Are ordinary people really populations of interests rather than something more solid? It’s disturbing to think of yourself as so fluid, so potentially unstable, held together only by the shifting influence of available rewards. It’s like being told that atoms are mostly empty and wondering how they can bear weight. Yet the bargaining of interests in a society can produce highly stable institutions; perhaps that’s also true of the internal interests created by a person’s rewards…these patterns look like familiar properties of personality. – George Ainslie, Breakdown of Will, p 44
Productivity methods and self-help advice that promises to improve one’s effectiveness at achieving goals (Getting Things Done, Lifehackers, etc) are all the rage these days, but I have mixed feelings about them. On the one hand, who can argue with people trying to improve themselves and become more effective? But something about this form of discourse makes me suspicious, something doesn’t quite add up. How can one will oneself to be more willful? Becoming more dedicated to your goals sounds good, but that is true only if those goals are the right ones to have; where did they come from, how did they get chosen out of all the available goods in the world? And how do you know when it is time to let go of your goals and revise or replace them? People occasionally have to pivot just like startups do; and a narrowly-focused dedication to one goal can mean missing out on better ones. In short, the management of goals and the willpower that they direct is a fundamental mystery of human action, and the productivity experts seem to blithely ignore all the theoretically interesting aspects of it.
This literature reads as if Freud never existed. If there is one valuable insight to be gleaned from his problematic legacy, it is that our conscious intentions are at best the tip of a very large hidden iceberg of unconscious motivations. Our true purposes are obscure; the mind is a disorderly riot of conflicting drives, we are constantly tripped up by desires we are not even aware we have.
Freud and his inheritors are distinguished by their method of anthropomorphizing internal mechanisms of the mind – treating a part of the mind, say the id, as an autonomous being with its own goals, agency and some limited intelligence. Marvin Minsky’s Society of Mind theory is probably the most worked-out contemporary theory in this mode. I’ve studied with Minsky, and have internalized much of his worldview, but this post is an attempt to grapple understand the work of George Ainslie, whose approach is rather different. Where Minsky tends to be more cognitive in his approach, Ainslie’s theory is deeply rooted in drives, rewards, behavior, and quantitative utility theory. His version of this he calls “picoeconomics” – that is, the internal and very-small-scale internal economy of the mind.
All these thinkers, themselves no doubt motivated by disparate goals, focus on unpacking the seeming solidity of the self and unveiling an underlying disunity. Viewing the mind as an internal society of conflicting agents has some conceptual problems, of course – mechanisms, after all, are not people, and it is easy to criticize such theories for positing an infinite regress and hence explaining nothing. As science then it may be problematic, but as a matter of practical personal self-understanding, it seems to be an extremely powerful technique, and also a possibly dangerous one.
The actual self seems extremely slippery in such a model. There is no central control of the mind, it’s just a loose collection of desires, agents, plans, and assorted bric-a-brac. How then does a coherent self get pulled together from this mess? One of Ainslie’s key points is that the main or only reason we have a self at all is to construct and enforce long-term bargains between independent behaviors and interests, and the self is better understood as a temporary alliance than an organ or a structure.
A practical person might ask, why you would want to see yourself this way? Isn’t it a somewhat destructive intellectual goal? Even if the self is fictional, it is surely a necessary fiction, and undermining it might be a bad idea. Integrity is a virtue, why the emphasis on its opposite? In at least the cases of Ainslie and Freud, the motive is at least in part therapeutic. Freud wanted to treat people who were suffering from problems caused by the irruption of repressed desires; and Ainslie is motivated by his work on addiction, a paradigmatic case of the failure of the unitary self. An intergrated self, if it is anything, is a construction, an achievement, an emergent fact out of a pre-existing disorder, and one needs to understand this process in order to avoid or correct the pathologies that the process is prone to. But one thing Ainslie is not doing is offering a self-help manual on increasing your willpower. In fact he is deeply skeptical of the idea and includes a chapter on “The Downside of Willpower” pointing out that an overactive rational pursuit of will can be self-undermining.
Outline of Ainslie
First, an apology: this is a very compressed account of a book that is already quite dense with ideas. For more detail, see this summary, and if you are just starting down the road of self-division, you might do better to start with Minsky or the popularized Freudianism of Transactional Analysis.
