Consciousness and the formation of ‘I’

Consciousness – where is it?

The preceding section elaborated on in-time pattern representation and processing in cognition – Spatial embedding and isomorphic spatiotemporal accommodation of patterns projecting upon a sensory perimeter. In this section I propose some conceivable, and hopefully insightful, operating principles for machinery that may generate ‘mind’ and ‘self’. Building on the ideas presented in the preceding section, here there is an attempt to define the objective and outcome of “computations” that may support the emergence of consciousness in systems.

Since this section aims to consider mechanisms pertaining to the emergence of Consciousness, the so called ‘Hard Problem’ of consciousness needs to be looked-at up-close. One assumption that is not questioned here is that consciousness is a natural phenomenon that humans can hope to understand better, just as we hope to gain a deeper, more accurate, wider reaching, understanding of other aspects of physical nature. For progress to be made, a perspective is required, in respect of the application to consciousness theory, of the distinction between advancement that may lie adjacent to the horizons of our scientifically (and mathematically) expressed models of the world, vs aspects of phenomenology that are out of reach, lying beyond a chasm of theoretical shortfall. We may tackle the former and constrain the latter. Even if proposed delineations between that which may be modeled and that which awaits further developments in physics can only tenuously be supported, or informally and intuitively conjectured – a set of multiple conceivable possibilities regarding the constraints applicable to ‘hard’ aspect of conscious phenomena can be something to work with. The exploration of the consequences of assumptions regarding such constraints, and of models compatible with these constraints, may yield some truth.

The proposed schematic model of ‘I’ described below is complementary to and dependent on some notions and assumptions regarding nature of phenomenality. In both the preceding section and this one it is assumed that consideration of phenomenology can be dissociated from the discussion of computational models and deferred. This may be allowable if the aspects being dissociated manifest on different orders of physical scale, and will be explored further in the next section of this work, which considers constraints conceivably applicable to phenomenality as a natural aspect. It is however important to note in this context that a locus bound aspect of phenomenality is implied by the above.
It seems likely that the computational process extends in the lower spectral range of information capture, whereas the phenomenal aspect, to the extent that it may be computationally processed, is manifested in perimeter bound local patterns characterized by smaller and shorter spatial and temporal orders, respectively. Intuitively it seems that the phenomenal may be manifesting on a far smaller (therefore local) order of scale, so that it may be ‘envelope’ modulated by a ‘neural like’ computational process.

Thus, although what is meant by “sensory” and “perimeter” – terms used throughout this work – begs to be more finely defined, the adoption of an extended notion of a “perimeter” as a bounding “surface” upon which all phenomenal content is manifested – suffices to create a conceptually useful dissociation between all loci of phenomenal origination and the workings of the cognitive machine concerned with the representational/computational aspects of cognition.

Inverted mind theory

Introducing Inverted Mind Theory, a physically orientated treatment (grounded would be too strong a claim) proposing an almost Ptolemaic-Copernican change of perspective relative to conventional ‘internal model’ and ‘neural paradigm’ theories of cognition. The theory presented in this section and the next is essentially a variant of the Dynamic Core notion (Edelman, Tononi), that is consistent with Global Workspace theory (Baars, Dehaene, GNW, provided one is willing to accept a dynamic core as a micro-scale correlate of the contents of the workspace at any given time, and some constraints on the core’s dynamic trajectory), and is implemented using Adaptive Resonance mechanisms (Grossberg) that integrate information to produce Main Complex Phi – as would be required per Integrated Information Theory (Tononi)…
On the view that applies to the phenomenal/experiential level it is also compatible with Reflexive Monism (Velmans) by virtue of dual-aspect fundamentals and the role of engineered particulars in a pan-proto-phenomenal or pan-experiential world. It is tainted by ideas tangential to, and reminiscent of, some aspects of moderated sensorimotor theories of consciousness; In it’s reliance on re-enact-able invert-able isomorphic encoding of sensory projections upon an embodied form that interacts with the world, it appeals to something like ‘sensorimotor contingencies’ (Noe, Regan, but see Gamez. Grush gives an interesting account of sensorimotor contingency, which is compatible with the theory presented here).

