 IPUs to agree on the presence of mutual consciousness within one another, there must be some shared aspects of the two IMERs.

Information transmission directly influences consciousness

As CSA-H arises from the combination of information flow (R₁/R₂ circuit) and information processing (N₁ node computing the output of the N₀ node). Multiple regions of the human brain have already been identified which correspond to centers of consciousness consistent with the composite N₁ node. These include the superior temporal sulcus and temporo-parietal junction. However, IFT predicts that manipulation of flow alone, independent of activity within these centers, should result in an altered manifestation of consciousness. This is evident in several examples derived from clinical medicine. General anesthesia has been found to work through the generation of oscillations or waves of neuronal activity which transiently interrupt normal information signaling between regions of the brain involved in consciousness. Similarly some patients who have undergone a corpus collosotomy (transection of interhemispheric connections) manifest the so-called “split-brain” syndrome in which each hemisphere exhibits an independent conscious experience.

IFT and artificial CSA

The principles of IFT apply equally to biologic and artificial systems capable of processing information. Even the most complex neural networks remain analogous to a C₀ IPU with individual neurons connected through multiple layers functioning as a composite N₀ node. Self-awareness at would therefore require a second network capable of receiving and processing the output of the first network. This C₁ IPU would “experience” awareness but solely within the context of its learning algorithm. The addition of an R₂ circuit within this artificial system would provide the IPU with the ability to communicate its state however this communication could only occur within the context of the information it is designed to process.

Of critical importance to future technology capable of artificial general intelligence, IFT teaches that there is no “reality” distinct from the constitutive components of the IMER. A sensory based human-like IMER built into the design would therefore enable a machine experiencing conscious self-awareness in silico (CSA-IS) to be able to communicate its state in concepts familiar to CSA-H through a common language. Perhaps most importantly, pro-social concepts such as altruism and empathy should be strongly represented within this IMER. To a machine, unless these pro-social concepts are built into the IMER, they would not even exist.

Language facilitates superhuman CSA (CSA-SH)

IFT predicts that human natural language functions as a type of input/output flow which uses formalized symbols to organize and transfer information regarding the IMER. Sharing of knowledge has facilitated an increase in the human IMER/IOR ratio over time, not within individuals but rather within the species as an aggregate. The capacity to extend the IMER beyond the limits of intrinsic sensory input represents a new functionality as it requires C₂ IPUs working in concert. This coordinated system functions as a superhuman consciousness which is able to dynamically iterate upon itself, increasing its collective IMER/IOR ever higher. The cultural and scientific progress of the human species may therefore be conceived of as a single C₃ IPU consisting of a collective of C₂ IPUs (e.g. individual humans) sharing input and output streams of information (e.g. language) both in series (e.g. over time) and in parallel (e.g. over space) to produce an aggregate superhuman CSA (CSA-SH) (see Figure 2).

Testing for consciousness: beyond the Turing test

IFT predicts several experimental models to probe for the presence of the recursive N₁ processing within an indeterminate IPU required for CSA. The addition of N₁ computation to the output stream should have several detectable differential effects. Relative to information undergoing N₀ processing alone, the added R₁/R₂recursion and N₁ computation may be discriminated by a relative delay in the shortest possible output response to a given input signal. Similarly, N₁ processing would require a larger energy input resulting in information in a more organized or lower entropic state than that of N₀ computation only. Finally, as CSA arises not from isolated processing N₀/N₁ processing but rather from their interconnection through R₁/R₂ recursion, interruption of recursive information flow alone should be capable of reversibly eliminating CSA for the duration of the interruption. In fact, each of these experimental methods have already been demonstrated, to some degree, in human subjects. Time delays of 500ms have been documented between conscious and unconscious processing in classic experiments of the so called “readiness potential”. Similarly, functional imaging studies which rely on the differential use of blood and glucose uptake to detect active brain regions have identified discreet regions of increased energy use associated with the conscious performance of various tasks.

IFT and panpsychism

By providing a discreet architecture both necessary and sufficient for self-awareness, IFT introduces both upper and lower constraints on the presence of self-awareness. In principle, any system possessing two computational nodes linked by at least one unidirectional connection may be considered self-aware, regardless of degree of computational complexity. These arrangements may throughout nature both in isolation as well as within larger composite nodes. However, the character of their “awareness” will be dictated solely through their informational input. A unique IPU arrangement may also be predicted at the level of the entire universe where the input and output information streams themselves become recursive as there is, by definition, no external environment at this scale. Within this “universal” or C₄ IPU, all other IPUs may be seen as constituents of larger composite nodes allowing for the concept of a universe conscious of itself (see Figure 2).

References

This article "Information Flow Theory" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:Information Flow Theory. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.

📰 Article(s) of the same category(ies)

• Noogenesis
• DeepSea Technologies
• Multimodel Deep Learning
• Expected float entropy
• Deep feature synthesis
• Causal Neural Paradox (Thought Curvature)
• Bing Predicts

---