Emergence of Basic Value Orientations Anticipation and Individual Differences

Since the animat's training depends on a number of random factors, each animat develops a different cognitive system (classifier set and decision rules), even though final performance may be similar. In order to show general tendencies despite these individual differences, mean values over large populations were obtained. These dealt with (1) results of the training process in two (otherwise identical) environments having different variety and variability, and with (2) performance of animats after transfer from their training environment to environments challenging them with more variety, or variability, or change.

One remarkable result from these experiments is that individuals achieve comparable performance in a given training environment with very different cognitive systems, and in particular with different orientor emphasis. While this may not provide any particular advantage in the training environment, it may provide distinct fitness advantages if the animat is moved to a different environment. Three particular types of individuals stand out: generalists (type F) stressing freedom of action, specialists (type E) focusing on effectiveness, and cautious type (type S) emphasizing security. Figure 4 shows the different orientor stars for these three types.

The ability to develop a cognitive system reflecting its environment makes the animat a suitable vehicle for investigating goal function emergence and value orientation. Genetic algorithms are very effective processes that seem to capture the essentials of real processes found in the evolution of organisms and ecosystems. In the animat, they very effectively build up a cognitive model (or goal function) that enables anticipatory behavior; since rewards flow back to earlier rules leading to later pay-off, the activation of the initial rules in a pay-off chain means that the system suspects possible pay-off and anticipates the near future, that is, it has an internal model of the results of its actions under the given circumstances.

In the animat experiments (and similarly, in real life), implicit and (more or less) balanced multidimensional attention to the basic orientors emerges from the simple one-dimensional mechanism of rewarding success in the given environment. Thus, in the course of its evolutionary development in interaction with its environment, the system evolves a complex multidimensional behavioral objective function from the very unspecific requirement of fitness. Conversely, this also means that balanced attention to the emergent basic orientors is necessary for system viability and survival - they would not have emerged unless important for the viability of the system.

Balanced attention still leaves room for individual differences in the relative emphasis given to the different orientors. Individuals belonging to the populations used in the animat experiments evolve significant differences in value emphasis (e.g., specialist, generalist, cautious type). These individual variations, while not significantly reducing performance in the standard training environment, provide comparative advantage and enhanced fitness when resource availability, variety, or reliability of the environment change. They also result in distinctly different behavioral styles. However, pathological behavior will follow if orientor attention becomes unbalanced (e.g., dominant emphasis on one orientor).

Training of animats in different environments, the performance of animat individuals in environments that differ from their training environments, and the simulation of adaptive learning in a changing environment, lead to some general conclusions that are in full agreement with everyday observations and general systems knowledge:

• Generalists have a better survival chance than others if moved to an environment of greater variety.

• Cautious types have a better survival chance than others if moved to a less reliable environment.

• Training in more unreliable and/or more diverse environments increases satisfaction of the security and/or freedom of action orientors at the cost of the effectiveness orientor.

• Training in an uncertain environment teaches caution and improves fitness in a different environment.

• Learning caution (better satisfaction of the security orientor) takes time and decreases effectiveness, but increases overall fitness.

• Investment in learning (exergy cost of learning in the animat) pays off in better fitness; the learning investment is (usually) much smaller than the pay-off gain.

Animat individuals not only develop behavior that can be interpreted as intelligent, they also develop a complex goal function (balanced attention to basic orientors), or value orientation. Serious attention to basic values (basic orien-tors: existence, effectiveness, freedom, security, adaptability, coexistence) is therefore an objective requirement emerging in, and characterizing self-organizing systems. These basic values are not subjective human inventions; they are objective consequences of the process of self-organization in response to normal environmental properties.

Worm Farming

Worm Farming

Do You Want To Learn More About Green Living That Can Save You Money? Discover How To Create A Worm Farm From Scratch! Recycling has caught on with a more people as the years go by. Well, now theres another way to recycle that may seem unconventional at first, but it can save you money down the road.

Get My Free Ebook


Post a comment