Focusing on "Autonomy"
Continuation of: Superfoo
Rafe and I had a great chat on the phone today about Superfoos. I think we agreed that there will be multiple instances of agents emerging in the level immediately above humans but there is always a single top-level network in local space. I think we also agreed that the “awareness” at this level will be different from human -awareness. It probably won’t subsume our awareness (at least without a technological singularity) but will exhibit properties such as self-preservation.
Where we got stuck was on the concept of autonomy. Stuck isn’t really the right word. We both greatly expanded our conceptual space around autonomy. But we didn’t come to agreement on a definition. However, it was a very productive conversation, so I thought I’d put my impressions down here.
I look at autonomy from a decidedly economic standpoint. How many different alternatives on how many difference axes can I possibly choose? Rafe and I agree that increased economic complexity leads to increased choices because everyone has more resources. I think Rafe looks at autonomy from a decidedly psychological standpoint. How many different alternatives do people feel are reasonable for them to choose? Rafe and I agree that the average person probably feels more constrained in their behavior as societal complexity increases because they have more responsibilities.
The difference between these two perspectives raises some interesting questions of free will. Is it truly a free choice if social isolation is the consequence of one alternative? Now, here’s where I believe the multiplicity of superfoos becomes extremely important. If there are multiple superfoos, there’s likely to be a spectrum of the degree to which they impair psychological autonomy. Some may value and promote diversity. Then you’re free to choose a superfoo that gives you the degree of psychological autonomy you prefer. These superfoos then compete at the higher level for advantage. It’s an open question as to which superfoo strategy is better.
Even this analysis is incomplete. In economics, a major insight is that people choose how to allocate their time between labor and leisure. I would refine this to say that people choose how to allocate their time between transactional labor, social labor, and leisure. The difference between transactional and social labor is that the primary purpose of the former is to generate monetary capital and the primary purpose of the latter is to generate social capital. Think of the difference between going to work as a programmer and volunteering to be an officer in your school’s PTA. So different superfoos could potentially offer different bundles of transactional labor, social labor, and leisure autonomy.
One thing to note here is that I am putting a premium on what I’ve described in the past as executive function, which includes the ability to choose your goals. I contend that humans (and potentially some birds and other mammals) are the first agents in local space to possess this capability. Therefore, the next level of agency will obviously build on it in how it organizes its human component. To me, that obviously means offering choice.
I’ll hope Rafe will expand on his thoughts.
Related posts:
- Going Meta on "Autonomy"
- Superfoo
- Response to "Superorganism Considered Harmful"
- Convergence
- Superorganism as Terminology
8 Responses to “Focusing on "Autonomy"”
Leave a Reply
We’re almost on the same page.
“Psychological autonomy” doesn’t really capture what I am after because it minimizes the reality of the constraints on people’s actions. The example we spoke of yesterday illustrates this. A person claims that he really wants a Rolex watch, but he doesn’t get it because it’s a luxury that he chooses not to indulge in, perhaps he’d have to save up first, or decide not to go out to dinner this month, etc. He does have the choice though and in this sense he is an autonomous actor in this scene. Now shift the object of desire to a mansion that costs more than the sum total of his net worth. Here, I would say that this is out of his space of alternatives (at least for the moment) and thus clearly he has no autonomy in this second scene. The contentious ground is in the middle where actors are interdependent. For instance, the person wants a modest house but will require a loan from a bank, or needs to convince his parents to buy it for him. Here I would say that he has the choice, just like with the Rolex, but he has less autonomy in the modest house scenario.
I liken the relationship between decision space and autonomy space to that between “total market” and “penetrable market” to bring it back to an economic analogy. That is, autonomy cannot exceed choice, but they are not the same. In particular an increase in the average level of interdependence between agents decreases the average level of autonomy, all else being equal. The confounding factor is that all else is not equal. The decision space increases as time goes on, thus depending on how fast interdependence increases relatively speaking, you could see autonomy net increase, net decrease or remain the same.
I lost you when went on about the importance of superfoos and choosing superfoos. Perhaps you can expound on that a bit further.
Also, while you are at it, I’m not sure I understand what you mean by “local space” in this context. And how does it relate (or not) to my usage of “decision space” and “autonomy space”?
- Local space just means the inner solar system. I don’t think we can extend our view of the organized matter network beyond that given current technology.
- I think we agree that your definition of autonomy space is a subset of alternative space. However, beyond this mere subset relationship, I don’t see why there’s any further analogy to total market and penetrable market. In fact, selecting an economic example of such a relationship is potentially misleading.
- Your example about the modest house illustrates the problem here. Without the existence of the bank, the modest house is outside choice space. The person simply doesn’t have the ability to acquire the house. However, with the increased interdependence on the bank, all of a sudden that choice is open through the possibility of a loan. This can’t possibly be a sufficient condition for a decrease in autonomy if the only axiom you have is that autonomy space is a subset of choice space. So no, I don’t agree that increased interdependence reduces choice. It increases it. You’ll have to define autonomy better for me to see how autonomy could decrease when choice space increases.
- I’ll expound on the concept of humans choosing their membership in superfoos. Higher level agents must somehow organize lower level agents, right? Molecules take advantage of the bonding properties of atomic electrons. Cells take advantage of the folding properties of proteins. Multicellular organisms take advantage of the chemical signaling of cells. Very often, the agents at the lower level can move from one higher level agent to another based on competition at the higher level.
