In animals, this ability to abstract concepts from
perceptions should be given by the plasticity of the
animal’s neurons. This implies that some concepts
might be innate, determined by the prenatal wiring of
the neurons, which is dependent on the genome (and
perhaps also proteome). How and why these innate
mechanisms evolve, including the ones allowing the
concept abstraction, are still open questions. One
could argue that it is an advantage to have them, but
recent studies (e.g. Alexander, 2001) have put a
question on wether natural selection is the only (or
even in some cases the main) driving force of natural
evolution. Here we will not discuss this issue, just
assuming that this ability has been already acquired.
We will only say that work in evolutionary robotics (see
Harvey et. al., (1997) and Gomi (1998) for reviews)
might lead to answers for these questions.
So, we can say that an animal is able to abstract
regularities from its environment. We will not be aware
of them if the animal does not exploit the acquired
concepts in its behaviour. But if the animal manipulates
the acquired concepts in order to adapt to its
environment, we can say that the animal has abstracted
a logic ofitsenvironment. Knowledge of its environment.
As we stated, this is dependant on the observer, since
we believe that knowledge is an emergent property of
a cognitive system, not an element2. We can also say
that the “proper” use of concepts gives them a certain
meaning, grounded through action.
If we are searching for explanations of our logic, the
logic of our environment, then we should take other
issues into account. First of all, the fact that we live in
a society3, which is shaped by us, and shapes us. We
have a language, which allows us to externalize and
share our concepts. This allows us to have an access to
the concepts of others, enlarging our knowledge.
Language and human thought are so interrelated,
interdependent, and internecessary, that some people
even seem to have forgotten that they are different
things4. We believe that language is also necessary for
complex manipulation of concepts (Clark, 1998), and
since an individual can develop a language only in a
society (Steels and Kaplan, 2002), it is only in a society
that an individual can develop higher cognition
(Gershenson, 2001), as it seems has been in nature
2We should also be careful with language games,
such as “Does a tree knows when spring came because it
blossoms?”.
3The social factor has been proposed to be also
responsible for the evolution of our “big” brains (Dunbar,
1998).
4Though thought seems to have all the properties
of a language...
(Dunbar, 1998). Through generations, a culture is
formed, accumulating past experiences.
Summing up, the epigenetic stageswe should follow
to reach knowledge from behaviour, should be:
1. concept abstraction.
2. grounding of concepts though action.
3. sharing of concepts through social interactions
(language).
4. manipulation of concepts (logic5).
5. evolution of concepts (culture).
Note that knowledge is not acquired only until
completing all the stages, but it is developed gradually
with every stage. And we will not say that our
knowledge cannot be improved as well, i.e. we can
always add more stages. Also, steps 3 and 4 could
exchange places. In fact, there have been models and
theories which address most of the steps described
above, but separately (e.g. Scheier and Lambrinos,
1996; Gardenfors, 2000; Zlatev, 2001; Cangelosi and
Parisi, 2001; Prince, 2001; Steels and Kaplan, 2002),
and thus they answer only partially the question of how
knowledge could evolve from adaptive behaviour. Of
course, it is necessary to have such models and theories
before attempting to model all the path, and they all
can be considered as BBKS or BBKS theories.
These requirements for acquiring higher order
cognition seem quite sensible, and have been proposed
with similar approaches (e.g. Kirsh, 1991; Steels, 1996;
Clark, 1997; Balkenius et. al., 2001; Zlatev 2001;
Gershenson, 2001; Steels and Kaplan, 2002).
Another way of convincing ourselves to follow this
path is to analyze it backwards: if we take from humans
each of the stages described, how do our knowledge
would be diminished? Without culture we would not be
able to accumulate knowledge from generation to
generation, and only by our physiological abilities we
would be at a level even lower than social primates.
Without being able to manipulate concepts we would
not be able to make inferences nor predictions.
Without a language we would not be able to learn what
other individuals have learned, and we would be
restricted to our individual learning. Without concepts,
we would just have reactive behaviours, without the
possibility of integrating our sensory experiences to
produce complex behaviours. But of course, for using
these concepts they need to be grounded.
We believe that following this approach, consistent
with Epigenetic Robotics (Balkenius et. al., 2001;
Zlatev, 2001), we will be able to build systems which
develop their own logic, consistent with their
environment, which will be able to do reasonings in the
sense a KBS does. Of course, this will not replace KBS,
since their manageability at a knowledge level is much
higher than it would be in BBKS (as noted in
5We mean logic as a tool, not logic as a science.