Re: "The Evolution of Agency" by Michael Tomasello

On 07/04/2024 15:51, *Hemidactylus* wrote:

Arkalen <arkalen@proton.me> wrote:

>

[snip]

>
Here the difference he proposes between organisms with and without
"agency" isn't that we know the biochemistry in one case and not in the
other, it's a specific claim about how they function and how many
degrees of freedom the individual organism has with respect to their
evolutionary hardwiring.
>

Degrees of freedom is one notion Dennett explores to bootstrap “free will”.
 

>
He defines a "feedback-control system" for agency that has the following
features:
>
- a goal
- behavior(s) suitable to reaching the goal
- perception that verifies whether the goal is achieved
- a feedback loop between them to repeat the behavior until the goal is
achieved
>
>
The thing is, living things don't *have* to function with this kind of
organization. You could have an organism that chemotaxes towards
nutrients & absorbs all it encounters and chemotaxes away from toxins or
chemicals associated with predators/bad environments and reproduces once
it's reached a certain size, and evolutionarily speaking that's a
perfectly cromulent organism; if it can survive and spread this way it
will. Insofar as it has the "goal" of eating or breeding or avoiding
predators however that goal is evolution's goal more than the
individual's. The nature of the goal and the specific behaviors it
engages in to meet it change over the generations via evolutionary
processes. Insofar as there is a feedback loop between perception and
behavior that optimizes things towards the goal, the "perception" is
"how does this organism interact with its environment" and the feedback
loop is "is this organism reproductively successful".
>
>
But if you have a living thing that *does* have this kind of
organization then "goals" have a different definition for it.
"Evolution" still has the "goal" of it eating but the way this is
behaviorally implemented means the individual itself can be described as
having this goal in a completely different sense, that manifests
differently. You can even get differences between "evolution's goals"
and "the individual's goals" - usually not with eating but for example
you can have goal-driven animals evolve the drive to have sex, although
from an evolutionary point of view the actual goal is reproduction.
>

Hmmm, is this evolution’s goal stemming from Tomasello himself?

It doesn't *not* stem from him :p Let's say I'm pretty sure he uses the same metaphor phrased in a similar way somewhere in the book but I'm not certain and I'm not going to check so I feel bad implicating him in my phrasing choices.

Not liking
the idea of evolution having goals. First the outcome of evolution itself
can stem from several factors, selection being one. Given drift, neutral
evolution, and the prevalence of junk DNA in humans and other organisms,
evolution seems too happenstance for goals. Goal directed evolution is the
stuff of orthogenesis or omega point. Complexity of human brains or
evolution of complexity itself if I recall Gould on this is a drunkard’s
walk constrained against a lower boundary.

Honestly on reflection I agree that "evolution" having "goals" might be a bad metaphor here, although maybe not for the same reasons as you. First I want to narrow things down to *adaptations*, because those are what we're talking about here. The book fully assumes the cognitive mechanisms it discusses are adaptations, an assumption I think is reasonable but whether it is or not, it's normal for the book to use language and make arguments that make sense for adaptations even if not all evolved traits are adaptations and that language wouldn't work for those that aren't.
With that out of the way I've fully moved away from Dawkins' idea we should avoid notions of "purpose" with respect to adaptations and talk about "appearance of purpose" instead. I think it confuses more than it clarifies in most cases. I think it makes more sense to redefine "purpose" (and "function", "design" etc) in a way that covers both evolved adaptations and human engineering, because their underlying commonalities justify it. For example the way selective pressures for flight lead a wing to have the structure it does justify thinking of it as "for flying" the same way an airplane wing is and in a way a rock isn't "for having the precise shape it happens to have". There is an interplay between the structure suiting the thing to a function because the function was causally involved in making the structure that's common to both evolutionary adaptations and human design and accounts for the superficial similarities.
That kind of reasoning is why I wasn't bothered about talking about evolution metaphorically having a "goal" but I'm still rethinking that choice somewhat because I'm not confident the metaphor worked. Like, I explained how "purpose" can be redefined based on commonalities between how adaptation and human design work but in this context we're talking about the "goals" of a system called "agency" that was defined a specific way in the book, and I'm not sure "evolution" can be massaged into that definition even for a metaphor. This bit here was very much my attempt to do so:
 >> Insofar as there is a feedback loop between perception and
 >> behavior that optimizes things towards the goal, the "perception" is
 >> "how does this organism interact with its environment" and the
 >> feedback loop is "is this organism reproductively successful".
but the more I think of it the less confident I am that it worked. Evolution doesn't do loops!
I definitely don't want to say something like "bacteria seem to have goals but definitely don't and goal-directed agents do have goals, ignore the appearance of a commonality as it is pure illusion" because I don't think it's pure illusion, there's got to be a good way to account for why one looks like the other and express it simply.
(other than "the division Tomasello proposes isn't a thing at all" of course, which I don't buy)

 That said teleology should be watered down to teleonomy (Mayr) or apparent
goal direction in organisms due to their “programming” and is an outcome
not a target. One needs to differentiate also between the proximal focus
and distal (ultimate) when looking at evolutionary outcomes. Proximate
causation happens at the level of physiology and so called “goals” obtain
here as organisms negotiate their environment for food and such. Failures
resulting in reduced reproductive output will “reprogram” future
generations away from those failures.
 Said reprogramming may result in long term trends over generational time,
but that trending (eg- cognitive complexity) cannot be interpreted as an
evolutionary goal as trends can result in devastating dead ends especially
if the ecological context or fitness landscape shifts dramatically.

Long-term trends in evolutionary change is definitely not what I meant by "evolutionary goal".
OK I've thought on it a bit and I think I've figured out what I meant; turns out it's very similar to the "purpose" thing after all. You'll tell me if it makes sense, and if so, whether you can think of a pithy way of saying it.
1) When/why do we think systems have goals? When there it behaves in ways that yield a certain outcome, and the behaviors seem optimized so that this outcome will happen. As if the outcome caused the behavior and not just the other way around.
2) Humans have a whole cognitive system where this is indeed the case, with internal representations of the desired outcome, different possible behaviors, what outcomes they might lead to, and processing to ensure the behavior that's actually displayed leads to the desired outcome.
3) Tomasello describes his basic "feedback-control system" as a system where this is also the case, with an internal representation of the goal and control over whether (and which) behaviors are displayed depending on whether (and until) the outcome is achieved. He even argues that this is the minimal possible system that can be goal-seeking like this.
4) Bacteria (Tomasello would argue; I'm interested in counter-arguments) do NOT have such a feedback-control system. They do not flexibly adjust their behavior according to whether their perceptions match up to some internal representation of a goal.
5) Bacteria DO display behaviors that are causally related to an outcome as we expect of a system that has goals - the difference is that the process enabling that causal relationship, the one that adjusts the behaviors in such a way that they end up yielding a specific outcome, is not within the bacterium but is the process of evolution that produced it.
Hence, poetically but more misleadingly than I guess it's worth, "the bacterium's goals are not its own but evolution's".
4 is definitely the weak point of that chain IMO, my intuition is that it's true but I'm curious how rigorously it can be demonstrated.

 Sure humans and octopods have converged upon cognitive complexity, but so
many other species haven’t.
 And this is where adaptive evolution is being considered. I dare say most
evolution is not adaptive.
 [snip rest]