Surely all of this must amount to a real mess, not so? Well, it serves to highlight that while all these well thought out, well articulated and well researched theories hint at some direction, they are not differentiated along facets that truly constrain their combination. With some latitude, a widely consistent, even unifying, combination is conceivable (trivially, unless some of the above are grossly misguided).

This treatment constitutes an attempt to inch closer towards an understanding of why such a combination must be missing, and how it could come about. It proposes and builds upon some simple natural principles, here deemed necessary for the manifestation of phenomenally conscious agents. It takes care to eschew semantic glossing-over, and tries to bite into some essential difficulties left opaque by the above theories.

Why have I called it ‘Inverted Mind Theory‘? Several distinct perspectives of ‘inversion’ apply:

  • The agent’s world model, encoded according to the principles outlined in the preceding section, is represented by the literal iconic tracing of projections upon the agent’s embodiment. Through the establishment of relational consistency across multiple angles the agent hones it’s encoded representations [by segregating (factorizing) and binding] until it can sustain dynamic in-time constitutive patterning that, from within the agent’s embodiment, is equivalent to the world model being virtually ‘out there’ – i.e. to that which would project (elaborated below).
  • The encoding is invert-ably enact-able, as in a universal approximator. Network connectivity is reciprocal, encoding/recognition is isomorphic, and the system’s target function is to maximize resonance, approximate equilibrium, with dynamics that are constrained to particular activity modulation frequency bands. The encoded modulation of configurations of ‘dense activity’, representing momentary states of the agent, in the patterns, effects the ‘diffusion’ of the agent into the patterns, it’s coming to be permeating the patterns (assuming a simple, cross-order-of-scale, compositional approach in respect of the constitution of a conscious agent, a position adopted throughout this work).
  • An even more left-field aspect of inversion comes out with regards to the notion of the agent’s phenomenal ‘perimeter’. If phenomenality is the result of some form of combination and integration of (proto-)phenomenality, mediated through ‘binding’, then what physical (biological, cellular, sub-cellular) elements may partake in the generation of (proto-) phenomenality, vs binding? If (proto-) phenomenality is a dual aspect property of physical interactions, or if panprotopsychism or panpsychism of any sort manifest in the world, what then, in such a setting, would be the role of a conscious agent’s neural-like and sensory system pertaining to the formation, ‘tuning’, combination and binding of consciousness? The beginning of a suggested response to this last question – with regards to binding – can be found in this section, whereas some exploration of conceivable arrangements in a (proto-) phenomenal world, and the relations between (proto-) phenomenality and computational processes, are examined in the next section considering computation and phenomenal content. Some strange ideas fall out: In such a setting it is not inconceivable, not even implausible, that some (proto-) phenomenality, corresponding with sensory organ stimulation, may arise within the peripheral sensorium rather than within the brain. According to this idiosyncratic conception of ’embodied cognition’, the brain effects most of the modulation and binding through the creation of, and association with, a spatiotemporal ‘I-space‘ construct (see below). Phenomenality comes about from (proto-) phenomenality by containment and association in the spatiotemporal order imposed by relata and relations inherent in the ‘I-space’ construct, but not all information contributing to phenomenality must necessarily be communicated to the core, and conversely (proto-) phenomenality may manifest locally e.g. in sensory organs.
  • Inversion also applies to the time dimension; here, a notion of presentism refers to the manifestation of consciousness over the ‘now line‘. this requires temporal pattern compression into spatial representation – to support a concurrently distinct conscious impression of temporal passage, and a multitude of A-theory (temporal theory type) ‘temporal indexing’ situations – temporally ‘retentionist’ associations of unfolding events.

The formation of ‘I’ in the world model

The virtual projection of the (external and internal) world model on the phenomenal sensing perimeter of an agent is it’s iconic (and only) presentation to the agent. If an agent’s world model, including the representation of it’s own volume within the world, traces iconic projections of the environment upon it’s (abstract) sensory-phenomenal perimeter, then invariant pattern elements derived from, and driving, that projection correspond to invariant pattern elements in the world that produces the patterns (recognition corresponds to the in-time inverse of isomorphic encoding) – and the agent’s world model is therefore equivalent, within the limits of representational specificity and resolution, to the world that would project these patterns.

