Walk Toward One, Back Away from the Other (Notes on Going Nowhere Efficiently)
At 8:40 on a Thursday night I was standing in the hallway of my own house, at the midpoint between my office and my kitchen, a phone in one hand and a coffee cup in the other, and I had been there long enough for the motion-sensor light to give up on me and go dark.¹
Behind me, on the desk, a laptop sat open to a half-built thing I have been calling a side project for so long that the phrase has become a pet name. Ahead of me, in the kitchen, was a coffee maker, which was not the point. In my hand was the phone, which had buzzed twenty minutes earlier with a friend asking whether I was coming to dinner on Saturday. I wanted to go. I also wanted Saturday, the one stretch of the week long enough to finish something, and I knew, with the confidence of someone who has run this experiment before, that saying yes to the dinner meant losing the day, and saying no meant the dinner would go on without me, pleasantly, and I would find out how pleasantly from a photograph.
What I had been doing for the better part of half an hour was this. I would walk toward the office, feeling the project pull, telling myself an hour would do it. Three steps in I would think of the unanswered text, and how an unanswered text does not stay one size but accumulates, and I would turn around. Three steps toward the kitchen I would think of the project, and how momentum perishes (the half-built thing was close to the stage where an idea becomes a chore, which is the stage where things either get finished or get quietly retired), and I would turn around again. The net result of thirty minutes of labor was zero displacement, one cold coffee, and a mild ache in my calves.
I would like to report I handled this with grace. What I did was stand in the dark and feel enormously busy.
There is a name for this, it turns out, and the name is a donkey.
Buridan's ass is a thought experiment in which a donkey, placed exactly between two identical piles of hay, starves to death because it has no reason to walk to one pile rather than the other. It gets taught as a puzzle about free will, and it has been taught for so long that nobody checks where it came from, which, having checked, I can report is nowhere especially dignified. Aristotle has a version, in On the Heavens, but his animal is a man, equally hungry and thirsty, and his purpose is cosmology rather than psychology.² The donkey shows up later, attached to a fourteenth-century French philosopher named Jean Buridan, who never wrote it down. The Stanford Encyclopedia of Philosophy says the example appears nowhere in his work and was probably a parody by critics of his theory of free choice.³ So the most famous starving animal in philosophy is a caricature wearing the name of a man who never owned it, which I found oddly comforting, having been on the receiving end of things I never said in meetings.
My first reaction, standing there in the dark, was relief of the wrong kind. The donkey is not me, I thought. The donkey has two identical piles. I have a project and a dinner, which differ in every way that matters, and nobody is ever standing at the exact midpoint of anything. The thought experiment is rigged. I was prepared to walk back to the office on the strength of this conclusion, and I got three steps.
It turns out someone with real credentials has been arguing for forty years that the game is not rigged. Leslie Lamport, a computer scientist best known for his work on distributed systems, ran into the donkey while studying circuits that must choose between two signals arriving at nearly the same instant, and in 1984 he wrote a paper called Buridan's Principle claiming the problem is a law and not a parable, one that turns up in ordinary life as readily as in hardware.⁴ His claim, as I understand it (I am not an engineer, and I expect to be corrected), is that any decision forced out of a continuously varying input cannot be guaranteed to finish within any fixed amount of time. Not that it can't finish. That nobody can promise when. His example is a driver deciding whether to cross a railroad track before a train arrives. And we do not find dead donkeys scattered across the landscape, he observes, with a dryness I admire, only because the odds of a donkey standing on the exact midpoint are infinitesimal.
Here is what this did to my hallway. The version of the problem I had rejected needed a perfect tie. This version needs only a close call. The nearer the two pulls come to equal, the longer the deciding takes, with no ceiling on how long, and the exact midpoint is not required, only the neighborhood of one. You do not have to be equidistant. You have to be near enough to a place where you would be.
I had, in other words, been standing at a yellow light for several years.
The phone buzzed in my hand. "No pressure!!" the friend had written, which in my experience is the most pressurized phrase in the language, and I put the phone face down against my leg as though it could see me.
