The Sockmonster

Is there any evidence to suggest that development is adaptive? Is it reasonable to believe that development, as a process, will behave differently in different environments and that this adaptive property of development will produce environmentally adapted phenotypes? For example if two genetically identical plants develop in two environments, one cold and one warm, will the product of development produce two different phenotypes, one that is adapted to colder environments and one that is adapted to warmer environments? If there is evidence to suggest this then how large can the change (in this case what range of temperature) be before development breaks down? And how does this evolve?

If this is reasonable then would it be possible to design a developmental robot instead of pre-designing a robot for an environment, design a developmental program capable of adapting to the specifics of the environment.

- – Tuesday 26th Feb – -

As it turns out there is lots of evidence for adaptive development, in fact its more common than not (Gilbert, 2001)! There is good evidence of adaptive development in both plants and animals, Jewelweed, for example, will develop elongated stems if grown in shade and short stems if grown in light. This is because in shaded environments the plant can increase its light capture by growing tall, whereas if the light is good it is more energy efficient to grow short stems (Dudley & Schmitt, 1996). Most of the support for adaptive development comes from studies on plants. I’ve also found good evidence of adaptive plasticity (as it’s known in biology) in animals, Wood Frog tadpole, for example, will develop a deeper tail musculature if they develop in tanks which contain predators (in a cages so they can’t eat the tadpoles). The deeper tail musculature allows the tadpoles to turn faster and swin faster, and the more predators in the tank the deeper the musculature. What structures of genes result in adaptive plasticity? Can we use these gene structures to develop robots that develop adaptively?

If novelty is produced from agent-environment interaction duringdevelopment, and so not written into the genes per se, then how is the novelty inherited if it is selected? Do other processes evolve which control the agents environment therefore increasing the likelihood of the agent developing in the environment of its parent? Is this what behavioural and cultural inheritance systems do, control the environment? And can these systems evolve with the agent?

I have just spoken to Luc about the last post (part 2), and after some discussion he managed to persuade me of his argument. Basically it comes down to this, before investigating something (in this case development) one doesn’t need to define exactly what it is they are investigating, science is not about definitions! The process of investigation creates the definition. If one defines the target of investigation prior to experimentation, and then conducts experiments based upon that definition, then what scientific inquery was there to validate the definition in the first place? The answer is none!

Basically this means that I don’t need to fully define development in order to start the project on critical periods. I need to find a feature of development that is widely regarded as being a necessary component of development, and build a simulation around this feature. If something arises from the experiments on the simulation that helps us understand development then this goes someway to defining development.

To summarise: I was kind of working backwards, trying to define something before experimentation, I was taking a very engineering view point; define something and build it. This is not necessarily a scientific approach.

I’ve had a think about the issue in part 1 and have spoken to a couple of people to get their thoughts on the matter. I think the list below goes some way to defining development.

• Most importantly development is change.
• The change must be toward some desirable state, if the change is detrimental it isn’t development.
• The change must be internally driven, change from external forces isn’t usually seen as development.
• The change isn’t necessarily in response to external factors but it can be.
• It is unidirectional, a system doesn’t undevelop.
• To separate development from learning; development works at a different time-scale to learning and is generally less flexible.
• Stages are common, stages condition further development to other stages.

I’m sure there are some more things that define development, but I think that the simulation for the critical periods project should exhibit at least some of these points.

I’ve been looking at a project I completed a few months ago on critical periods in development. Luc thinks that there is potential in this project and after talking to him about it I can see what he’s getting at. To turn this project into a publishable paper I think it necessary to re-implement the simulation so that I can run many more runs and know for absolute certainty what is happening in the code (the old code is messy and likely contains bugs, I wouldn’t feel comfortable writing a paper around a badly implemented simulation). But this has raised some interesting questions.

The old simulation used a genetic regulatory network as the developmental system to ‘grow’ a shape of cells on a 2D grid. Why is this system (the GRN) developmental? Why not use any dynamical system in its place? Why can’t I replace the GRN with a simple cellular automaton? These three questions are really addressing the same issue, “What characterises a developmental system? What sets development apart from other processes such as self organisation or learning?”

I’m sure there is a lot of literature on this but I haven’t come across (or at least I can’t remember coming across) any that defines a developmental system. This is really very important, if I am to study critical periods in development I must justify the system I implement as developmental. So I’m going to go and think about this and hopefully come up with some definition of a developmental system.

This is something that I’ve been thinking about for a little while but I’m not sure if it’s worth following up. From reading a few Neuroscience books I’ve noticed that Neuroscientists categorise the brain into function and/or structure, the cortex can be seperated into many areas (as many as 52 if you beleive Broddman) based on the layering of the cortex (the structure of neural connections) and these different areas seem to be attributed to different tasks. Is it possible that some structures of neural connections are better suited to performing certain tasks? Are the layers conditioned by the stimulus they recieve or are they predetermined by evolution to perform some function?