If there were an ‘always on’ core of activity (e.g. Heartbeat and Default Mode Network, Park et al, Nat Neurosci 2014; implicated interoceptive processing in the insula, etc.), corresponding to projections of ever-present internally generated and ambient projections on the abstract perimeter, then that ‘always on’ relatively isotropic configuration would correspond to some continuous sense of ‘what it is like’ to be the agent, while attracting and and forming associations with the patterns effected by the various interactions of the agent with the world. The latter and the former aspects would be co-dependent, any demarcating boundary would be arbitrary. This is ‘I-space‘. The primary ‘I’ experiential quality in ‘I-space’ would amount to some degree of ‘pre-reflective’ conscious perception of patterned qualities in ‘I-space’; a rudimentary ‘absorbed’ form of self. An ‘always on’ core would obviously be dynamic i.e. have a spatial configuration that varies in time, but may have components exhibiting relatively stable activity e.g. corresponding to regular tonic and phasic-oscillatory somatic patterns. With extreme simplification, ‘I’ could grow from the composite generated by relating, (including modulating by filtering and re-enacting) spatiotemporal patterns of stimuli projected upon sensing loci, through an effective continuous common core of activity.

Such notions applied to biological ‘self’ may be consistent with Damasio’s or Merker’s proposals, regarding a confluence of sensory and internal somatic sensations effected through a brainstem processing hub (due, amongst other causes, to intrinsic structural proximity, the promotion of continuity, and Hebbian learning, the relatively stable flux of signals that originates in the portion of the perimeter that is involved with internal body functions, regulating and homeostatic mechanisms may produce a reference signal source/target that, since always present, becomes entwined in filtering transformations of sensory patterns. These notions could conceivably extend to accommodate hedonic and emotional expression, being applied through a combination of wide-ranging humorally mediated modulations of somatic phenomenal expression, coupled with canonical modulations of effective network connectivity, having both efferent and afferent filtering implications).

Focal Symmetry

How could the so called ‘binding’ of ‘I’, and of the intentional contents of ‘I’ be effected? This section is limited to considering the computational aspect of binding.

The ideal representation of an invariant element in the agent’s world model would be constituted of an isomorphic compression to a point. In a non-ideal network implementation, a pattern projected on the perimeter (or filtered off the perimeter) would be in symmetric correspondence with a dynamic configuration of activity within the agent’s system.
An approximation of an invariant point would be constituted of population activity tending to synchronous, isotropic, activation within temporally stable spatial boundaries. By extension the complete momentary state of the agent, insofar as it is encoded, constitutes a configuration of isotropic population activity.
If quanta of activity within the agent’s system are computational proxies of sensed experience then a momentary spatial configuration of isotropic activity qualifies a corresponding momentary experience of ‘what it is like’ to be the agent.

A perimeter pattern is isomorphically transformed into and out of an isotropic ‘cloud’ of (synchronously, homogeneously) active modular points.
The ‘shape’ of the active population of modules or units varies in-time with the ‘imagined’, anticipated and actual stimulus patterns projecting upon the perimeter; Some of these arise from continuous internal and ambient stimulation that drives an ‘always on’ core of activity. The morphing of the spatial configuration of isotropic activity is a realization of a ‘dynamic core’ (Edelman & Tononi).

An effective focal point coupling the computational patterns corresponding to ‘what it is like’ to be the agent could be mediated by a set of interacting points on the isotropic activity side of pattern transformation, because the spatiotemporal constructs of the agent’s world model are expressed solely through the set of synchronously active points..

Since patterns projected on the perimeter (or filtered off the perimeter) are in symmetric correspondence with dynamic spatial configurations of isotropic activity within the agent’s system (i.e. a morphing set of active populations of units or modules, each one an invariant representation of a pattern factor, each playing out as isotropic activity throughout it’s particular encoded duration), modulations and manipulations of these configurations effect modulations and manipulations of the agent’s perceived world (inclusive of perceived internal state), insofar as the state is of patterned constitution. If the isomorphic efferent is a filter function then manipulation of internal state would affect receptive sensitivity and selectivity (i.e. the system’s “tuned” state would change).

