Haugeland offers a theory of systems along the following lines:
A component is a relatively independent and self-contained
portion of a system in the sense that it relevantly interacts with
other components only through interfaces between them (and contains
no internal interfaces at the same level). An interface is a
point of interactive “contact” between components such that the
relevant interactions are well-defined, reliable, and relatively
simple. A system is a relatively independent and
self-contained composite of components interacting at interfaces
(Haugeland, 1998, p. 213.)
Where Haugeland’s theory of system might work well for some
electronic systems, it fares much less well for some biological
systems. There are biological systems that are individuated, in
part, by their structure and the kinds of processes they carry out,
rather than exclusively by their interfaces. As an example, consider
the human muscular system.
The set of muscles in the human body constitute the muscular system.
In general, the muscles in the muscular system do not interface with
each other; hence do not constitute a system in Haugeland’s sense.
Even antagonistic muscles, such as the lateral and medial rectus
muscles of the eye, which move the eye left and right, do not connect
to each other. But, even if one were to say that antagonistic
muscles have a kind of indirect interface, there are other
combinations of muscles that do not stand in such antagonistic
relations. The lateral and medial rectus muscles of the eye do not,
for example, interface in any natural way with the gastrocnemius
muscles of the calves. What appears to unify the muscles of the body
into a system appears to be a commonality of function and a
commonality of underlying mechanism.
I guess it is somewhat common parlance to talk about the “muscle system”, but I’m not persuaded that its a great example of a system. Surely a system its at least true that a system is a reasonably independently function collection of entities. The “muscle system” is neither independently functioning, nor is a unified collection.
I’m much more persuade that the muscles, bones, tendons, etc. of the arm, for example, are a “system” than that the collection of muscles in my body is one. That’s just a collection of similarly realized components that play partial roles in a variety of systems in the body.
I’m not so sure about this objection. First, commonality of function is shared by my two mobile phones. Are they a system? Definitely no. Moreover, me and the prime minister of Nigeria definitely share some biological function but we are not a system either. So this is definitely too broad a criterion.
Second, commonality of underlying mechanism criterion is fulfilled in both cases I mentioned, as long as two mobiles are of the same kind.
On the other hand, it is not a good philosophical argument to say that some X is a system just because someone calls X a system.
Haugeland’s metaphysical notion is definitely too strict, as it disallows systems with components that are not detachable. Anyway, Herb Simon’s architecture of complexity is definitely much better here – it distinguishes between systems with relatively isolable and detachable parts and systems where interactions disallow any detachment.
I take the example from a college science textbook, Human Anatomy & Physiology, 5th Edition, Elain Marieb. It contains a chapter on the muscular system, one on the endocrine system, one on the urinary system, etc. So, my example does not come from mere “folk science”. I’m supposing that a philosophical theory of what a system is ought to somehow account for these scientific usages and that Haugeland’s theory does not account for these usages.
I think you are right regarding your first two points. The last sentence of the entry is too strong. How about this instead:
“Part of what appears to unify the muscles of the body into a system appears
to be a commonality of function and a commonality of underlying
You might naturally ask, then, what is the remaining part of what unifies them? I don’t think it can be interactive “contact” of the sort that Haugeland refers to, but I don’t have a positive account either. I’ll have to work on that.
= = =
What I take myself to be doing in this entry is a bit of philosophical explication of a scientific concept. I think Haugeland is doing the same. If you google “muscular system” you will find a number of web sites that are “sciency” in nature, so that it does not seem to me to be an idiosyncratic feature of the particular textbook that I picked that there is this use of the term “muscular system”.
= = =
“Haugeland’s metaphysical notion is definitely too strict, as it disallows systems with components that are not detachable.” Is there a typo here? I’m thinking that Haugeland’s notion disallows systems with components that are detachable, since such putative systems fail the interacting “contact” condition.
I agree that Haugeland’s account is unsatisfactory. I don’t think appealing to underlying mechanisms is going to help, because a mechanism is in turn a system in the relevant sense (made out of components … so again, one needs an account of what makes something a component of a mechanism).
In his forthcoming book, Carl Craver discusses this issue at length. He criticizes Haugeland and Simon’s accounts as inadequate and offers an account in terms of mutual manipulatibility of the mechanism and its components. Roughly, something is a component of a mechanisms iff it is a spatiotemporal part of it, you can wiggle the part by intervening on the whole, and you can wiggle the whole by intervening on the part. You should definitely check it out if you haven’t already.
Thanks for the reference to Craver’s book. It’s probably in the top three things on my reading list, so I’ll definitely look into it.
