Default behaviour

Definitions

So what about the default behaviour? First lets define a few terms in order to introduce the default behaviour:

• DNA: a sequence of 0s and 1s, of a given length. Be it interpreted as nucleobases (G, A, T, C) or something else. In a computer it always ends up as 0s and 1s. It could also be multiple sequences, or sequences with branches.
• Initial DNA: the DNA of all your population at the beginning of the simulation, in the case where they all have the same.
• Initial Random DNA: the DNA of each member of your population at the beginning of the simulation, in the case where they all have a randomly generated DNA (and potentially of random length as well).
• Action: the lowest-level element in the simulation defining how a Creature can interract with itself and its environment, including reproducing itself, or doing nothing.
• Behaviour: the logic internal to the Creature (its reasoning) which translates its perception of the environment and itself in a given Action (could be one or more at the same time). An initial population with the same Initial DNA would typically have the same Behaviour.
• Behaviour Generator: a piece of program that generates the behaviour of a Creature based on its DNA. The generator may constantly modify the Behaviour during the life of the Creature.
• Intentional Action: an Action resulting from the Behaviour of a Creature.
• Default Behaviour: the Action made by a Creature when its Behaviour does not lead to any Intentional Action.

In a loosely constraint life simulation, it is highly probable that the Behaviour of the Creatures will sometime lead to no Intentional Action. So a Default Behaviour must be defined, even if it is just "do nothing".for a given time.

Default Behaviour in Evo(rand)

In Evo(Rand) you must define a Default Behaviour. That Default Behaviour can be "do nothing" or any list of Actions each with an associated probability and range, i.e. the distance to the Ground where the Action will be performed. The distance is applicable to the Actions that start with a mouvement to the point in space where the Action will actually be performed.

In Evo(rand) you can define an Initial DNA of 0s, an Initial DNA of 1s or an Initial Random DNA. The probability of having an Initial Random DNA which allows a Creature to reproduce itself is so low, that it is usually not used. Instead I usually use:

• An small Initial DNA of 0s.
• A Default Behaviour that includes reproduction as one of the possible Actions.

In the examples of this article I used:

• An Initial DNA of 20 bytes of 0s.
• A Default Behaviour with 2 Actions, each with a different probability: Move (with range of 0 or 1), and Clone (with range of 0).

Un-suitable Default Behaviour

Similar to the first article, the model used for this experiment is very simple:

• Each Creature has an Energy attribute. The Energy is passed from one generation to the next, i.e. the Energy of a newborn comes from its mother (no sexual reproduction).
• 4 possible Actions: Move, Clone, Nill (do nothing) and Deciding (used as a transitory Action when the Creature is thinking, i.e. deciding what will be the next Action).
• Creatures regularly loose Energy.
• Creatures also gain Energy (potentially more than what they loose), but to avoid exponential increase of the population, less Energy is given when there are many Creatures on the same Ground: the Energy is distributed over the Creatures present on the Ground.

In Evo(rand) it is shown like this (Aging is discussed in this article)

:

As seen in this article, a very simple simulation can lead evolution to select the Default Behaviour as the most fitted survival strategy. Let's explore this a bit more. With a range of 1 in the Move Action (of the Default Behaviour) the Creature can move away from the Ground where it is born. Changing the relative probability of each default Action, we see that:

• A high probability for Clone Action is the best strategy. with a Clone probability of 90%, we have 98% of the population relying on the Default Behavour.

• With a Clone probability of 5%, we have 17% of the population relying on the Default Behavour.

In both cases, the total population ends up being the same: a bit more than 7100 Creatures. This means that over the generations, both populations have found an equal level of fitness, the first relying totally on the Default Behaviour, the second relying partially on the Default Behaviour. But the second populations had to go through an adaptation period to develop an adequate Behaviour. This can be seen, for example, by the bigger DNA in the second population.

So the lower is the Clone probability in the Default Behaviour, the less suited it is. In a world where the newborns are independant and do no rely on the teachings of the parent, the parent can die at birth, passing all its Energy to the child. Reproducing as fast as possible is the best way to populate the world.

Intentional random Action

But what happens if the range of the default Move Action is 0? The Creature is stuck to the Ground where it is born, to reproduce here, and to have its Children all on the same Ground. That does not seem very suitable. And indeed it is not. Here is the evolution with a Clone probability of 5%:

What do we see? A population of hardly 5000 Creatures, with a longer DNA than the above experiments, struggling to survive. Running the simulation longer, the population never reaches 6000.

But if we give the possibility for the Creatures to move to a randomly selected Ground, in other words an Intentional Action to a random Ground, then this is what we obtain:

After an initial transition period, an adapted Behaviour is found: a population of nearly 7100 Creatures with less than 2% of the population relying on the Default Behaviour. The population has the same level of fitness than when relying purely on the Default Behaviour with 90% Clone probability. But this time it is using Intetinal random Actions instead of the Default Behaviour.

My Child died!

In the above experiments, a Creature dies when its Energy is 0. Here we are going to test what happens if we define that a Creature dies when its Energy is below 0.1. It means that, contrary to the previous experiments, if a Creature with a low Energy tries to give birth, it has a high probability that its Child will die, where before only the parent would die.

Here is the evolution with a Clone probability of 5%:

And with a Clone probability of 20%:

In both cases the population is less than before: around 6600 instead of 7100 Creatures. This is normal because Creatures die when their Energy is higher than before. We also see that this aditional constraint make the default behaviour less suitable, although only slightly less.

Conclusion

To avoid having your population relying on the Default Behaviour, make sure:

1. That the Default Behaviour is suitable for your initial population to survive a few generations. It must include Clone Action, but with a probability as low as possible.
2. That the Creatures have access to information that allow them to develop an adapted Behaviour, like the Intentional random Action in the example above.
3. Add complexity in the world that is simulated. The example above is very simple, but it shows that a small added complexity (Creature dying at 0.1 Energy) requires a more complex adaptation: a more complex Behaviour that relies less on the Default Behaviour.