How shall we understand goal-directed systems? What explains the ability of a bacterium to home in on its food? How is a cell is able to migrate to a target location in a developing embryo?  How is a homing torpedo able to track and move in on a target ship? What explains human purposefulness, our ability to formulate wants and to act to satisfy them? What explains teleology?

There is something magical about these systems, in which entities seem to be guided by the future. Somehow a migrating cell’s target location, a goal lying somewhere in its future, seems to be able to reach back in time and direct the cell’s movement right now. The growth of an acorn seems to be controlled, directed, by the oak tree that lies in its future. It was on account of this seeming future directedness that Kant said there would never be Newton for a blade of grass, that science would never explain the teleology built into living things. It is now generally recognized that Kant was wrong. Darwin was that Newton, and natural selection explains where teleology comes from, explains the origin of teleological mechanisms in evolution. The mechanism that turns an acorn into an oak tree is the product of millions of years of selection-driven design.

But explaining the origin of teleology is only half of the problem. The question remains: how do teleological systems do it? Natural selection does not direct the bacterium while it is swimming, nor the cell as it migrates, nor a plant while it is tracking the sun across the sky. Selection does not guide their behavior from moment to moment. Nor is it selection that gives these behaviors their teleological feel. In addition to a theory of origins, we also need a theory of mechanism, a theory that explains how goal-directedness works.

I propose that all goal-directed systems share a common physical organization, that they all consist of a goal-directed entity that moves within a larger directing field (McShea 2012). For a homing torpedo, the larger field that directs it is the “field” of sound emanating from the target ship. Cells in developing embryos often move within and are directed by chemical gradients or “fields” of gene activation. Nested within a field, goal-directed entities are able to find their way to a target trajectory no matter where they start, from anywhere within the field. And wherever they wander, wherever deviations in trajectory take them, the field is there too, directing them back to the right trajectory. In other words, movement within a field is what accounts for the two signature behaviors of teleological systems, what Ernest Nagel and Gerd Sommerhoff have called persistence (ability to recover from errors) and plasticity (ability to home from multiple starting points).

The proposal is that the key to goal directedness is nestedness, inside-ness, an entity nested inside a field. Notice a consequence of this view is that the seeming magic of teleological systems does not involve time. No goal in the future reaches backward in time and directs teleological entities. Rather, it involves space. What directs a teleological entity is something spatially large, a field that is large enough to direct the entity almost no matter where it starts and almost no matter where it wanders. From the perspective of the teleological entity, the field is everywhere.

This line of thought builds upon the work of a number of systems theorists, especially Herbert Simon, Donald Campbell, Stan Salthe, and Bill Wimsatt.

Human Purposes

Human behavior is the teleological phenomenon we are closest to. We think of our behavior as driven by wants, preferences, motivations, desires, and passions. I swung the tennis racquet because I wanted to hit the ball. I tidied my yard because I didn’t want my neighbors to think badly of me. We don’t know much about what wanting, preferring, and desiring are, but we think we know they are things that only a mind can do. We think the teleology of mind is a phenomenon wholly distinct from the teleology of simpler systems like homing torpedoes, plants tracking the sun across the sky, and acorn development. It seems obvious that human goal directedness is different.

But suppose that’s wrong. Suppose human wanting is not special, that its goal directedness arises from the same sort of causal structure as every other kind of goal directedness, an entity moving or changing within a field. The suggestion is at least superficially plausible, if only because behavior under the influence of wants shares the trademark persistence and plasticity of other goal-directed systems. I want to go to the post office to mail a letter. As I leave the house, I see the newspaper in the driveway and stop to bring it into the house. As I get into the car, I get a phone call, and I pause to talk, starting up the car and continuing on my way when the call is done. Along the way to the post office I run into a road closed for repairs, and I take a detour, eventually arriving at the post office. This is persistence. The want, the motivation to go to the post office, is an ever-present drive of some kind, a mental “field” that directs my thinking and behavior from above. My thinking and behavior are free to some extent, free to vary under of the influence of idle observations and thoughts, as well as the diversionary wants they evoke. But to the extent that I am on some mission, under the influence of some single want or coalition of wants, I am directed. And the suggestion here is that a want is a larger-scale directing field of some kind. And that whatever entities in the brain generate thought and behavior must lie within that field. They must be directed by it.

Since I am speculating here, there is no reason not to broaden the scope of the claim. All affective phenomena — wants, preferences, motivations, desires, and passions — are directing fields of this sort, and all thought and behavior (including speech) is directed from above in this way. This suggestion contradicts nothing that neuroscience has to say about how affective systems work. And it is consistent with folk psychology. It is also consistent with certain (admittedly impressionistic) observations about affect. From the perspective of the conscious mind, affective states do seem “large.” And like large things generally, when they change, they change slowly. Affective states change slowly compared to the movement of thought and muscle, compared to the “movement” of the conscious self. Further, like many larger things, they are resistant, sometimes impervious, to our efforts to modify them from below, from within. We cannot ordinarily change a want or feeling by an act of conscious will. Nor can we easily summon wants or feelings. We cannot simply decide to want or feel something. Affective states seem to just happen.

And finally, consistent with all this, wants and feelings are extra-logical, operating outside of the rules of conscious rational thought. Wants, preferences, motivations, desires, and passions do follow thoughts in time, of course. And there is a causal connection there, with conscious thoughts causing, or evoking, affective states. I imagine a crying baby and the thought evokes sympathy in me. But the connection is not a logical one. Lightning follows thunder as a matter of physics not logic. And sympathy follows from the thought of a crying baby as a matter of brain structure not logic.

