Theories of Consciousness: two examples

Theories of Consciousness:

two examples

28.11.2019

[email protected]

The Global Workspace

Bernard Baars, 1988, 1997, 2002.Baars

Background

In many respects, there is a strong

contrast between the organization of

the brain and that of consciousness

Brain

• Billions of neurons, firing at an average rate of 40 nerve impulses/sec

• Dense interconnections: we can reach any single neuron from any other in less than seven steps!

• Processing is massively parallel and highly distributed

• Many specialized processors or modules that operate in a largely independent, fast, automatic, often mandatory, and inflexible fashion

• We could speak of a 'society of modules'

Brain: specialized modules

• Examples of specialized modules in reading:

– letter and word identification

– letter-to-phoneme mapping

– eye-movement control

– linguistic analyses such as

• matching words to a mental lexicon

• syntactic analysis

• semantic analysis

Brain: specialized modules

• Important: These specialized modules operate unconsciously

call them USMs (unconscious specialized modules)

The Stroop effect:

Blue

Red

Green

Red

Blue

Green

Red

Blue

Green

Consciousness

• In contrast to the USMs, consciousness

– is slow (calculate 15 x 15 x 15)

– is serial (contents enter consciousness one-by-one)

– has limited capacity (we can be conscious of only one

content, e.g. one of two simultaneous spoken stories; one

ball game on screen; only the currently rehearsed item in

WM; we can execute only one action that demands

consciousness)

– is flexible (can adapt to novel situations)

– is context-sensitive (responses can be adapted to

context)

– is relational, associative (can establish links between

any two conscious contents classical conditioning!)

Baars' answer:

A Global Workspace

What architecture could explain

these largely opposite properties

of the brain and consciousness?

Global workspace (GW)

• The idea: Consciousness is a global workspace that serves to exchange and distribute information between different brain processors

• Consciousness = central information exchange of the brain, the publicity organ of the brain

• Consciousness makes information globally available for all the unconscious specialized modules and other brain processes

• Consciousness creates access to many (all?) brain processors

Consciousness creates access

• Simply by being conscious, information gains access to the vast collection of unconscious brain resources, such as USMs, memory stores etc.

• This is impressively demonstrated in (associative) learning: Simply by relating two (or more) items in consciousness, we learn them. Consciousness of something is all that is required to learn it. Completely unconscious learning, on the other hand, does not seem to be possible.

• Similarly, simply by consciously looking at pictures, we gain access to and can store them in the vast memory system of our brain

Consciousness creates access

• By simply being conscious of a problem, we access the problem-solving machinery of our brain, which will unconsciously deal with the problem, and, if we're lucky, come up with a solution and present it to consciousness

• Another impressive demonstration of global access is the fact that conscious biofeedback enables access to virtually every brain system, from alpha waves to single neurons

Global workspace

• On the input side, many contents, inputs, or stimuli compete for access to the global workspace= consciousness, but only one content at a time can become conscious

• limited capacity, serial, slow

• Attention is the mechanism that selects which content gets access to consciousness

Consciousness

brain processors (USMs etc.)

global

distribution

or access

bottleneck (attention)

input

competition

potentially conscious contents

GW as a theater

• This is the basic idea of the global workspace: input competition and global distribution of information in consciousness

• But the complete GW model has more features, which Baars has summarized into a neat and powerful metaphor: the theater metaphor of consciousness

Elements of the theater:

actors & the spotlight

• The actors correspond to the many contents that compete for consciousness.

• Only the actor in the spotlight, however, is conscious

• The spotlight corresponds to attention ('spotlight of attention'). As we saw last time: attention creates consciousness. We become conscious of what we focus our attention on.

• The actor in the spotlight announces his message to the large audience in the darkness of the theater hall


the audience

• The audience is unconscious. It corresponds to the all the unconscious processors in the brain. Most notable these are the unconscious specialized modules (USMs) we met before. The message proclaimed by the actor in the spotlight is globally broadcast to all these modules.

• It is important to note that not all the USMs respond to the global message. Only those that can "read" the global message will respond to it. So each module decides locally what to do with the global message.

• A USM will respond to the global message if it can read it, and if the message is novel and informative. In that case, the USM will adapt to the novel information. After that the information becomes redundant and the USM will habituate and no longer respond. If the information changes again, the USM will dishabituate.


the audience

• Other elements of the audience are memory sytems, as e.g. those related to vocabulary, syntax, meaning, autobiographical events, knowledge, beliefs etc, or motivational and emotional systems


the fringe

• Around the center of the spotlight there is a fringe. It corresponds to those contents that are only vaguely conscious. These could be feelings of knowing something, of familiarity, beauty, goodness, or rightness. Fringe events are somehow conscious, but we can't tell exactly how we experience them.

• The qualia associated with fringe events are unclear, fuzzy, vague


the stage

• The theater stage corresponds to working memory (WM)

• As we saw last time, items in WM can be made conscious immediately, they are held in an active temporary information store

• As we also saw, however, only the currently rehearsed item in WM is actually conscious. This item corresponds to the actor in the spotlight, whose message is conscious.


actors competing for access to the stage

• The actors competing for access to the stage are the many unconscious contents that could potentially become conscious.