Synchronicity: From implied equivalence to ‘dense activity

By the theory being proposed, factorization (encoding), recognition and completion (decoding) occur contemporaneously and interdependently, as facets of canonical modular processing effecting isomorphic transformations of perimeter patterns unto (equivalent) spatial configurations of isotropic activity, and the concurrent inverse of these, i.e. the in-time tracing of perimeter patterns. On one ‘side’ of the process there are spatiotemporal patterns, on the other a residual or mediating spatial configuration of isotropic state. Almost trivially, isotropic state implies synchronicity. Whichever point belonging to an isotropic active set one looks at, it behaves equivalently to all others in the set, which exhibits a degree of coherence (see below), regardless of how the points may be inter-connected. Equivalence is in respect of the signal expressed synchronously – the more properties that evolve synchronously the greater the degree of set equivalence.

Synchronous oscillation is information poor; isotropic oscillations are the penultimate residual following temporal information extraction and embedding, as a synchronous motif may only devolve into (rectified) noise.
Considering spatial configurations of isotropic activity and assuming activity is packed in discrete quanta, each with a characteristic short duration, we may speak of “dense activity” in a population of active points. A densely active population would manifest a degree of coherence throughout a spectral band even in the absence of distinct oscillations e.g. where and when activity of units is correlated but stochastic, and power is distributed throughout a spectral band.
Spatiotemporal relations between sub-patterns, be these explicit iconic representations of stimuli or representation manipulation controls, can be conveniently represented as relations (spatially and temporally contiguous or dis-contiguous affinities) between the equivalent spatial configurations of synchronous activity.
Therefore at a minimum, with regards to learning, synchronicity is correlated with (successful) pattern factorization and embedding, and may serve as a promoter of modular inter-connectivity as conventionally proposed by many theorists (mediated by spike timing dependent plasticity), whereas with regards to the exercising of learnt capacities synchronicity indicates contemporaneous factorization, recognition, manipulation and reconstitution/composition.

In addition, depending on how phenomenality arises and is associated with the ‘computational’ process, synchronicity could be a correlate or a medium of instantaneous phenomenal manifestation. Computational constructs supporting a range of possibilities are elaborated in this section. Phenomenology is considered in the next section.

From quanta of activity to dynamical spatial shaping of noise sources

The notion that the system’s processing involves dynamical spatial and amplitude modulation imposing patterning over a population of noise generating loci aligns well with:

  • The notion that the ideal final residual of a spatially-embedding isomorphic transform of patterns is a distinctly spatially shaped source of noise.
  • The temporal ‘smearing’ effected through recursive [pattern <==> isomorphic transformation <==> isotropic activity configuration] that would be required to sustain cross-modality/sub-pattern combination relations that accommodate system sensory relay and processing delays, bringing about a ‘specious’ present (see next section, headings related to temporal binding).
  • The description of a densely active noise generating population as a universally synchronous activity core that functions as an isotropic virtual focal point.
  • The conjuring of imagined patterns being reliant on an always present spontaneous activity of noisy sensors and in sensory tracts (inclusive of peripheral extensions in the absence of blocking) and somatic feedback from tonic activity, that stochastically generates quanta of activity densely enough for filters to effectively shape the imposed patterns.

<-Prev                      Next->

2 thoughts on “Consciousness and the formation of ‘I’”

  1. You are re-stating Graziosi’s “theory” recently “explained” at Conscious Entities, where I saw a post by you. Both of you provide various philosophical and computer waffle o try to build a “system” around a simple core idea- that biological sites are directly represented in a brain by sites’ neurons – they belong to sites, not a brain. Your sensory perimeter, and Graziosi’s sensory systems, and neither of you do anything more than show enough faith in that core idea to build abstract computer waffle around it.

    How did you get so much faith in that idea? You are clearly not a neurologist, and neither is Graziosi, as neither of you deconstruct the neuron for the magical capacities you hope they have to represent biological functions And neither of you set out any useful specifics about your supposed sensory perimeter or system, just vague spatial-temporal properties that must emanate from sites somewhere. I gave Graziosi a link to my work at Conscious Entities late last year before his sudden new idea. Did you read my work, or Graziosi’s “new idea” before developing “your” idea?


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s