the members of a social group.
-Cognitive processes may involve coordination
between internal and external (material or environ-
mental) structure.
-Processes may be distributed through time in
such a way that the products of earlier events can
transform the nature of later events.”
The study of individual consciousness has again become
academically respectable, after nearly a century of ideolog-
ically enforced unpopularity, and Baars’ Global Workspace
Theory (GWT), (Baars, 1988, 2005; Baars and Franklin,
2003) is the present front runner in the Darwinian competi-
tion between theoretical approaches (Dehaene and Naccache,
2001). Wallace and colleagues (Wallace, 2005a, b, 2006, 2007;
Wallace et al., 2007; Glazebrook, 2006) have developed a fairly
detailed mathematical model of GWT, using a Dretske-like in-
formation theory formalism (Dretske, 1981, 1988, 1993, 1994),
extended by tools from statistical physics, the Large Devia-
tions Program of applied probability, and the topological the-
ory of highly parallel computation.
Institutional cognition involving simultaneous, multiple
global workspaces is, however, a far more complex and var-
ied phenomenon, significantly less constrained by biological
evolution, and far more efficient in many important respects
(Wallace et al, 2007).
As the cultural anthropologists will attest, the structures,
functions, and innate character of institutional cognition are
greatly variable and highly adaptable across social and phys-
ical geography, and across history. Individual human con-
sciousness, by contrast, remains constrained by the primary
biological necessity of single-tasking, leading to the striking
phenomenon of inattentional blindness (IAB) when the Rate
Distortion Manifold of consciousness become necessarily fo-
cused on one primary process to the virtual exclusion of oth-
ers which might be expected to intrude (e.g. Mack, 1998;
Dehaene and Changeux, 2005; Masuda and Nisbett, 2006; Si-
mon and Chabris, 1999; Simons, 2000; Wayand et al., 2005).
Generalizing a second order mathematical treatment of
Baars’ Global Workspace model of individual consciousness
to organizational structures, Wallace et al. (2007) suggest
the contrasting possibility of collective multitasking, although
that is mathematically a far more complicated process to ana-
lyze and describe. That work uncovered an institutional ana-
log to individual inattentional blindness, and additional fail-
ure modes specific to the complication of multiple workspaces.
Here we will recapitulate that analysis from the perspec-
tive of machine design, specifically invoking devices having
as many global workspaces as a large, highly capable human
institution, but operating with characteristic times similar to
individual consciousness - a few hundred milliseconds.
By contrast, small, disciplined groups involved in hunt-
ing, combat, firefighting, sports, or emergency medicine, may
function in realms of a few seconds to a few minutes. They
are adaptive and ‘self programming’ in the face of dynamic
challenge. The appropriate time constant probably follows
something like a logarithmic scaling, i.e. τ x log[N], where
τ is the time constant and N the number of individuals or,
more likely, workspaces, which must intercoordinate across
the organization.
Similarly designed fast, self-programming, highly adaptive,
multiple-workspace, machines could be expected to do far
more than just play variants of chess well, although their
canonical and idiosyncratic failure modes would not be stabi-
lized by the tens of thousands of years of cultural and ‘market’
selection pressures which have come to structure the various
cognitive human institutions. Creating an ‘ethical’ stabilizing
context for them seems a difficult challenge.
We begin with a review of recent work on individual con-
sciousness, as a kind of second order iteration of simple cog-
nition, and then examine the nontrivial extensions needed to
describe highly parellel machines having multiple workspaces.
We will then attempt to adapt the approach to multi core ma-
chine programming.
INTRODUCTION TO THE FORMAL THEORY
1. The Global Workspace model of individual con-
sciousness
The central ideas of Baars’ Global Workspace Theory of
individual consciousness are as follows (Baars and Franklin,
2003):
(1) The brain can be viewed as a collection of distributed
specialized networks (processors).
(2) Consciousness is associated with a global workspace in
the brain - a fleeting memory capacity whose focal contents
are widely distributed (broadcast) to many unconscious spe-
cialized networks.
(3) Conversely, a global workspace can also serve to inte-
grate many competing and cooperating input networks.
(4) Some unconscious networks, called contexts, shape con-
scious contents, for example unconscious parietal maps mod-
ulate visual feature cells that underlie the perception of color
in the ventral stream.
(5) Such contexts work together jointly to constrain con-
scious events.
(6) Motives and emotions can be viewed as goal contexts.
(7) Executive functions work as hierarchies of goal contexts.
Although this basic approach has been the focus of work by
many researchers for two decades, consciousness studies has
only recently, in the context of a deluge of empirical results
from brain imaging experiments, begun digesting the perspec-
tive and preparing to move on. Currently popular agent-based
and artificial neural network (ANN) treatments of cognition,
consciousness and other higher order mental functions, to take
Krebs’ (2005) view, are little more than sufficiency arguments,
in the same sense that a Fourier series expansion can be em-
pirically fitted to nearly any function over a fixed interval
without providing real understanding of the underlying struc-
ture. Necessary conditions, as Dretske argues (Dretske, 1981,
1988, 1993, 1994), give considerably more insight.
Wallace (2005a) addresses Baars’ theme from Dretske’s
viewpoint, examining the necessary conditions which the
asymptotic limit theorems of information theory impose on
the Global Workspace. A central outcome of that work is