Psychologists arrived at the same spot from a different direction, with rats. In the 1940s Neal Miller ran hungry rats down a short alley toward food. The ones that were shocked at the food end approached part of the way and then stopped, and they stopped farther back the stronger the shock had been. Later experiments, in which rats wore a harness that measured how hard they pulled, suggested that the pull toward the food and the pull away from the shock both grew as the animal neared the end of the alley, with the retreat growing faster, so that the two lines crossed.⁵ Where they crossed, the rat hovered. Not at the midpoint of the alley, which would have been a geometric accident, but at the place where the forces balanced, which is a different thing and a far more common one.⁶
Notice what this adds to the donkey. The donkey stands still. The rat does not. The rat advances, and the shock looms, and it retreats, and the hunger looms, and it advances, and this can continue for as long as the alley stays the length it is, which is to say indefinitely. Nobody had told me what happens near the balance point is motion. Motion that looks, from the outside, as though it were trying.
Miller's rats were not choosing between two goods, though. They were caught between a want and a fear, which is nearer to my hallway than the donkey ever was, and I will come back to that.
The motion has a price, and two lines of research put numbers on it, both with caveats I will confess to in the footnotes. In 1992, Amos Tversky and Eldar Shafir reported that when each option in a set had real advantages and real drawbacks, people grew more inclined to put off deciding, to keep searching, or to take the default, and that improving the set could make this worse, which is the opposite of what a tidy account of value predicts.⁷ A dozen years later, Jiwoong Shin and Dan Ariely tested what it costs to keep options alive, using a computer game with three doors. Participants, who were paid real money, had a hundred clicks to spend. Clicking inside a room paid a random amount, switching rooms cost a click, and the best strategy was to sample the rooms briefly and settle in the richest one. Then the researchers added a twist: any door left unvisited began to shrink and eventually vanished. People could not let them go. They spent clicks rushing back to reopen doors that were closing, and their earnings fell, and this held even when the doors were of little interest, and even with practice, information, and a visible cost.⁸ The authors read it as loss aversion more than a taste for flexibility, which is to say people were not keeping doors open because the rooms were good. They were keeping them open because closing registered as a loss.
I had always assumed keeping an option open was free. The door game says it costs clicks, and clicks were the only currency the experiment had, which is also true of Thursday evenings. Ariely put a playable version of the game on the website that accompanied his popular book, Predictably Irrational, and I have not played it, because I suspect I know how it goes.
I will confess what all of this did for me, which was relief. For years I had read the hallway as a verdict on my character, evidence of something soft in the middle of me, a failure to commit I would have noticed in anyone else and called by its proper name. Miller and Lamport offered another reading: that near a balance point anyone hovers, that hovering is what the structure produces, and that it requires no defect in the hoverer. For a page and a half of reading on my phone in the dark, I felt exonerated.
Then I noticed I was still in the hallway, now with a theory, which is worse than standing there without one, because a theory is something to do with your attention besides deciding. I had upgraded from busy to informed. The coffee was just as cold.
Here is the part of the story the donkey is never allowed. The donkey gets a single decision, made once, with both piles frozen in place forever. In life the donkey can walk to the hay, eat, and then walk to the water. The water will be a little farther away by then, and it will taste different, because the one drinking it is no longer the creature who was thirsty at 8:40. I do not want to wave this cost away, since it is real, and sequence is not the free lunch it gets sold as. But it is smaller than the cost of never eating, and smaller than the cost of the oscillation itself, and it was the argument I had been preparing to make to you: that these dilemmas are rarely either/or, that you can often take the hay and then the water, and that most of us take a step toward one, notice how much farther we have drifted from the other, and compensate, and compensate again, until we are back where we started and a little more tired. People I know stand in versions of this hallway with heavier doors at either end. The promotion against the soccer practices. The business against the friendships that thin out while it gets built. I had meant to tell them, and you, to walk to the hay.
I had planned to stop there. The evening declined to cooperate.
Sequence only helps if you know which of your two things is hay and which is water, which is to say which one keeps and which one spoils. Standing in the hallway, I could not tell. The dinner happens once, and by Sunday it will be gone, so surely the dinner is the water. But the project, I have been telling myself for years, is waiting for me, and I have been telling myself that for exactly as long as it takes a thing to stop waiting and start dying, so surely the project is the water. Depending on the minute, each of them was the one that spoils. Each claimed the perishable position, because that is the position that gets the vote.
The philosopher Ruth Chang has a name for the kind of difficulty this is, and it is less comfortable than the donkey. Hard choices, she argues, are not hard because we lack information, nor because the things we are weighing cannot be measured against each other. They are hard because the options are on a par: neither is better than the other, neither is worse, and they are not exactly equal either.⁹ That last clause is the strange one. If two piles of hay were equal, the donkey's problem could at least be solved with a coin. Parity is stranger. One test of it, which I have borrowed from a commentary on her work, is that improving one option slightly does not make it determinately better than the other. Sweeten the dinner (a better restaurant) and the decision does not flip. Sweeten the project (an extra hour on Friday) and the decision does not flip. They are not identical hay. They are not unequal either. They are a third kind of thing, and the third kind is where most of my evenings take place.
