Systems, Man & Cybernetics - April 2017 - 9

phase B for inference, prediction,
scene or music with previously
decision making, and (re)actions
seen examples and concepts. Data
The specific role of
2) a certain number of outliers that
clouds are like clusters but are
the emotions is very
(for good or bad) have been memshape free and form a Voronoi tesorized (mainly subconsciously).
sellation in the feature space [12],
important in regard
There are many studies that
[14]. This process reduces the vast
to understanding
report that, during sleep, people
amount of raw data into a much
(unconsciously) revisit and reorgasmaller number of concepts while
the nature of
nize the raw data, and various outliidentifying possible outliers [16]
unsupervised learning
ers come to their attention [17], [18].
(Figure 2). In real life, examples of
in individuals.
We are aware that better results may
data clouds can be images of or
be achieved using electroencephasounds from cars, dogs, or any
lograms or electrocardiograms,
other sense stimuli we encounter
which are more informative, but we
in the course of our lives.
deliberately do not consider these in our aim for nonintruAccording to the CyberMind architecture, at the secsiveness or minimal intrusiveness. Indeed, [19] demonstratond layer of the hierarchy, these data clouds that repreed that different emotional states such as happiness,
sent semantic concepts can be linked to rules, story
sadness and grief, anger, or a neutral state can be recoglines, and graphs, and labels can be optionally assigned
nized based on physiological cues such as heart rate, skin
to them. Our hypothesis is that by memorizing as landconductance, and blood oxygen level.
marks only the concepts (i.e., the focal points of the data
An example of data from a real experiment conducted at
clouds) plus the outliers identified, instead of the huge
Carlos III University, Madrid, Spain, in September 2015 is
amount of raw data people encounter in their lives, they
depicted in Figure 3, where two persons' heart rate and skin
are able to cope with this real-life big data problem. In
conductance were measured on one day in response to the
this way, individuals transform this overwhelming
same sequence of images and audio. To make the measures
amount of heterogeneous data into a manageable and
comparable, the values are centered using the subtracted
individually specific subset of
mean per person per day, which also applies to all the fig1) data clouds forming rules, labeled and organized into
ures in this article. This demonstrates the importance of the
other structures for successive use in the conscious

Raw Data

Reasoning and
(RE)Action

Extract Rules from Data

Decision Making

Visual

Heterogeneous

Semantics

Emotion 1

Concept A

Concept B
Concept O

Decision
Making

Concept B

(Re)Action

Concept A

...
...
Audio

Emotion P
Concept G

Goals

Concept R
...

Concept K

Cognitive Feedback

Figure 2. a layered hierarchy of making sense of data (Cybermind architecture).

Ap ri l 2017

IEEE SyStEmS, man, & CybErnEtICS magazInE

Table of Contents for the Digital Edition of Systems, Man & Cybernetics - April 2017