Daniel sent us this one. He's asking what compromise actually means — not in the greeting-card sense, but structurally. What frameworks actually produce fair compromises when two people have conflicting requirements, and he's grounding it in something painfully concrete: a shared housing search. Two people, one apartment, incompatible wishlists. What do you do? And he's asking for heuristics — actual tools, not just philosophy. Which is good, because philosophy doesn't pay the security deposit.
It's a question that's only gotten sharper lately. I was looking at the May Zillow data — median rent in the top twenty US metros is up eight percent year-over-year. More people are sharing housing out of necessity, not choice, which means more of these two-sets-of-requirements collisions. The stakes are higher.
Let's step back and ask: what does compromise actually mean, mathematically? Because most people treat it as a soft skill — be reasonable, don't be difficult, meet in the middle. But underneath that, there's a formal structure that nobody talks about at the kitchen table.
And I think there are three layers here. Layer one is the colloquial version — compromise as "everyone gives something up." It's framed as loss. Layer two is the game-theoretic version — compromise as reaching some point in what's called the bargaining set, the space of possible agreements that are better for both parties than no deal at all. And layer three is the constraint-satisfaction version — finding a feasible solution where no single person's constraints are fully met, but the overall configuration works.
That third one is interesting. It reframes compromise from "we both lost" to "we discovered a configuration that didn't exist in either of our heads before we started.
And that's the shift I want to make. When you're searching for a shared apartment, you're not really negotiating — or you shouldn't be. You're jointly optimizing over a multi-attribute utility space. Each person has a vector of preferences: price, location, square footage, amenities, lease length, pet policy, roommate gender preference, noise tolerance, commute time. The goal isn't to win concessions from the other person. The goal is to find a single point in that high-dimensional space that maximizes joint satisfaction.
Yet most people approach it like a tariff negotiation. "I'll give you the neighborhood if you give me the extra bedroom." Which frames the whole thing as win-lose before you've even looked at the map.
The core misconception is that compromise means splitting the difference. Meet in the middle on price, meet in the middle on location. But splitting the difference often produces what economists call a Pareto-inefficient outcome — a situation where both people are equally unhappy, but a better solution exists that neither considered.
Define Pareto efficiency for the person who didn't spend their weekend reading welfare economics.
A state is Pareto-efficient if there's no alternative that makes one person better off without making someone else worse off. In housing terms: if you and your roommate settle on an apartment, and there exists another apartment you could have taken where one of you would be happier and the other would be at least as happy, your compromise was Pareto-inefficient. You left value on the table.
Which is the housing equivalent of ordering a pizza that neither of you wanted because you couldn't agree on toppings, when there was a third topping neither of you had tried.
That's actually a perfect analogy. And it happens constantly. I want to walk through a real example from this month. Two roommates in San Francisco. Person A works in SoMa and wants to live within a fifteen-minute walk of the office — hard constraint — with a budget of twenty-two hundred dollars max. Person B works remotely and wants a two-bedroom with a dedicated office space, budget eighteen hundred max. The traditional compromise: find a one-bedroom in a middle neighborhood for two thousand. Person A gets a twenty-minute commute instead of fifteen, Person B loses the office, both are paying near their ceiling, and neither is happy.
That's the "meet in the middle" outcome. Both equally miserable.
Here's the Pareto-improving alternative they actually found. A two-bedroom in a neighborhood slightly farther out — twenty-minute commute for Person A, which is still under their twenty-five-minute walk tolerance they hadn't stated explicitly — with a den that Person B uses as an office. Total rent: twenty-one hundred. Person A pays eleven hundred, Person B pays a thousand. Person A gets the commute they can live with, Person B gets the office, both pay less than their maximum.
The key thing is — this alternative wasn't on either of their initial lists. They discovered it by expanding the search space rather than negotiating within the narrow set they each brought to the table.
Which brings us to the formal framework. To really understand what's happening here, we need to borrow a tool from game theory — the Nash Bargaining Solution.
Not just Russell Crowe shouting about Adam Smith.
Nash published "The Bargaining Problem" in nineteen fifty, and it's elegant. You have two players, each with a utility function — a way of mapping outcomes to how satisfied they'd be, on some numerical scale. You also have a disagreement point — what happens if no deal is reached. For housing, the disagreement point is each person finding their own place separately. The Nash Bargaining Solution says: the fairest agreement is the one that maximizes the product of each person's gain over their disagreement point.
Mathematically — and I know you want to write this on an imaginary whiteboard — it's argmax of U-one minus D-one, times U-two minus D-two.
Maximize the product of the gains from cooperation. Not the sum, which is important — the product. Because the product penalizes extreme inequality. If one person gains a hundred and the other gains one, the product is a hundred. If both gain ten, the product is also a hundred. But if both gain fifty, the product is twenty-five hundred. The Nash solution inherently favors balanced outcomes without requiring equal sacrifice.
It's not "we both give up the same amount." It's "we arrange things so that the marginal pain of the next concession is roughly equal for both of us.
That's the insight. Apply it to housing. Let's say Person A values location on a zero-to-one-hundred scale, and Person B values square footage on the same scale. The disagreement point — finding separate apartments — gives Person A a utility of fifty and Person B a utility of forty. The Nash solution finds the apartment where, if you plotted both utility functions against each other, the trade-off rate equalizes. You don't just split the difference on the map. You find the point where moving one block closer to work costs Person B as much utility as it gains Person A.
That point might not be the geographic midpoint. It might be much closer to one person's ideal if their utility drops off a cliff past a certain distance.
