Re: "The Evolution of Agency" by Michael Tomasello

On 4/5/2024 4:13 PM, Arkalen wrote:

On 05/04/2024 16:02, John Harshman wrote:

On 4/5/24 4:13 AM, Arkalen wrote:

Hello all,
>
Has anyone here read "The Evolution of Agency" by Michael Tomasello ?
>
I thought it was a really interesting (and very short) book that kind of blew my mind, and months later I can confirm it still impacts how I think about human consciousness and social living. I'm still not sure though how much of that is just being dazzled, or reading things for the first time that are actually already well-known, or if the book is plain wrong and if so on what.
>
>
I'd toyed with the idea of doing a book report here, and still might if motivation arises, but I figured now it's been out long enough that someone else might actually have read it and have takes.
>

I for one would be interested in a summary.
>

 Maybe really short to start with:
 Tomasello defines "agency" as a kind of goal-seeking system: a system that has a goal, is capable of perceiving the environment, verifying whether the goal is met, if not deploying a behavior that would move the goal forward, and looping between verification/behavior until the goal is met. Analogy is a robot lawnmower.
  He argues that animal evolution has seen progressive complexification of agency that basically involves adding layers, with higher ones monitoring/controlling the lower ones. He focuses on the history of human evolution specifically and takes some example organisms from lineages that presumably match the level of agency a human ancestor would have had (he acknowledges convergent evolution of various levels of agency in other lineages but leaves it at that).
 The levels he describes are:
 * no agency - nematodes. There isn't goal-seeking, just stimulus-response. The animal eats food if it runs into it, escapes danger if it's present etc but doesn't really *plan* or anticipate beyond its immediate environment

You have to go back to single celled microbes to have "no agency", and then it would be trying to define what you are talking about.  If you want to define agency as something that requires a brain and that type of decision making, you can do that, but that is just an extension of what organisms were doing before they had brains.  Take a simple behavior undertaken by bacteria.  There is something called the SOS response.  A bacterium finds itself in an unfavorable environment, a physiological response is started that results in genetic mutations occurring faster than normal.  The bacterium does this because it obviously has worked to improve the individuals situation at a high enough frequency that the bacterial lineage survives as a population.
Bacteria can use their flagellum to move to better environments, and they have sensors and decision making apparatus in terms of changing direction and moving towards that better environment.  Nematodes have a more sophisticated system to do the same thing.  Humans have an even more sophisticated system to do the same thing.  Chimps and humans have group agency, they can form hunting parties more sophisticated than wolf packs and lion prides, but wolf packs and lion prides still have less sophisticated group agency.  Have you watched Planet Earth on the BBC channel?  They show group agency among sea creatures.  Sea snakes and fish cooperate in order to be more successful in hunting prey.  They also show octopus and fish cooperating in order to hunt prey.  They show group agency among fur seals to trap fish, and they show dolphins and whales cooperating with each other to be more efficient predators.
Agency just seems to have levels of being able to interact with the environment.  It is a general aspect of life because organisms that can do it have an obvious advantage.  Life has evolved more sophisticated means to interact with the environment, and it has resulted in what we call consciousness.
Ron Okimoto

  * first level of agency - early vertebrates. Basic goal-seeking: a lizard will have distinct goals at any given time (seeking food, resting, etc) and their behavior will depend not just on their immediate environment but also on what their goal is at any given time. They'll have hardwired behaviors that are deployed to meet the goal according to the above described feedback loop, with a global inhibition mechanism that can shut everything down in response to danger.
  * second level - early mammals. They have a second layer of goal-seeking feedback-loop system that pilots the first in order to not just achieve a goal, but achieve it in the best way possible. Includes the ability to mentally stage potential behaviors, anticipate their outcomes and then pick which will work best. Also much finer inhibition abilities, able to inhibit one behavior and switch to another in service of the same goal instead of the global shutdown of lizards. And the ability to generate new behaviors instead of the earlier hardwiring.
  * third level - great apes. I can never remember quite what this one does so I had to re-skim a bit but basically it's an extra layer monitoring & controlling the mammalian decision-making one that results in higher-level reasoning, including controlling not only how to achieve a goal but which goal to seek to achieve (thus resolving cases where goals conflict), understanding other's decisions, and understanding causality - not only the basic mammalian "if I do this, that will happen" but "if this happens *in general*, that will happen".
  * fourth level - humans. Tomasello argues that the human "secret sauce" is essentially collective agency - reasoning agents like great apes that are also able to function as parts of a collective goal-seeking agent that uses the same basic "goal-perception-behavior-perception" feedback loop as all other agency. This means simultaneously modelling different kinds of agent - the "self" agent analogous to great apes (what are your personal interests) but also the "role" agent (what is your job as a member of the group) and the collective agent (what are the group's goals), as well as the "self" agency & "role" agency of collaborators. This also implies/explains moral dilemmas: whereas other animals only ever need to worry about one agents' goals (themselves), humans need at every point to decide whether they'll behave in service of their personal goals, or those of a group they're part of and if so which one (the immediate task-oriented team? Their family? Their tribe?).
 [Tomasello doesn't mention (but I immediately thought of) how a species with an innate sense of collective agency where said agency was both greater than any single individual but also analogous to their own agentic self might rationalize that sentiment...]
  He proposes this happened in two steps, the evolution of pairwise collaboration somewhere in the hominid lineage, and the evolution of full group agency in direct Homo sapiens ancestors, with the development of strong cultural variation & norms that provide a kind internal regulation to the collective agent and allow collaboration between strangers that share a culture. He also proposes actual group selection acting at this point, with groups with strong collaborative cultures outcompeting others.
 [this meshes beautifully with the "social cohesion signalling" hypothesis for the evolution of music btw]
 [another aside - he doesn't say this but it occurs to me that "arbitrary cultural variation + collaboration" seems like it could drive the evolution of universal computation (so children can learn whatever 100% arbitrary cultural baggage is required to function in their band) as well as truth-seeking (the only grounding on which one *could* base collaboration in the face of arbitrary cultural variation), which is kind of the core of what we think of as "our intelligence"]
   He proposes the driver of the evolution of each of these layers is a kind of unpredictability in the organism's environment. Animals that live in very stable environments can rely on hardwired behaviors that change at the speed of evolution. Early vertebrates would have been hunters that had to quickly adapt to also-evolving prey and agency/goal-seeking would have allowed that. Early mammals were social animals that were in competition not only with the prey but with their peers; an extra control layer to optimize behaviors would give them an edge. Early great apes that foraged for fruit would have had even higher competition for this resource that's concentrated in isolated patches with few access points.
  OK maybe that didn't end up that short ;) I hope I didn't mess it up too badly given it was mostly from memory, but that's basically it.