The temporal inconsistency of preferences
Ainslie starts from the observation that human preferences tend to be heavily biased in terms of favoring immediately available rewards over those that are distant in time. In other words, we discount a goal based on how far away in time it appears. The ice cream we are trying not to eat is easy to resist if it is down the road in a store, harder if at home in the freezer, impossible when sitting in bowl in front of us. The rewards of ice cream and the rewards of dieting both produce a curve that discounts rewards over time, and they compete (along with other rewards and behaviors) for dominance
If these curves obeyed the normal rational rules of of economics, they would discount rewards at a constant rate over time, producing a curve of preference that decayed exponentially with increased distance in time. However, it appears that we actually discount the future at super-exponential rates, producing a hyperbolic curve, more sharply curved than an exponential. This hyperbolic discounting takes place at a broad range of time scales (from sub-second to weeks or longer) and has been experimentally verified in both humans and animals.
One of the key consequences of hyperbolic discounting is that our preferences can be inconsistent over time. A purely exponential discounter doesn’t have this problem; a temporal graph of perceived utility over time of various exponential reward curves will not have any of them intersecting. But with hyperbolic discounting, short-term goals, when they are imminent, can override longer term ones, or in other words short-term preferences can dominate longer-term ones (figure from Ainslie):
In B, the behavior with a short time perspective is temporarily capable of overriding the longer-term one. This temporal inconsistency, according to Ainslie is the mechanism underlying not only succumbing to temptations, but a wide variety of other common problems as addiction, procrastination, and compulsions. At a more fundamental level, the inconsistency of our preferences is a problem for a would-be rational agent, because it means we can never trust ourselves to pursue the same goals in the future as we hold now.
The need for long-range goals to somehow suppress temporarily dominant short-term goals leads to an existential impasse. Because we can’t naively trust our future selves, we are thus in some sense radically alien to ourselves. But our future selves are not wholly out of control: their behavior can be predicted, their choices can be influenced. So what methods are available to enforce our current preferences on these future versions of ourself? Ainslie catalogs some techniques for what he calls “intertemporal bargaining techniques”, including:
- Extra-psychic commitment: this means changing one’s external circumstances so that future temptations are avoided or resisted. The classical example is Ulysses’ strategy of resisting the call of the Sirens by having his men tie him to the mast of his ship. A more modern example would be a drug addict who checks themselves into a rehab center to force themselves to be away from drugs, or an alcoholic who doses themselves with a drug that changes the metabolism of alcohol so that drinking leads to severe nausea. In all of these cases, a person acts in the present to alter the choices or rewards available to them at a future time.
- Personal rules: You swear on New Year’s to forego desserts, and somehow hope that the resolution has enough strength to overcome future temptations. Sometimes it does, but more often it doesn’t. The act itself seems rather mysterious: how can this purely mental action affect future choices? How can you bind yourself to follow a rule that you are also motivated to break? The tenuous possibility of success of such a resolution seems to involve the substitution of an abstract principle for the concrete situation which causes the undesired short-term choice, in essence implementing a sort of Kantian categorical imperative. If this mental trick succeeds, then breaking your diet is no longer merely a single act (whose isolated consequences are, of course going to be minor) but an act that destroys a valuable abstraction – not only the diet, but the fact of commitment, and the person’s own image of their strength of character. In other words, having a rule raises the stakes of an individual choice are raised to the point where long range goals can override short term ones. The well-known Seinfeld technique for self-enforcement of personal rules is a partly externalized and explicit version of this technique.
Commitment as recursive self-prediction
Ainslie is essentially saying that a person is in a situation of playing an iterated prisoner’s dilemma game with one’s future selves – like the classical IPD, one can cooperate (act in a way that serves the shared long-term interests) or defect in favor of more “selfish”, local, short-term interests:
Hyperbolic discounting makes decision making a crowd phenomenon, with the crowd made up of the successive dispositions to choose that the individual has over time…. Participation in the acts of this crowd of successive choice-makers is an extremely self-referential process, hidden from the outside observer and even from the person herself facing it in advance. She can never be sure how she herself will choose as she tries to follow this crowd and also lead it from within. (BoW p130)
If simple rational utility maximization with time-consistent discount curves was the norm, we wouldn’t need anything that looked like decision-making: we’d just act on whatever drive promised the most utility, with no more complex mental machinery required than a simple maximizer function. But due to temporally-inconsistent preferences, such a simple scheme doesn’t work this way. Instead, our conflicting agents are forced to negotiate with each other, creating bargains and alliances using the techniques described above.