So I did what a person does at 9:15 in a hallway when his theories have stopped helping. I sat down on the floor, found an envelope in the pile of mail, and drew two columns. The psychologist Shalom Schwartz's theory of basic human values holds, on the strength of decades of cross-cultural surveys, that human values sit on a circle, with compatible values next to each other and conflicting ones on opposite sides, which is the sort of geometry that makes a hallway feel scientifically justified.¹⁰ I was not trying to place myself on a circle. I only wanted the reasons for each pole written down where I could see them, in handwriting.
Under the project I wrote: to make something with my name on it. To find out whether the last few years were building toward anything. To not be the person who only talked about it. Under the dinner I wrote: to be in the room. To be someone who shows up. To not be the person who is always about to arrive.
I read both columns twice. The reasons were different. The sentences underneath them were the same. Both columns, when I leaned on them, said some version of: so that it will have counted. So that when somebody adds up the years, something is in the column.
That is not what I had planned to find. The donkey has two wants, and Miller's rats had a want and a fear, and what I had on the envelope looked less like two things I was walking toward than one thing I was walking away from, the possibility of reaching the end with nothing to show or nobody to show it to. A person walking away from something everywhere has, as a practical matter, a hallway. A project and a dinner are two exits from the same room, and "so that it will have counted" is not a reason for either of them. It is a fear that has learned to sound like the reason for both.
I am going to say the next part plainly, because I have run out of jokes for it. I am tired. Not of the project, and not of the dinner, but of being always about to be somewhere. I do not know when it started, and I cannot tell you whether the oscillation is how I manage the fear or how I keep it going.
Leibniz, according to the encyclopedia article on the donkey, had the sharpest answer to the whole business: no donkey is ever truly balanced, because below the level of noticing, tiny perceptions are always tilting it one way or the other.¹¹ In his picture the animal never starves, because something it cannot feel has already decided. Spinoza took the opposite line, conceding that a creature in perfect equilibrium would die and then declining to be impressed by the fact.¹² I like both answers and trust neither, the first because it sounds like a consolation and the second because it sounds like a dare. But I thought of Leibniz toward the end of the evening, because of what I did next, which was count again.
My legs had gone to sleep from the floor. I stood up, carefully, and since I had a theory and an envelope and no other instrument, I returned to the spot where I had been standing and counted the steps in each direction. Toward the office: seven. Toward the kitchen: eight.¹³
I had been standing one step closer to the office the whole evening. Not at the midpoint. Not balanced. Tilted, by a margin small enough to hide in a floor plan, and unable to feel it, which is just what Leibniz said would happen.
I would love to tell you this settled it. A lean toward the office, so I want the project, so I go. But a lean is not a reason, and it is not even a desire. A floor tilts the way it tilts. It does not tell you whether you are sliding toward something or away from something else, and that distinction, I now believe, was the whole argument of Miller's graph.
So I took the step. One step, toward the office, in the dark, and I can give you two accurate descriptions of it. In the first, I walked toward the thing I wanted. In the second, I walked away from a table on Saturday night and the photograph I would have to look at on Sunday. I cannot tell you which description is the true one. I suspect both are, and the gap between them is what I have been calling indecision.¹⁴
The phone, still in my hand, still held the unanswered text. I turned it over and typed, "Let me get back to you," which is the donkey's favorite move, a step that counts as movement and commits to nothing. I looked at it for a while. I did not send it, and I did not delete it. The screen dimmed. The coffee was cold in my other hand.