• These include sensory stimuli, visual imagery, inner speech, emotional feelings, thoughts and ideas


backstage

• Behind the scenes many unseen operators control or influence what is going on on stage

• These operators include a director, a spotlight controller and local contexts

• Less metaphorically speaking, these context operators correspond to processes that are themselves unconscious, but shape our conscious experience


backstage: director

• Note that the global workspace itself is not an executive, it does not itself decide which contents become conscious and are globally distributed

• The global workspace is controlled by executive systems in the brain that use it to achieve goals for which consciousness is required (e.g. learning)


backstage: spotlight controller

• Attention is what controls the spotlight (or maybe attention is the spotlight)

• Attention can be controlled either consciously or, very often, unconsciously


backstage: contexts

• Contexts are by far the most varied and numerous unconscious sources that constrain conscious experience

• Every unconscious process or state that shapes the way in which we consciously experience something is called a context

• Contexts are by definition unconscious. They can, however, be made conscious, but then they cease to be contexts

Neural global workspace (NGW)

• We're looking for brain systems that could implement a global

workspace architecture with input competition for access to

consciousness, selection of a content for consciousness

("attentional bottleneck") and global broadcasting / distribution of

this content

• These are:

– brainstem reticular formation (RF)

– reticular nucleus of thalamus (NRT) and intralaminar nuclei of

thalamus (ILN)

Horizontal cut through the brain, view from above

NRT(reticular nucleus

of thalamus)

ILN(intralaminar nuclei

of thalamus)

Brainstem reticular formation

• RF = diffuse, global activation system

• "Diffuse", because the RF has widespread, unspecific

projections to the cortex

• Stimulation of RF activates cortex arousal

• Shut-down of the RF leads to coma

• "Reticular" means "network-like", because RF neurons

form a densely interconnected network

• The RF receives input from all major sensory and motor

system of the brain, incl. extensive collaterals from

sensory pathways on their way to the thalamus

reticular formation (RF)

NRT(reticular nucleus of thalamus)

• The NRT is like a sheet covering the lateral surface of the thalamus

• It consists only of inhibitory neurons

• It is extensively connected with the specific (incl. sensory) nuclei of the

thalamus modulation of specific nuclei

• Nearly all information from the thalamus to the cortex passes through the

NRT the NRT is in an ideal strategic position to gate, control, select the

information that flows to the cortex

• The NRT is therefore said to consist of "gatelets", i.e. an array of gating

circuits

• The NRT itself sends no projections to the cortex, but it receives collateral

projections from nearly all of the cortex

• The NRT is also interconnected with

– ILN

– midbrain/brainstem reticular formation

"gatelets"

NRT

NRT(reticular nucleus of thalamus)

• The NRT shows input competition: it selects from

among a number of inputs (e.g. visual, auditory,

somatic) one for access to the cortex, by opening the

corresponding gatelet, and by shutting down the other

gatelets

• Thinking back to the theater metaphor of the global

workspace, the NRT could therefore be the mechanism

implementing the spotlight of attention that selectively

highlights one particular content of consciousness

• But what does the global broadcasting of this content?

Thalamus and Cortex

• All or nearly all connections from thalamus to cortex are

reciprocal

• In fact, the cortex sends back 10 times as many

projections to the thalamus as it receives from it!

• The thalamus provides about 90% of the extrinsic input

to the cortex!

• However, only about 10% of projections ending in the

cortex are extrinsic, 90% of them are intracortical

connections!

Thalamus and Cortex

• The strong reciprocal connectivity between thalamus

and cortex suggests the presence of thalamocortical

feedback loops

• Baars suggests that what is selected by the NRT are

such loops that are established between specific nuclei

of the thalamus and their cortical projection area

• The NRT will support the establishment of one such

loop, simultaneously suppressing the establishment of

other competing loops

Global broadcasting

• According to Baars, global broadcasting of content could be achieved through

– massive axonal projections from posterior to anterior cortex and between the left and right hemispheres of the brain

– the so-called "tangential intracortical network" (TIN) in the upper layers of cortex, consisting mainly of axon collaterals and dendritic ramifications of pyramidal cells

– perhaps also by diffuse projections to the cortex from the unspecific thalamus (ILN)

Neural global workspace

• The big picture

1. Various streams of input arrive at the specific nuclei of thalamus

2. The NRT selects one input for establishment of a thalamocortical feedback loop with the corresponding cortical projection area

3. The NRT itself is controlled by input from attentional systems, e.g. executive systems in the frontal cortex and anterior cingulate

4. The selected input is globally broadcast via intracortical connections (e.g TIN), and maybe via the diffuse, nonspecific thalamocortical projection system of the ILN

5. Global broadcasting is associated with consciousness

6. The recipients of the global message are various distributed systems in the brain: USMs, memory systems, motivational, emotional systems...

7. The reticular formation controls the overall activity level of the system, it is like a power switch that can turn on or off the workspace. Activating the RF leads to inhibition of NRT and thus opening of the gates

NGW in the theater

• Actors: competing input streams to thalamus and/or thalamocortical loops

• Spotlight of attention: the NRT

• Director, spotlight controller: attentional systems and/or executive systems

• Actor in the bright spot: selected content as represented in thalamocortical loop

• Global broadcasting: distribution of content through TIN etc.