What’s the primary mechanism for organizing humans? I say it’s exchanges based on incentives, in other words, choice. So the superfoos (on the level immediately above us) will probably take advantage of this mechanism. Human action as components of a superfoo will be organized based on providing the humans choices. Different superfoos may offer alternative choice regimes. So I don’t see how choice space will get smaller. In fact, as superfoos become more organized and competition at that level increases, choice space out our level should increase.
I believe our difference here is that where you see an increase in choice, I see an increase in the illusion of choice. When alternative space expands there does tend to be an increase in actual choice too, but also an increase in illusory choice. That’s because of interdependence, which is not to be confused with cooperation where there really is choice involved.
–
I now understand your point about choosing superfoos. But the problem comes when choice of superfoos is illusory or non-existent. The political system in the U.S. is a good example. As a candidate for Congress you have the choice to be Democrat or Republican. Is that really autonomy? When it comes time to vote on an issue, sure you have “choice” (illusory), but not choice for real. By and large, you depend on your fellow party members to get things done and if you don’t tow the party line, you will not be effective for very long. This is interdependence masquerading as cooperation. The stronger the agency of the superfoo, in this case the stronger the party is, the less autonomy its constituent agents have. Choosing your superfoo here doesn’t help the situation because your autonomy in the other superfoo is the same. In fact, if your autonomy were greater in superfoo #2, then superfoo #2 would be a less coherent agent than superfoo #1 and thus not be able to compete well with superfoo #1.
I’m having a hard time seeing the notion of illusory choice as well-founded. Some choices have more unpleasant consequences than others. But they’re still choices. Suppose you have choice set A and choice set B. Every choice in choice set B has unpleasant consequences. However, I think it’s pretty clear that A+B dominates A, because sometimes those consequences may be worth it. Your expected utility over all potential future states is clearly higher.
Given your contention that the superfoo level is rather opaque to those of us at the human level, I’m having a hard time buying any arguments from first principles about superfoo dynamics. We might be able to measure some aggregate statistical properties, but I’m leery of any deduction or analogizing. I’m pretty sure I can always come up with a counter deduction or counter analogy.
For example, in the US Congress, there is quite a lot of choice. It’s just at the level of factions within a party. In fact, I would argue that the party with more factions is likely to be successful more often because they will be more likely to adapt to changing sentiment among the electorate.
The point about superfoos with more choice being less competitive is obviously incorrect given history. The US, a multi party democracy with strong legal protections for individual autonomy (nee freedom), out-competed the Soviet Union, a single party dictatorship that attempted to stamp out any semblance of autonomy. Also, at the multicellular level of organization, there is clearly a premium on diversity.
If you look at economics and sociology, the two fields of study that I think are closes to examining superfoo dynamics, you find that there are indeed control mechanisms. However, stronger control mechanisms do not result in monotonically increasing organizational effectiveness. There is an optimum range. Whether that’s higher or lower than we have right now in the US, I don’t think anyone knows.
- A+B where B is illusory is simply A.
- Opacity is your best argument. My only hope against it is induction/extrapolation, not deduction. I think where we diverge is that I am looking at a broader range of systems and you are focusing on the human level (and the level’s just around it). I urge you to look at all systems and find the commonality, while I realize that the levels closer to human are likely to be more predictive.
- I do think the U.S. Congress is a perfect example. By invoking factions, you are simply changing the superfoo and making my point for me. If factions are necessary to get things done then the individual rep is not autonomous. Compare this to a system where there are no factions and no parties and every rep simply votes as befits their mandate.
- Regarding the Soviet analogy, it is a red herring because it’s a case at the margins and a “small numbers” situation. You are right about non-monotonicity. What I hate about this discussion is that we keep bringing it back to concrete examples but these also are too simplified to illustrate the macro trends when dealing with large numbers of agents. Forget the onesies and twosies and let’s just look at systems where there are at least a couple hundred agents interacting.
- If you define the set of illusory choices as those where A +B = A, you can’t simply turn around and say a choice is illusory therefore A + B = A. That’s circular reasoning. You have to tell me how to classify a choice as illusory based on some other rule.
- I’d be happy to look at other levels _if_ you could define autonomy so that I see how it applies to the human level. If I can’t figure out what it means at the level with which I’m most familiar, how can I possibly extrapolate from or to other levels?
- If I as an individual I can choose to assemble a coalition that changes the direction of a superfoo, how am I not autonomous? I may have made your point for you, but I can’t simply accept your statement without some sort of “autonomy predicate” that allows me to evaluate your assertion.
- The standard argument against induction is to find a counter example. You have asserted by induction that superfoos whose constituent agents are less autonomous are more competitive. I have provided a counter example. Therefore, your induction is false. If you are asserting that this is merely a statistical probability, fine. But then I’m going to want to dig into where your priors come from and their confidence intervals.
[...] December 28, 2008 by kevindick Continuation of: Focusing on “Autonomy” [...]
[...] • Response to Superorganism Considered Harmful • Superorganism as Terminology • Superfoo • Focusing on Autonomy • Going Meta on [...]