And the light, which had been dark for an hour, came back on, as motion-sensor lights do, for anything that moves, whether or not it is going anywhere.¹⁵
¹ I should say, for the record, that I had counted the steps: seven to the office door, seven to the kitchen threshold, from the spot where I was standing. I counted them in the spirit of someone hoping the answer will settle something. Keep the seven in mind. I am going to cite it as evidence later, and I am not yet certain it deserves the weight. ↩︎
² The man appears in Aristotle's On the Heavens, where he is used to mock Anaximander's argument for a stationary Earth. So the first recorded case of paralysis by equilibrium was a defense of the planet's refusal to move, and I read that and thought, finally, someone who understands me. ↩︎
³ A later and better-known version comes from Al-Ghazali, who used two equally good dates, the fruit, and argued that free will can break the tie; Averroes disagreed. It says something about the problem that when it became personal, the stakes were snacks, which is about where mine are on a bad Thursday. ↩︎
⁴ Lamport has described the paper's reception: when first told of the principle, people usually find it unbelievable and propose mechanisms to get around it. It was written in 1984 and published in Foundations of Physics in 2012, which means a man spent twenty-eight years being told that indecision was not a real physical problem, and I take that personally on behalf of everyone who has ever stood at a crosswalk. One account also says his earlier paper on circuit glitches was rejected by a journal for making the case that the glitches could never be eliminated entirely. I am telling you all this partly to remind you of the seven steps. Hold the number loosely. ↩︎
⁵ The steeper-retreat picture is a model, and a contested one in its details. One later study trained rats to press levers for food or to avoid shock, matched the two closely on other factors, and found no reliable difference in the slopes of approach and avoidance. I am using the graph for the general layout and not for the street names. ↩︎
⁶ The encyclopedia article on the donkey credits rat experiments on two attractive goals to Kurt Lewin. The runway work I could actually verify was Miller's, and a business-school paper on decision conflict traces the research tradition to Lewin in 1933 and Miller in 1944. I have credited the rats to Miller. If you know otherwise, I will correct it, assuming I ever turn the comments on. ↩︎
⁷ In 2023, Ioannis Evangelidis, Jonathan Levav, and Itamar Simonson reported forty well-powered direct and conceptual replications of the conflict-and-deferral effects, with a combined sample of more than twenty-six thousand people, and concluded that the results cast doubt on the replicability of the earlier work. So the finding I just leaned on may not hold up. I kept it because it describes the feeling, and I am aware that is not a scientific reason. ↩︎
⁸ The game had a correct strategy (sample, then settle), and my problem has no strategy known to me, so the analogy is a good deal weaker than I am making it sound. I am using it anyway because the part that translated was the rushing back. The authors' last experiment suggested the mechanism was aversion to loss rather than a desire for flexibility, and I find the distinction harder to sit with than the result. ↩︎
⁹ I am working from Chang's abstract and a skim of the paper, not a mastery of it, so treat the next sentence as my reading and not hers. Her keywords include commitment and something she calls hybrid voluntarism, which I take to mean that when reasons run out, we can put our agency behind an option and make it ours. That is a hopeful idea, and I have not figured out how to tell a commitment from another step. Her longer treatment of the subject is in Making Comparisons Count. ↩︎
¹⁰ The original theory had ten values, and a 2012 refinement by Schwartz and colleagues has nineteen, still arranged on the circle, where adjacent values are the most compatible and opposing ones the most conflicting. The structure has been tested across many cultures. I mention this to be fair to the theory, which did not ask to be used as a decoration for my hallway. ↩︎
¹¹ The Stanford Encyclopedia entry on Leibniz's philosophy of mind discusses these unconscious perceptions at length, and notes Leibniz's claim that they matter to the science of minds as much as invisible particles matter to natural science. The entry does not mention the donkey, so the pairing comes from the encyclopedia article on the donkey and not from the philosopher's own words as I have read them. The Preface to his New Essays on Human Understanding is where the idea is stated most strongly. ↩︎
¹² Spinoza's version of the verdict, in the Ethics, is that such a creature would perish, and that this makes it irrational rather than a paradox. I would have been dead by Friday, and I would have been annoyed to be told it was a failure of reasoning. ↩︎
¹³ I recounted in daylight the next morning, to make sure the hallway had not been lying to me by lamplight. Seven and eight again. In note 1 I said seven and seven, and I would like to claim I miscounted, but I think I counted as I count everything on that kind of evening, by arriving at the number I needed. Note 4 told you to hold it loosely. I am glad one of us listened. ↩︎
¹⁴ The same movement described two ways is, I think, much of what Miller's graph is about. A rat that stops short of the food is not only a rat that wants the food less; it is a rat whose fear has been rising as the distance closes. I would like to claim I can tell my approach steps from my avoidance steps in daylight. I cannot, and I have been trying for most of my adult life. ↩︎
¹⁵ Note 1 promised I would cite the seven steps as evidence, and here is the citation: the evidence was wrong, and being wrong by one step was the finding. I was never the donkey. I was tilted and could not feel it, and I don't want to be flippant about what that implies, because "I secretly wanted the project" is as tidy a story as "I was the donkey," and the floor told me only which way it leaned, not why I was standing on it. ↩︎