The Life & Mind Seminar Network

Including Constitution in Models of Behaviour

Posted in Seminars by matthewegbert on August 13, 2009

Thank you to everyone that showed up for my presentation yesterday. I appreciated the comments, questions and criticisms that were raised at the end of the talk. I thought I would try to summarise some of these comments and my responses to them here.

Also, here is the paper which the talk was based upon.

Egbert, M., Di Paolo, E. A. and Barandiaran, X. (2009) Chemo-ethology of an Adaptive Protocell: Sensorless sensitivity to implicit viability conditions in Proceedings of the Ninth European Conference on Artificial Life, ECAL09, Budapest, September 13-16, 2009, Springer Verlag. (forthcoming)

Comment #1: “How are you distinguishing between behaviour and constitutional processes? Why don’t you consider processes of self-constitution to be behavioural processes?”

This question likely comes from early in my talk where I speak about how models tend to either concentrate on `behavioural’ or`constitutional’ processes. I don’t think we need a formal definition of `behaviour’ or self-constitution to see that different types of phenomena are being modelled in, for instance, Anil Seth’s model of action selection and Varela, Maturana and Uribe’s model of an autopoietic system.

The former is inspired by one type of biological phenomena (I called it behaviour) and the latter is inspired by a different phenomena (autopoiesis, or self-maintenance). It is also clearly the case that in these two models (and many others) the two types of phenomena are not included in the same model. Typically in models, if type-a phenomena (what I called behavior) is in the model, then the type-b phenomena (what I called constitutional processes) is absent and vice versa.

One of the points I tried to make in the talk is that this division (made by modellers that choose to include only one of these phenomena rather than including both) is sometimes misguided. I mentioned as I was describing the first scenario that the behaviour could not occur without the metabolism, and that the metabolism could not continue without the behaviour. This chicken and egg cycle of contingency suggests that indeed the behavioural and constitutional (metabolic) processes are highly integrated (at least in this scenario). What we have thought of as behaviour can be thought of not as separate from metabolism, but as an extension of metabolism. This raises questions, at least in my mind, about modeling behaviour without including metabolism (in a non-trivial manner).

So, returning to the comment, perhaps the best response to the question “Why don’t you consider self-constitutional processes to be behaviours?” is that I do! At least in the sense that I think that self-constitutional processes play a vital role in their interaction with other behavioural processes. I am trying to blur the distinction between behaviour and metabolism, not emphasise it. The distinction is one that has already been made, perhaps unintentionally, in the modeling community by the rare incorporation both phenomena in the same model.

Comment #2: “Why are you looking at a system where the behaviour is so directly integrated with the autopoiesis? Surely the more interesting situation in when the two phenomena are at least more disjointed, or perhaps even in conflict?”

There are a variety of ways that behaviour and constitution can be related. I chose to study what I believe to be one of the simpler situations first — a relationship between behaviour and constitution that I think is most likely to come about in early or proto-life organisations: a relationship by which behavioural mechanisms serve autopoiesis. Audience participants gave examples of behaviours that do not serve system health, but serve some other ‘motivation’, such as sexual drive, drug addiction, etc. I like the idea of studying these kinds of systems in which behaviours are conflict with the constitution of the behaver, but as a first exploration in this direction, I thought it made more sense to start with what seems more typical rather than with the exception.

Also, there has already been quite a bit of research around behaviours that have no relationship to processes of constitution. As I’ve mentioned above, many models of adaptive behaviour ignore self-constitutional processes. So it appeals more to me to try explore the less well studied topic of systems with highly integrated mechanisms of behaviour and constitution.

Furthermore, it seems to me that for the studies that look at behaviour without constitutional processes, we run into a philosophical problem not dissimilar to the grounding problem that computationalists have experienced. Specifically, it is difficult to transition from the notion of ‘just a physical, dynamical, system’ to an agent with a 1st-person perspective. I know that this is still somewhat tenuous, but I think (and I’m not the only one!) that relating cognitive processes to processes of self-constitution is a methodology that may be a stepping stone to understanding what 1st-person perspectives are.

When you have a system that must act to counter its relentless degradation, environmental phenomena have value with respect to the ongoing existence of that system. Things are either ‘good’ or ‘bad’ for that system — and here perhaps is the beginnings of the most primitive imaginable kind of 1st-person perspective. These kinds of systems do not just have a ‘good’ or ‘bad’ relationship to environmental phenomena, but also react appropriately to them. They behave! This is what I have been trying to model. I think that the model I described shows some primitive forms of systems that act on their own behalf.

There were more interesting comments and questions, but I fear I have already written too much here!  So if there is something important you feel I’ve missed or want to add, please comment.  Thanks again to everyone who came along.



One Response

Subscribe to comments with RSS.

  1. Tom Froese said, on August 24, 2009 at 12:48 pm

    Hi Matt,

    Thanks for posting this summary. I usually find the following to be a good rule of thumb to distinguish between the two kinds of phenomena which your model tries to integrate:

    constitutive: any phenomena that are involved in the ongoing construction of the system as an entity

    behavioral: any phenomena that are involved in changing the relations between the system as an entity and its environment

    Of course, for natural systems this separation can indeed be problematic (see, for example, the discussion in our recent ECAL paper here), but in terms of artificial systems it is usually quite clear which of these two domains is being targeted.

    As such, the fact that this distinction has become harder to defend in the case of your model might not be a drawback, but rather an indication that the model has become more biologically relevant!


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: