Posts

Goal-Directed Agency and Intentional Agency

/27 Comments/in , /by

Index to posts in this series

Lizards

Lizards are a representative of the first category of agency according to Michael Tomasello in his book The Evolution Of Agency. As with the other categories, Tomasello focuses on the way lizards forage for prey. C. elegant, the nematode we looked at in the last post, moves and food either does or doesn’t go into its mouth. That obviously won’t work for lizards. Most species survive largely on a diet of insects, small creatures that move and flit about unpredictably.

So lizards have a different strategy. They have a goal, finding food, but they have to do several things to reach that goal. They have leave their hiding places and move to a place where there are insects. They have to spot a specific insect and then catch it and eat it. At the same time, they have to make sure they don’t become prey for other creature.

Here’s a video of a guy feeding crickets to his pet crested lizard. The lizard is in a special feeding cage. The crickets must be alive or the lizard won’t eat them. Bonus: you learn a new term: gut-loading crickets.

Following Tomasello here’s how I read this video. The lizard stands still orienting itself to its surroundings. The guy throws a cricket into the cage. The lizard sees the cricket. It takes a few steps forward. It pauses. It charges and grabs it and eats it. The guy throws another cricket into the cage. The lizard sees it, turns to chase it, misses, and pursues it in short bursts. It catches and eats the cricket. The guy throws a bunch of crickets into the cage. The lizard singles out one, chases, catches and eats it. The other crickets run around. The lizard spots one, chases, catches and eats it. This is repeated.

In this video we see a frilled lizard, a native of Northern Australia and southern New Guinea. It’s about a meter in length, and weighs over a pound. This video is heavily edited but again following Tomasello, here’s how it can be read. The lizard descends from it’s hiding place in a tree. It forages for prey, meaning any insect it might find on the ground. It sees one, approaches, catches and eats it. While chewing it looks around. It spots a predator, a black-headed python. It’s frill blows up. The snake approaches. The lizard runs really fast (also amusing). It eventually runs up a tree, making a full escape.

Tomasello says that the lizard has hard-wired goals: including eating and avoiding predators. Both it’s prey and its predators are quick and unpredictable. It has evolved to deal with that unpredictability by paying attention to the crucial aspects of the situation it perceives at any moment. It then acts to achieve a goal applicable to that situation. In both videos, the lizard eats while checking the situation. In the second it detects danger. It freezes its eating, then flares its frill, then runs. It achieves its goals by a series of go-no go decisions. That’s a clear step up from C. Elegans.

Squirrels

Squirrels are Tomasello’s example of intentional agents. He gives an example of a squirrel on a tree branch spotting a nut on a lower branch. The problem is whether to jump to the lower limb or retreat to the trunk and run down to the lower branch. He says the squirrel looks at the nut then at the tree trunk and then back at the nut. He interprets this as the squirrel cognitively considering two alternatives, then forming an intention to act, then acting. This two step process is evidence of an executive tier of mental control of behavior.

I have a young cat who does something similar. Winston knows there is interesting stuff happening on the counter when I’m making coffee. I see him looking at me, then at the counter as if contemplating jumping up. Am I paying attention to him? If I say No in a loud voice, he won’t jump. Maybe. But if I do nothing or am not paying attention, he works out an answer and acts.

Winston also knows how to open a door. Here’s a video of cats opening doors. Note that in several of the scenes the cat stares at the handle before acting. Tomasello interprets this as the cat forming an intention and then acting on the intention. Again, that implies an executive tier of mental activity.

Here’s a video of squirrels hiding nuts. Nut-hiding is a hard-wired activity, but it’s a complex problem. The squirrel has to decide where to bury the nut, and it has to be able to find it in winter. There are other squirrels waiting to steal the nuts, and predators. There are obstacles, including roots and hard soil. The squirrels stop and look around several times. In each case it looks like the squirrel makes a choice.

This is a 20 minute video of a guy operating a squirrel maze in his back yard. It’s not exactly relevant, but it looks like the squirrels are playing, something we don’t see lizards do. Also I enjoyed it.