• Audience: USMs, memory systems, motivational, emotional systems...

• Context: various brain systems

• Without equivalent in the theater metaphor: reticular formation

• The RF - thalamus axis supports the global state of consciousness, while the cortex provides its particular content

Integrated Information Theory (IIT)

Giulio Tononi et al. 2004, 2014…

Starting points

Consciousness is both highly

• integrated

• differentiated

Integration: each conscious scene is unified, experienced

from a single point of view

Differentiation: one can experience a huge number of

different conscious states in a short time; the selection of

one single conscious state out of the millions of possible

conscious states constitutes a huge amount of

information, in the sense of reduction of uncertainty

consciousness is informative

Measures of integration and differentiation(derived from information theory)

Integration:

• Mutual information (MI)

• Integration (I)

• Cluster Index (CI)

Differentiation / Informativeness:

• Neural complexity (CN)

Functional clustering

• With these measures in hand, we can now identify functional clusters

within the brain

• A functional cluster is a subset of brain regions that interact much

more strongly among themselves than with the rest of the brain

• Such functional clusters indicate highly integrated systems that are

relatively functionally isolated from other brain parts

• More formally, we can say that a functional cluster has a high

Integration I among its brain regions, but low Mutual Information MI

with the rest of the brain

• We can formalize these criteria by defining a measure called the Cluster

Index (CI) defined simply as I divided by MI:

Functional clustering

Finding highly integrated systems in the brain is important for

consciousness, because consciousness is itself a highly integrated

phenomenon

Functional clusters could be the neural basis of consciousness

Dynamic core hypothesis (...see later...)

Measures of differentiation:

Neural complexity (CN)

• Now we have a measure to identify highly integrated neural processes in the brain.

• But according to ET, "to account for the fundamental properties of conscious experience, we still must find a way to characterize a neural process in terms of how differentiated or informative it is... we do so by considering a measure of neural complexity."

Measures of differentiation:

Neural complexity (CN): Examples

A: normal brain:

high specialization,

high integration

high complexity

B: reduced connectivity:

high specialization,

low integration

low complexity

C: high, uniform

connectivity:

low specialization,

high integration

low complexity

Neural complexity

If this is all too complicated , a good metaphor

is this:

A complex brain is like a collection of

specialists who talk to each other a lot.

Back to consciousness

Remember our starting point:

To find the neural substrates of consciousness we must

1. find neural systems that are both integrated and differentiated

2. develop appropriate measures of integration ( Cluster Index CI) and differentiation ( Neural Complexity CN)

• Now we can define the neural substrate of consciousness as a functional cluster (high CI) with high Neural Complexity

• Later version: A system‘s degree of consciousness can be measured by a measure of its intergrated information content: The Phi function Φ

The dynamic core hypothesis

Only elements of a functional cluster can contribute to

consciousness

A functional cluster can contribute to consciousness

only if it is highly differentiated, i.e. has high complexity

Such a large functional cluster of high complexity is

called a dynamic core

Dynamic core

The IIT hypothesis accounts for the general properties of conscious experience by linking these properties to specific neural processes that can give rise to them." again, the assumption of structural isomorphism

IIT is a process, not a thing

It is not a fixed set of brain regions, but changes its composition over time. The same group of neurons may at one time be part of the dynamic core, thus contributing to consciousness, but not at other times

The composition of the dynamic core may vary between individuals

A dynamic core is largely, though not exclusively, generated with the thalamocortical system, achieving high integration within hundreds of millisecons through reentrant interactions

Evidence for dynamic cores is still largely absent, but a prediction from the hypothesis is that the conscious state of a subject should directly correlate with the neural complexity of the dynamic core

Before we can do empirical research we need methods to estimate neural complexity over short integration periods of hundreds of milliseconds

Qualia

IIT say that every single unified, integrated conscious state that we experience corresponds to a quale (quale=singular of qualia)

Thus, on their hypothesis, every total state of the dynamic core implements a quale

That means, qualia are always defined by the total activity of the core, not just by the activity of some parts of it

E.g. even the "simple" experience of pure red is defined not only by the activity of red-sensitive neurons in the core, but also by the (lack of) activity of green-sensitive neurons, or of auditory neurons, indeed of all other neurons that are part of the dynamic core.

Neural implementation

• only certain brain areas support consciousness at one time, preferentially those in the thalamocortical system

• a certain strength, duration and stability of neural interactions may be a prerequisite of consciousness

• there can be anatomical (e.g. reciprocal connectivity) and functional (e.g. inhibition, desynchronization, neuromodulation) mechanisms that support functional clustering

• Mechanism for integration/binding:

– reentry: signaling parallel loops leading to synchronization

– convergent connectivity (marginal)

• Mechanisms that support neural complexity

– high density connections

– strong local connectivity

– patchy connectivity between neural groups

– many short reentrant circuits

Theories of Consciousness: two examples

Documents

Transcript of Theories of Consciousness: two examples