Systems, Man & Cybernetics - July 2015 - 14

den neuron is aware (its output is
nonzero). The output neuron is also
aware for the input (-1,-1), although
one of its direct inputs is not clearly
SSSs for Universe
defined. For this network, the first
hidden neuron defines a concept
that contains at least (-1, 1). It may
or may not contain (-1, -1) and
(1, 1). The second hidden neuron
SSSs for Nations
clearly defines a concept that contains only (1, -1). The output neuron clearly defines a concept that
contains (1, -1) and (-1, 1). Thus,
using the three-valued logic, it is
possible to understand the behavior
of each component. If we can design
SSSs for Organizations
a three-valued logic aware system
to emulate the decision-making process of the brain, we may explain
the mechanism of tacit awareness.
SSSs for Individuals
In a three-valued logic aware system, if a subsystem is not aware for
Figure 2. The hierarchical SSSs for individuals, organizations, nations, and
a given input, its output is zero, and
the universe.
this subsystem will not contribute
anything
for
its
successor
(node) to make a decision. The
a general subsystem, we can use a support vector regressubsystems never try to make unclear decisions. This
sor (SVR). In this case, the aware function is the function
can be a key point of awareness theory. That is, even if
defined by the SVR; Tl and Tu can be -1 and +1, respecsome subsystems are not fully aware, the whole system
tively; and the unaware region is the margin.
can still make correct and definite decisions if some
Note that the inputs of a subsystem may come from
other subsystems can provide enough information. That
some physical sensors or from some other subsystems in
is, someone else will fix the problem can be a principle
the same system. In the former case, we can use a neuron
of awareness.
or an SVR as defined above to convert the analog data into
three-valued discrete ones. Without loss of generality, we
Computational Awareness
can assume that all the subsystems in an aware system
in Human Societies
have only discrete inputs. With this assumption, we can
In the "Awareness: Sometimes an Enigma'' section, we disdefine the aware function using a logic formula, and the
cussed awareness in two senses: 1) explicit awareness and
value can be calculated, say, based on Kleene's three2) tacit awareness. Awareness is defined in Cambridge
valued logic [30]. This type of aware system can be interDictionaries Online as "knowledge that something exists,
preted directly based on the three-valued logic.
or understanding of a situation or subject at the present
In practice, however, it can be computationally diffitime based on information or experience." This definition
cult to obtain a logic formula directly. To solve the probcovers only explicit awareness. For tacit awareness, full
lem more efficiently, we can train an analog aware system
understanding is not needed.
first using some existing algorithm and then analyze the
In recent years, we can find many application examples
concept defined by each subsystem. To illustrate the idea,
of CA, such as location aware, context aware, explanation
let us consider a simple example. Suppose that the aware
aware, wisdom aware, mood aware, power aware, energy
system is realized by a three-layered feedforward neural
aware, hardware aware, user aware, life-cycle aware, prinetwork. Each neuron is a prime aware system or AU. We
vacy aware, green aware, cluster aware, QoS aware, fault
can train the network using the backpropagation algoaware, network aware, platform aware, copy aware, people
rithm and then discretize all neurons using Tl ^= - 0.5 h
aware, situation aware, activity aware, client aware, emoand Tu ^= + 0.5 h . Suppose that the problem to solve is the
tion aware, and so on. Some realizations of CA take into
XOR problem. Table 1 shows one of the results.
consideration the first principles of the operation of comFrom Table 1, we can see that, for any input, we
puter systems-for example, developing cluster computing
can easily interpret the decisions made by all neurons,
considering energy condition. In these CA applications,
including both hidden and output neurons. For instance,
the knowledge in a specific field is taken into account in
when the input is (-1, -1), although the first hidden neutheir computation. To realize social awareness, we should
ron is not aware (its output is zero), the second hid14

IEEE Systems, Man, & Cybernetics Magazine July 2015

Table of Contents for the Digital Edition of Systems, Man & Cybernetics - July 2015