The difference

Tomasello says that when lizards perceive a change in the situation, they freeze their current behavior and then respond to the new situation. He contrasts this chain of go/no go decisions with

… an either-or process of decision-making in which the individual simultaneously considers more than one behavioral option simultaneously (which mammals arguably do….) Pp. 33-4.

Tomasello says squirrels and other small mammals have an executive tier in their psychological processes that controls their operating behaviors, like running and burying nuts. This is a function of their larger brains, especially a larger pre-frontal cortex, more complex neurons and neural connections, and increased memory. Larger brains and longer time spent as juveniles increase the possibility of learning about the environment and experimenting, including play, which we might see as rehearsal. Mammals seem to have whet we would call emotions that also provide input to the executive tier.

The executive tier of the squirrel brain coordinates all these inputs. It considers alternative courses of action, “weighs” the costs and anticipated benefits of each and chooses one. That choice is communicated to the operant systems. The executive tier monitors the outcomes. It can inhibit one choice in favor of another if the situation changes or if the initial choice meets an obstacle or fails. Small mammals don’t use words, so it does this with some from of remembered perception.

Here’s Tomasello’s version of an executive tier:

The executive tier oversees the operational tier, as it were, and attempts to facilitate behavioral decisions via action planning and cognitive control. … It requires individuals to cognitively simulate in an organized way their own potential actions, the potential obstacles and opportunities for those actions, and the probable outcomes of those actions. They do this by perceptually imagining all these action elements in the common cognitive workspace and representational format provided by an executive tier of operation. P. 49.

It is this executive tier that gives mammals a wider range of choices of action, which presumably increases the chances of survival of individuals.

Discussion

1. Tomasello says that there isn’t any way to find mental processes in brains, so scientists infer the mental processes from the overt behavior of the creature. I think this raises the potential of anthropomorphism, leading people to ascribe human characteristics to other kinds of creatures. The executive function in humans is a defining feature of our species, so this is a real possibility.Of course, here we’re talking about a simple form of executive control.

The videos seem to provide at least some reason to think Tomasello is on the right track in ascribing an executive tier to small mammals. The empirical studies he cites may also be persuasive evidence.

2. I have described only a small part of Chapters 3 and 4 for this post. It’s allfascinating, especially the discussion of the executive tier, but it’s not relevant to my purposes in reading this book. As a reminder, this series is aimed at thinking about the origins and roles of rights in our society. I’ll get there, I promise.

creative commons

Free Will, Agency, And Evolution

/79 Comments/in , , /by

Most of us think we have free will, and we certainly act as if we do. We expect ourselves to do certain things and not do other things, and we feel responsible for those choices. We have the feeling, the sense, that we control those behaviors, or at least that we have the ability to control decisions about which things we do and which we don’t. We attribute to other people their own agency, which we take to be just like ours, even if they may have different ideas about proper behavior.

There’s a school of thought that says we don’t control those things. Here’s a recent article about Robert Sapolsky, a Stanford University neurobiologist, who doesn’t agree. He’s not the only one. Perhaps recognizing that this is an intractable problem, many scientists use the term agency instead of free will.

One is Michael Tomasello, whose book, The Evolution of Agency, I’ll be examining in the next few posts. Agency carries less moral baggage, and it’s something that can be described and studied neutrally; at least more neutrally. Tomasello doesn’t give a precise definition of agency. This is from the introduction:

…[I]n the current case, we may say that agentive beings are distinguished from non-agentive beings … by a special type of behavioral organization. That behavioral organization is feedback control organization in which the individual directs its behavior toward goals—many or most of which are biologically evolved—controlling or even self-regulating the process through informed decision-making and behavioral self-monitoring. Species biology is supplemented by individual psychology. P. 2.

The book rests on two assumptions. The first is that the basis of agency is a feedback control activity, a psychological mechanism, seated in the brain. The second is that agency is an outcome of evolution.