Here is a simple, but I hope not simplistic way of thinking of Haugeland’s account. It is unlabeled boxology; what defines a system is what is causally connected to what in a particular way. I’m suggesting we need at least some labeled boxology, with some account of what each box does. I know it’s a bold claim to make about Haugeland’s account, but I think that the labeling of what the boxes do, although not entirely absent, is also not very salient. I think this abets his larger project in the paper of arguing for extended cognition.
As noted, I have not yet read Craver’s book, but it seems to me to suffer at the hands of Marcin’s two mobile phones example. Phone A is a spatiotemporal part of the two, you can wiggle this part if you wiggle the whole, and wiggling the whole happens by wiggling the part?
It sounds as though this system stuff could still stand another paper….
“Part of what appears to unify the muscles of the body into a system appears to be a commonality of function and a commonality of underlying mechanism.”
I think this is not only too broad (as I pointed above) but two strict as well. Take systems with parts that play different roles (functions). A bike is a system, though the saddle doesn’t have the common function as the wheels. You could argue that the function of the whole system is what you mean by the common function. But the system function can be as well emergent, and some components could as well not have it.
I don’t even mention the trouble with defining ‘function’ here.
‘What I take myself to be doing in this entry is a bit of philosophical explication of a scientific concept.’
That is a good naturalistic approach, but I’m quite positive that you cannot simply account for all usages of such concepts as ‘functions’ or ‘systems’ in science in a single definition. I simply mean that you should be aware that scientific concepts, especially across different fields, will be ambiguous, and the part and parcel of good ontological work is to have univocal concepts. Simply use another concept for other cases that don’t fit.
Yes, that was definitely a typo. One “not” too much (three negations in a row… well…). You’re right about what I really meant.
Yes, the quoted passage is too broad. I admit that additional conditions are needed. Also note that I did not say that all systems are individuated in this way. I’m suggesting the the muscular system is a case where things seem to work differently. This, I think, blocks the “too strict” case.
Granted there are challenges to explicating both function and, as Gualtierro noted, mechanism.
What I guess is not coming through so clearly in this short entry is that I share with you the view that there is likely a plurality of concepts of “system”. I suggested that Haugeland may get some right, those in consumer electronics, for example, where there are others where Haugeland does not get things right, e.g. “muscular system,” “urinary system,” and “endocrine system.” This was only hinted at in the initial entry, but was not properly incorporated into the comment, “What I take myself to be doing in this entry is a bit of philosophical explication of a scientific concept.” A more neutral description of the project might have been, “What I take myself to be doing in this entry is a bit of philosophical explication of what one might take to be a single scientific concept.”
Well, that’s an interesting approach, but in Simon’s account, systems
could be abstract entities. And that makes Simon account great
for analyzing software systems. On the other hand, unimplemented
software systems or – even worse! – complex mathematical structures
which seem to be systems cannot be systems, as they’re not
spatiotemporal. In this respect, Craver’s account is biased. It would be
much more useful to use a concept of emergence here instead of
spatiotemporal manipulation (I like Bill Wimsatt’s account of emergence,
and I would use it here).
I disagree. Mathematical structures are just a different topic. The subject matter is physical systems/mechanisms. And software, when implemented in a machine, is no less concrete than any other physical thing.
Just to further flesh out this “multiplicity of systems” poitn, there
are “systems” that amount to something like a method. Rating and ranking systems and systems of
assessment are probably systems in this sense.
Games of chance, such as blackjack and poker, have systems of this
sort. Sometimes they are called
“strategies.” The systems in blackjack
are systems of card counting. They
include systems known as “Hi-opt I,” “Hi-opt II,” “Uston Advanced Plus-Minus,” and “Revere
Point Count.” Maybe these are non-scientific. They do seem to constitute a different target conception of “system”.
And what of “systems of linear equations”?
The aim of my initial post was to draw attention to cases that are apparently not covered by Haugeland’s account, but that are part of the target of the account.
Yes, in this respect the word “system” is clearly ambiguous.
Now, I tend to think that there is one sense of the word “system” that
is of philosophical interest (whereas the systems in blackjack or in
lotto or whaterever are not). This is a fairly complex structure that
has interacting parts that do not share some properties of the whole
structure, and seems to have at least relatively isolable parts; and in
most cases, it’s functional. This is not an attempt at defining it, I
just want to pin down the intuition.
Now, the question how to make the proper definition to include all
interesting instances, like biological systems (at least autonomous
biological systems), artificial systems, and…?
Note that if you analyze software systems, you really don’t care much
about hardware implementation. I haven’t yet come across a piece written
by an IT professional about a software system that goes on to say things
about spatiotemporal interactions…
So, I think we are entirely on the same page at this point. =)