Re: Dennett on agency

Posted by gepr on Oct 20, 2020; 10:58pm
URL: http://friam.383.s1.nabble.com/Dennett-on-agency-tp7599199p7599214.html

No, it's not entirely rhetorical to say that a computer virus "intends" to infect your system. The intent translates from the programmer's *behavior* to the program's behavior. It's more interesting if the virus fails to do what the programmer intended it to do. But if it succeeds, then the intent of the program was faithful to the intent of the programmer.

It's not clear to me which distinction you're arguing against: local vs. global transitivity?

I don't think it's too banal. RussA was pointing to a graded intent. Swapping out influence, persuasion, or control doesn't change the idea that a grading is necessary to make the argument work. If we talk in terms of networks, we could use something like hop distance (never mind edge weights for a minute). Sure, the Kazakh's intention/behavior might percolate all the way through the graph to me. But the strength of that intent would pale in comparison to the strength of the nodes more local to me. I.e. 'local intent'.

The question comes when we have to talk about larger scale intent, collective intent. Can the nodes go into some kind of harmonic where many nodes lose their local intent entirely? ... so that a large clique of the graph exhibits some kind of coupled/coherent, nearly atomic, intent? If so, then maybe something like Tononi's IIT mutual information measure could quantify the gradation RussA implies? And ... if so, then maybe *that's* why we tend to associate intent with organisms and things like cells rather than the efficient-cause-open structures like tornadoes?

And it might make sense to talk of cancer cells as having *more* localized intent than a well-behaved normal cell who gets along with her neighbors? But that could be twisted into a contradiction by considering parasites that kill their hosts vs. ones that live in some relatively successful symbiosis for awhile, retroviruses, etc. ... which, again, returns us to the concept of co-evolution.

And the whole thing smacks of the prior conversation about side effects vs. primary effects, wherein Nick, EricC, and Jon keep using the word "epiphenomenon" ... choose your poisonous word: "intent", "epiphenomenon", none of it makes any sense without some scoping, some distinction between local, non-local, and global. If it doesn't scale, it's useless.


On 10/20/20 3:27 PM, Steve Smith wrote:

>
> On 10/20/20 3:35 PM, uǝlƃ ↙↙↙ wrote:
>> Our common conception of a computer program defines it as a Pure Tool, no agency whatsoever. But perhaps *some* kinds of computer program (e.g. an individual S-expression grown by a genetic algorithm) might have a tiny bit more agency/intention than a hand-written program.
> And is it entirely rhetorical to say that a computer virus or worm
> "intends" to infect/co-opt your system, even though it was hand
> written?   Can the intentions of the creator transitively pass through
> the artifact? 
>> The fundamental problem, though, is the ideal, in the limit, ultimate agency/intention. I suspect we can narrowly escape that recursion problem by only allowing co-evolutionary structures where the "objective functions" are all implicitly defined by the churning milieu. That way agency/intention can be *locally* transitive but not globally transitive. E.g. I can be Renee's tool, but not the tool of someone I've never met in some small village in Kazakhstan. Similarly, a cell might be a tool of its local tissue, but not the tool of some distant cell in some other organism on Venus or somesuch.
> I like the juxtaposition of "agent" and "tool".   I suppose I would
> argue against the above distinctions by contriving a chain of
> influence/persuasion/control from the Kazakh to you?   Yes, on the face
> of it, you are more likely a tool of Renee's than that of the mysterious
> Kazakh, but maybe this (counter)example is too contrived/banal to  mean
> anything.

--
↙↙↙ uǝlƃ

- .... . -..-. . -. -.. -..-. .. ... -..-. .... . .-. .
FRIAM Applied Complexity Group listserv
Zoom Fridays 9:30a-12p Mtn GMT-6  bit.ly/virtualfriam
un/subscribe http://redfish.com/mailman/listinfo/friam_redfish.com
archives: http://friam.471366.n2.nabble.com/
FRIAM-COMIC http://friam-comic.blogspot.com/