Thus, I argue, affective states seem large because they are in fact large, larger than consciousness. They do not merely influence consciousness. They surround, they envelop, it. Wants, preferences, motivations, desires, and passions direct consciousness, and they do so from above.

David Hume

Following a venerable line of argument in philosophy, no want or feeling follows logically from any thoughts or observations, from any set of facts about the world (Hume 1740). It might sound like I’ve got things backwards here. Are we not rational creatures? Aren’t our desires, and therefore our thinking and behavior, the product of reason, at least some of the time? The answer is no, and never. Because reason in the sense of pure logic is incapable of motivating anything. Logic has to do with necessary relationships among facts. From the fact that rain makes things wet when they are not sheltered, the fact that it is raining, and the further fact that I am outside unsheltered, another fact follows logically, the fact that I will get wet. But it does not follow logically that I should care whether or not I get wet.

Consider a more extreme example. Imagine a purely logical being, walking through the woods and suddenly charged by a bear. It observes the bear charging, and perhaps draws factual conclusions about what will happen. “The bear will impact me,” the purely logical being thinks, “It will knock me down, it will eat me.” But the purely logical being doesn’t care. It has no affective response at all, not a whisper of fear and no concern for its own existence (McShea 2012). It is adaptive to care about one’s own survival. It is normal, in the sense that virtually all humans and other sentient creatures do care. It is even “rational” and “reasonable” in the soft sense of both words, meaning something like “appropriate for a normal person under the circumstances.”  But no want or care follows from pure logic. A purely logical being wants nothing, cares about nothing, and therefore does nothing.

So what is the function of reason in the sense of logic? It is to serve the passions. In Hume’s words, “Reason is and ought to be the slave of the passions and can never pretend to any other office than to serve and obey them.” Or in Thomas Hobbes’ analogy, reason is a “scout,” deployed by the passions to discover routes to their satisfaction, to find ways for them to get what they want. A normal person faced with a charging bear does experience fear, a desire to survive, a desire that emerges not from reason but from the way his brain is structured. Then, driven by his desire to survive, reason sets to work calculating ways to satisfy that desire, to achieve that end: run, yell, climb a tree, play dead. For reason, this is a routine, almost mechanical task, a task for which it supplies none of the driving force and none of the urgency. Reason is a tool of the passions. It is a screwdriver.

Wholes and Parts: The Balloon

How is causation being understood here? How do fields cause behavior in the entities that are contained within them? I have in mind a simple model. Imagine a helium-filled balloon hanging weightless and still in the air in the middle of a room. The helium atoms within it zig and zag, this way and that, bouncing off each and other and occasionally bouncing off the molecules of plastic that make up the skin of the balloon. I walk over to the balloon, reach up with both hands and corral it between my palms, holding it steady for a few moments. And then slowly and deliberately, I move it 3 centimeters to the left. I release it, and it stays. Now it is clear what the effect of this 3-centimeter shift has been. The average location of all of the helium atoms has shifted 3 centimeters to the left. For individual atoms, the movement may not be felt immediately. If we were to number the atoms and focus on, say, atom number 1632, we might find that in the time it took to move the balloon, 1632 did not experience a 3-centimeter leftward shift. Indeed, there is a fair chance it moved to the right. Still, if we follow the trajectory of that molecule over a much longer time we will discover that it has moved, on average 3 centimeters to the left.

Notice two things. First the system is described hierarchically. It is nested, a small object within a large object. A physical arrangement like this naturally invites us to take a two-level perspective. There is the focal level, the level of interest, which in this case is the level of a single helium atom. And then there is the next level up, one level of nestedness above the helium atom, the balloon as a whole (including the atoms of gas within it).

Second, one natural way to think about causation in this story is top-down. I moved the whole balloon – plastic, helium atoms, and all – together and this caused the on-average movement of atom 1632. Of course, this is not the whole causal story behind the trajectory of 1632. A complete story would involve the collisions of 1632 with other helium atoms and with the plastic molecules of the balloon skin. More formally, we can decompose the causes of the movement of any given helium atom into two components. First, there is the on-average movement to the left caused by the movement of the balloon as a whole, or what I shall call “upper direction.” And second, there is rest of its movement, governed by purely local interactions, movement that I call “free.”

A Delicate Balance between Upper Direction and Freedom. We do not want to call a helium atom in a balloon a teleological entity. The point of this example has been only to explain how I am thinking about hierarchical causation. But we can say something here about the organization of systems that are teleological, namely that there exists in all of them a delicate balance between upper direction and freedom. Upper direction is necessary to return teleological entities to their proper trajectories when they wander. And freedom is necessary so that they can wander. Without wandering, without occasional errors, there would be no persistence.

Reduction versus Emergence. I invented this odd term – upper direction – in order to avoid a confrontation with the reduction-versus-emergence literature. In particular I am avoiding the term “downward causation,” which figures so prominently in that discussion. I am allowed to sidestep that literature, I think, because for my purposes, it does not matter whether higher-level systems and causes are reducible – either in practice or in principle – to lower-level ones. It does not matter whether the properties of the higher-level systems are non-linear or emergent. My own view is thoroughly non-reductionist, but reductionists are welcome to think of upper direction as just a heuristic, a shorthand description of how causation works in a hierarchical system.

Links to papers:

Teleology (2012)

Machine Wanting (2013)