Transcript build a brain – The social aspect in evolution

How to build a brain – The social aspect in evolution
TECHNISCHE
UNIVERSITÄT
DRESDEN
Hagen Lehmann† & Tino Schmidt‡
†University of Bath, Artificial models of natural Intelligence
‡Technische Universitaet Dresden, Institut fuer Allgemeine Psychologie
INTRODUCTION
The question how the human brain works can be approached from different
angles. Scientists from disciplines like biology, neuroscience, psychology and
philosophy try to solve this problem.
The human brain is like all nervous systems an information-processing system.
Therefore the comparison of information-processing models from computer
science with results from empirical neuroscience, describing the evolution of
neural networks as well as complex social behavior can be very useful for the
understanding of this complex problem.
The high degree of complexity of social behavior on the neural level is the
reason why science was not yet able to find satisfying answers to most of the
questions concerning the meaning of neurophysiological correlates found in the
human brain. This complexity evolved over millions of years to enable humans
to interact in a purposeful way with a constantly changing and noisy
environment. The functions of the human brain are physical manifestation of
actual behaviors performed by an individual or a group of individuals. These
behaviors evolved due to challenging situations in the environment and the drive
to survive and reproduce.
Therefore each species has its own behavior patterns while interacting with the
environment. Social behaviors as well as individual behaviors can be seen as the
connection between brain activity and environmental challenges. One possible
approach to the question “How to build a brain?” is to create a model which will
help us to understand the complex social interactions between individuals, how
individuals plan and how decision making influences the action selection
process.
The understanding of social behavior will help us to understand the complexity
involved in all higher cognitive brain functions. It is not only important to
understand how the human brain works, but also why it works.
Real primate societies:
Tolerance is one of the most basic forms of conflict resolution in social species
(de Waal & Luttrell, 1989). It is for example expressed when a dominant
animal allows a subordinate to take advantage of a resource in its presence.
For species with low levels of tolerance most conflicts are unidirectional:
subordinate animals make no effort to retaliate for an attack by a dominant, and
would almost never attack a dominant themselves. Conflicts tend to be
infrequent, but when they occur, they are highly intensive (e.g. biting) and
reconciliation behavior afterwards is rare.
In species with high levels of tolerance conflicts are more frequent and
bidirectional, and the majority of these interactions is being met with protests
or counter-attacks, but the intensity of aggression is low.
HEMELRIJK‘S DOMWORLD
Figure1:
Hemelrijk
implemented
grouping rules and rules for interactions
in her model. Depending on their
internal state (in the model represented
by the dominance value) the agents will
engage in dominance interactions. This
will change the dominance value of the
agents, depending on the result. The
agent which looses the interaction will
then flee from its opponent. To maintain
the group structure she implemented a
search procedure during which each
agent, which is not involved in a fight
searches for its nearest partner in the
group.
ALTERNATIVE MODEL
WHY AN EVOLUTIONARY PERSPECTIVE?
Before creating a complex artificial system to simulate functions in natural
processes it is important to understand these functions and their meanings in
nature. This will help to avoid redundancy and enable us to create ecological
valid, working systems. Especially since the goal is not to rebuild the physiology
of the human brain but to understand and simulate its functions. The
evolutionary perspective can help to achieve this, since it is a top down
approach.
One of the main problems in AI is that agents have to recognize the
circumstances under which certain rules, on which these agents are based, have
to be applied to solve a task. These agent-situation interaction patterns are called
affordances in ecological psychology (Greeno 1994).
Animals and humans are very good in recognizing such affordances, which is
why they are flexible and efficient in their reactions.
COMPARING NATURAL PROCESSES WITH THE
MODEL
Social interactions are the central element for the understanding of the
evolutionary mechanism of higher cognitive functions. An ecological valid model
which can help us to understand the human brain will therefore have to include
functions to simulate social behavior found in real animal societies.
Thierry et al. (2004) propose that the genus macaca is a particularly good model
for studying primate social organization. There are twenty-one known macaque
species. From a genetic perspective they are very similar, but they differ widely in
their social interaction. Their styles of social interaction range from egalitarian to
despotic and their repertoire of social behaviors differs according to that.
Model primate societies:
As an example for current models of primate social interactions we choose
Hemelrijk`s DomWorld (Hemelrijk 1999, 2002), because it is the most widelypublished agent based models in biology.
Hemelrijk was able to simulate, with different additions to her basic model
(Figure1), phenomena like female dominance and sexual attraction (Hemelrijk
2002). Her model is supported by results of her own field research. But it is too
simple (it does not include any higher cognitive functions) to be useful to help to
answer the question of “how to build a brain”.
CHECK
INTERNAL
STATE
LEVEL
OF DESIRE
-
THRESHOLD
LEVEL OF RISK
+
SEARCH
FOR FOOD
MOVE TO
FOOD
EAT
FOOD
MODIFICATIONS
OF WEIGHTS
LEVEL
OF DESIRE
CHECK
INTERNAL
STATE
Our model focuses on different functions of social behavior. We integrate the
most basic variables of motivation. Social interactions are based not only on the
mechanisms underlying tolerance but depend on a variety of different
motivations. This system of different motivations is hierarchical and includes
specific processing units to act in a situation with the ability to stay flexible to
choose other goals, in case the situation changes. The mechanism operates on a
top down level. A full expression of all included processes would be very
complex, therefore we only picked the eating behavior to illustrate the sequence
of actions.
REFERENCES
- Greeno J.G. (1994). Gibson's affordances — Psychological Review 101(2), p. 336-342
- de Waal, F. B. M. and Luttrell, L. (1989). Toward a comparative socioecology of the genus
macaca: Different dominance styles in rhesus and stumptailed macaques. Am. J. of
Primatology, 19:83–109.
- Hemelrijk, C. K. (1999). An individual-oriented model on the emergence of despotic and
egalitarian societies. Proc. of the R..Soc.: Biol. Sc., 266:361 369.
Hemelrijk C.K. (2002). Self-organization and natural selection in the evolution of complex
despotic societies. — Biol. Bull. 202, p. 283-289.
-Hemelrijk C.K. (2002). Despotic societies, sexual attraction and the emergence of male
”tolerance”: an agent-based model — Behaviour 139, p. 729-747.
- Thierry, B., Singh, M., and Kaumanns, W., editors (2004). Macaque Societies: A Model for
the Study of Social Organization.Cambridge University Press. de Waal, F.B.M. (1989).
printed by
www.postersession.com