The self as the locus of intertemporal bargaining
Picking a strategy in an prisoner’s dilemma game involves predicting the actions of the other player, and hence modeling them. What Ainslie seems to be saying is that this recursive process precedes, generates, and underlies the self. In some sense we bootstrap our selves into being through this process of trying to wrangle our infantile drives into coherent longer-term actions. We base current behavior on both our past actions and responses, and our predicted future actions and responses, and our self-representations are tools for enabling and/or artifacts generated by this highly recursive and self-referential process.
Here is where Ainslie’s story becomes hard for me to follow. I deal with complex recursions all the time, it’s part of the job of a computer scientist. But there is always an element of mystery to them. Recursions involving the self, even more so. So I can’t quite grasp this process in its entirety, and so can’t fully judge whether I find it a believable theory of mind. Unlike some of the more straightforwardly cognitive models of the mind, Ainslie’s seems to be rooted firmly in the precognitive unconscious and animal behavior.
Ainslie includes a chapter on the experience of intertemporal bargaining. where he warns that his theory may not only be counterintuitive but actively resisted, because it is in the nature of our internal behavioral rules to resist attempts to weaken them, and becoming aware of the somewhat arbtirary nature of rules that we create for ourselves can only serve to weaken them. In other words, it is potentially dangerous to become too self-aware of your mental machinery and the way real way your motivations work. You have been warned.
The Politics of Mind
…this last is of special use in moral and civil matters: how, I say, to set affection against affection and to master one by another: even as we use to hunt beast with beast…For as in the government of states it is sometimes necessary to bridle one faction with another, so it is in the government within. – Francis Bacon (quoted in BoW, p5)
A person, like a society, is composed of parts with their own private agendas, all taking part in a continuously renegotiated dance of conflict, cooperation, and compromise. Our disparate motivations are like politicians trying to advance a faction, and the self, such as it is, is something like a prime minister – not powerful in its own right, but because it has managed to become the public face for the most powerful faction. Our inner life is a noisy parliament.
Given the dysfunctional reputation of external legislatures, this might be cause for despair. But consider two points: First, what are the alternatives? The older model of the self is equally a reflection of the monarchic system of government, or a military hierarchy, the king-general-self seated at a pinnacle of command and handing down orders to the lower parts who execute them. That model doesn’t comport well with either the reality of human action or a modern esthetic of systems. Second, given the inescapable fact of conflict in both society and the individual, we have no choice but to struggle towards systems of governance that work, where “work” means to successfully arrive at suitable compromise solutions that are stable and satisfy a reasonably large subset of constituent interests.
Of course it is common to have legislatures that are mired in stalemate, like the current US Congress. The internal parliament of mind has its own set of pathologies: akrasia, addiction, compulsion, irresolution, repression, etc. The distributed agency theories of Freud and his descendents open up the tantalizing possibility of a unified theory of power and action over distributed systems, one that can show how assemblages of agents form, compete, cooperate, and dissolve, that cuts across psychology and political science. Perhaps these two most intractable problems – the organization of society and the organization of the self – can some day illuminate each other.
So you and every person you interact with are each a whole society of separate agents, with an internal economy, government, and politics. Some people may be organized like monarchies with a strong central self, others may be more anarchic bundles of disparate impulses, others may be flexibly improvising democracies of interest. A kind of mental anarchy is probably the infantile ground state, with structures of governance emerging over time. We all probably are familiar with people who have either too much or too little governance over their impulses. Personal interactions are like diplomatic missions between countries, and our social selves the ambassadors, forced to represent a complex system in a simple, polished, and understandable form.
Most of us in the technology world I think find politics (the external kind) distasteful – because of its dysfunctionality, inelegance, and because when it does work at all it requires dealing with humans on a retail, personal level, rather than as abstractions. But if individuals are themselves loose collections of divergent agencies, with only cobbled-together alliances maintaining a semblance of unity and coherence, then politics in a sense can’t be avoided at all – it’s what we are made of. The dominant political philosophy of technologists appears to be libertarianism, an ideology that may be pretty accurately defined as the belief that politics can and should be replaced by something else (the market). Fighting this tendency has been an obsession of mine for decades, and while I am not going to fight that battle here I cannot resist the fresh insight gained from an immersion in Ainslie: that the flight from politics is in some sense a flight from authenticity, a denial of our true nature, and a proposal to replace it with something shiny and superficially attractive but utterly foreign to who we really are.