Global Workspace Theory

Global Workspace Theory (GWT) is a conceptual framework proposed by Bernard Baars that rather elegantly explains consciousness, at least its functional aspect, leaving aside the phenomenal epistemically privileged sort, in a manner that is intuitive and supported by the large and growing body of neuroscientific knowledge. I have been slowly gravitating towards this theory as my own knowledge in this field has grown and I have come to accept it, in its basic form, as valid for the time being. The fundamental tenet of GWT is that the computational structure of the brain organizes around local so called codelets, defined as small networks of neurons involved in discrete aspects of processing running in parallel, and that these codelets compete to project their message into working memory, at which point the message is said to become conscious and is then broadcast globally in order to recruit more codelets for additional processing. This is an intuitive conceptualization of consciousness as it gives a mechanism by which the contents of consciousness are reportable, can inform voluntary actions, and are self-referential. And, perhaps the most appealing aspect of this theory is its ability to explain the unity of consciousness as its contents stream serially through the mind by positing the flow of processed information through the working memory regions of the prefrontal cortex.

1) What governs the competition of the codelets and what determines the victors? 2) Are voluntary actions initiated in the working memory or is the ‘agency module’ distinct from the ‘consciousness module’? 3) Where does attention fit into this picture?

These are some of the questions that initially seem to have remained ambiguous and require further research. But I will attempt to address them in what follows:

1) The battle to reach the working memory can be resolved in several manners: a) on the basis of the intensity of the signal, i.e. a purely computational resolution of the conflict by means of summation of neural inputs. b) a higher level decision based on the goals and desires of the agent whereby inputs are filtered by their relevance. c) the outcome may be stochastically determined by chance co-incidence of the temporal patterns of the afferent signals. d) the salience or noiselessness of the signal may be the determinant of its success in projection into working memory.

2) My intuition seems to prefer the notion that consciousness and volition are conceptually distinct and thus that there must be a like functional distinction that needs to be drawn in the framework under current consideration. It seems to be asking too much of the working memory area of the brain to be simultaneously responsible for our awareness of the world AND our ability to willfully act upon it. While both of these functions probably reside in the prefrontal cortex, I maintain that they must remain strictly separate and acting independently, albeit with strong interconnections. Indeed, such a strong connection is precisely what allows the contents of consciousness to inform voluntary actions, namely through the global broadcasting of any messages that manage to battle their way into the working memory core, i.e. into awareness. The volitional core of the prefrontal cortex – we might call it the executive center – as compared with the working memory core, has a decidedly different role to play. Whereas the former is merely concerned with transiently retaining a faithful depiction of data streaming into it, the latter issues commands that are optimized to bring about a future which most closely fulfills the agent’s desires and goals. As such, it is very difficult to imagine a single processing unit that instantiates both functions simultaneously.

3) There are two forms of attention: endogenous and exogenous. The difference lies in the directionality of the cueing that is taking place, in other words whether it is top-down attention (issued from the executive center) or bottom-up (based on salience or other low level perceptual or semantic properties). There is much evidence for the notion that ‘attention is the gateway to awareness’ which means that attention facilitates a datum’s becoming conscious, i.e. projecting into working memory according to GWT. So, this offers a possibility to interpret my answer to question 1, and particularly answers b and d, as instances of attention flowing from the top down in the former case and from the bottom up in the latter. That is, when the executive center is actively monitoring its afferent data for something that is of interest because it impinges on the agent’s goals, this is facilitated by a filtering through the multitude of signals encroaching on the working memory and selecting those which have some bearing on the matter of interest. Alternatively, when attention flows from the bottom up, this is driven by signal noise or salience attributes of the raw neural activity, certain configurations of which are more amenable to processing by the brain and hence to winning the competition for ‘fifteen minutes’ in the spotlight of the working memory.

There is much more thinking through this model that is needed in order to flesh it out and complete this simple and elegant sketch of the emergence of consciousness out of a galaxy of biochemical computers called neurons. As of this moment, this framework appeals very strongly to my gut instincts regarding the brain and its circuitry. Further issues that need to be considered: what is the role of recurrent networks and signals, and specifically the finding that the recurrent signal to V1 was necessary for visual awareness? What is the role of synchrony of neural signals, which several experiments have correlated with perception? What are the necessary and sufficient conditions for the arising of a conscious percept? What does the neuropsychological data have to say regarding lesions of the working memory core of the brain?

Leave a Reply

Your email address will not be published. Required fields are marked *