Feedback control organization

Tomasello’s feedback control organization works like a thermostat. The idea is that a goal is set for the thermostat: keeping the temperature at a certain level. It has a sensor that measures the ambient temperature and compares it to the goal. It then turns on another device that brings the temperature closer to the goal. It continues to test the ambient temperature and when it reaches the goal, it turns off the device.

Tomasello claims that this is the only model that can work to enable things to control themselves. He points out that all efforts to get machines to operate autonomously work in accordance with this model.

Evolution and agency

Tomasello doesn’t think there’s a goal for evolution. He thinks that as brains become more complex, the feedback control activity takes on a different shape, a shape that takes advantage of the bigger brain. I’ll just toss in the observation that mutations happen all the time, and some become established in subpopulations whether or not they have any survival value. That might include hair color or a larger brain. If circumstances change, the mutation may suddenly have survival value, and the subpopulation thrives while the rest of the population suffers.

Studying psychological processes

Tomasello says agency is a psychological process, one that occurs in the brain of an individual creature. It cannot be studied directly. Instead scientists infer the existence of psychological processes from the overt behavior of subjects.

Scientists infer psychological agency when the organism acts flexibly toward its goal even in novel contexts. To behave in this flexible manner, the individual must go beyond a stimulus-driven, one-to-one mapping between perception and action. The individual must be capable of choosing to act or not to act, or among multiple possible actions, according to its continuous perceptual assessment of the situation as it unfolds over time (sometimes employing executive processes such as inhibition, as a further control process, during action execution). P. 27.

The layout of the book

Evolution has been at work on this planet for hundreds of millions of years. We say that different species split off from lines of evolution, as humans split off from the great apes; and as homo sapiens eventually split off from the first hominids, and then evolved into modern humans. The lines go back to the beginnings of life on the planet, to the earliest living creatures.

Tomasello thinks certain existing species have no agency, and the rest fall into four categories. He selects five of them to represent his five categories of agency.

1. No agency: C. elegans, a tiny worm-like creature (the image on the home page is a bunch of these creatures)
2. Goal-directed agency: lizards as representative of reptiles
3. Intentional agency: squirrels as representative of small mammals
4. Rational agency: great apes as representative of great apes
5. Socially normative agency, which has two subcategories
a) young human children as representative of hominids with a simple form of socially normative agency
B) adults humans who exhibit a more comprehensive socially normative agency

Tomasello treats each category of agency in its own chapter. The last chapter is mostly for his fellow scientists, discussing gaps in the research and proposals for future work on this model. In each chapter Tomasello explains how the agency works, the evolutionary pressures that might have led to it, and the nature of the world as perceived by the example creatures. These issues are supported by a empirical evidence from academic and field studies.

I’ll take a quick look at the first three levels of agency, and discuss socially normative agency in more detail.

Creatures without agency

Let’s start with C. elegans. This is a worm-like creature about 1 mm in length. We know a great deal about it: we have sequenced its genome; and identified its 302 neurons, their connections, and the role each plays. It has no sensory apparatus beyond the ability to sense nutritious and certain noxious substances. It lives in organic material, where it eats bacteria. It has rudimentary powers of movement. They are mostly hermaphrodites. For more, see this dense Wikipedia entry.

Basically it moves around in organic muck eating bacteria. If it isn’t finding any, it moves. If it detects a noxious substance it moves. That’s about it. Tomasello says that with the tiny number of neurons, it’s hard to imagine the creature could have a goal, let alone behave flexibly to achieve it. It is purely stimulus driven. It’s sensory apparatus is very simple, so it only recognizes a few stimuli, and it responds to them mechanically.

In Tomasello’s terms, this creature is non-agentive. He calls it an animate actor. There’s not much else to say about it.

Discussion

I’m not fond of the word “agentive”, which strikes me as an ugly neologism, but it points to somehting about human behavior. Not all of our behavior is agentive. Take breathing. We can control it, but mostly we don’t. It’s an interesting exercise to think about what parts of our actions are agentive.

Another way to put that is to ask how much we resemble C. elegans.