Suppose Person A's commute utility is basically flat from zero to twenty minutes — they're happy walking, they listen to podcasts — but drops sharply after twenty-five minutes because now they need transit. And Person B's square footage utility increases steadily with every additional square foot because they're home all day. The Nash solution might give Person A a twenty-two-minute commute and Person B an extra hundred square feet, even though the purely geographic compromise would be a different location entirely.
The math is doing something subtle. It's not measuring sacrifice in dollars or blocks. It's measuring it in utils — actual experienced satisfaction — and equalizing at the margin.
Here's where most real-world compromises fail. People don't articulate their utility functions, even to themselves. They state positions — "I want to live in this neighborhood" — rather than interests — "I want a commute under twenty-five minutes where I can walk and listen to podcasts." Positions are binary and brittle. Interests are continuous and negotiable.
The San Francisco example you gave — Person A's position was "fifteen-minute walk to SoMa." Their actual interest was "walkable commute under twenty-five minutes." Once they surfaced the interest, the solution space expanded enormously.
Which is heuristic number one. I call it "Expand the Pie." Before you negotiate any trade-offs, spend thirty minutes generating at least ten new search criteria or apartment types that neither of you initially considered. Sublets, co-living spaces, different neighborhoods, different unit configurations, mother-in-law units, rent-to-own, house-sitting arrangements. The goal isn't to find the answer in that thirty minutes. The goal is to expand the feasible set so you're not negotiating over a artificially constrained menu.
Because the menu you each walked in with is almost certainly too narrow. You've each done some Zillow browsing, you've each formed an image of "the kind of place we'd get," and those images are anchoring you.
The anchoring effect is brutal here. The "spotlight effect," too — coined by Gilovich and Savitsky in nineteen ninety-nine. People overestimate how much others notice and care about their specific choices. So Person A thinks "everyone will judge me if I live in that neighborhood," and Person B thinks "everyone will think I'm cheap if I suggest a lower budget." Neither of those judgments is actually happening at the scale they imagine, but the spotlight effect makes the negotiation feel higher-stakes than it is.
Heuristic one: expand the pie. Generate options neither of you had on your radar. What's heuristic two?
Weighted Priority Ranking. This is the one that transforms the conversation from subjective to transparent. Each person independently ranks their top five criteria by importance — one to five — and then assigns a weight, one to ten, to each criterion. You then combine these into a joint score for every apartment candidate.
Walk me through the math.
For a given apartment, you calculate a score for each person: the sum of weight times rank for each criterion. So if Person A weights location as a ten and the apartment scores a four out of five on location, that's forty points. Then you either average the two scores or — and I prefer this for fairness — you take the minimum of the two scores. Maximizing the minimum score is a Rawlsian approach: it ensures no one is left deeply unhappy while the other person is ecstatic.
The minimum-of-scores approach is interesting because it prevents the tyranny of the enthusiast. If Person A is a ten-out-of-ten happy and Person B is a two, the apartment scores a two. You can't paper over one person's misery with the other person's joy.
And what this process surfaces is hidden consensus. I've seen this repeatedly — both people independently rank "safety" as their number one criterion but never discussed it because it felt too obvious. Both people rank "natural light" highly. Both people care about in-unit laundry more than they admitted. The weighted ranking makes these convergences explicit, which reduces the felt conflict. You realize you agree on more than you thought.
For the areas where you diverge, the weights tell you where to spend your negotiation energy. If Person A weights location as a ten and Person B weights it as a three, you know location is mostly Person A's domain — and Person B gets disproportionate say on whatever they weighted as a ten.
Which leads to heuristic three: the eighty-twenty rule. Identify the twenty percent of criteria that drive eighty percent of utility for each person. Then explicitly label which criteria are dealbreakers versus preferences. Person A might have in-unit laundry as a dealbreaker because of a medical condition. Person B might have a hard requirement for ground-floor access. These aren't negotiable, and that's fine — you just need to know them upfront rather than discovering them three weeks into the search when someone rejects an otherwise perfect apartment.
The failure mode here is treating all criteria as equally important, which produces a kind of analysis paralysis where every apartment fails on some dimension and you can't figure out why nothing works.
The dealbreaker-versus-preference distinction is crucial because it prevents what I call the "veto cascade." Person A vetoes an apartment because the kitchen is small. Person B vetoes the next one because it's on a busy street. Person A vetoes the next because the bathroom has a pedestal sink. None of these were dealbreakers — they were mild preferences — but in the moment, each person feels pressure to advocate for themselves, and it's easier to say no than yes.
The veto cascade is the housing-search equivalent of a filibuster. Nothing gets through.
The antidote is to agree in advance: you each get three dealbreakers, maximum. Everything else is a preference that contributes to the score but can't singlehandedly kill an option. If the weighted score is high enough, you take the apartment even if the kitchen is small.
That's a constraint that forces honesty about what you actually can't live with.
Heuristic four: Temporal Escrow. Agree to the compromise now with a built-in review date — say, six months. This reduces the perceived permanence of the decision. Behavioral economics research on status quo bias shows that people resist changes more when they believe the decision is irreversible. If you know you can revisit the arrangement in six months, the emotional stakes drop considerably.
In practice, most of these reviews don't actually lead to a change. Six months in, you've settled into the place, the commute isn't as bad as you feared, the small kitchen turns out to be workable. The review clause is a psychological safety valve more than a logistical one.
There's a twenty twenty-four study from the Journal of Experimental Social Psychology that's directly relevant here. Couples who used structured preference elicitation — basically a formal version of what we're describing — reported forty percent higher satisfaction with joint decisions compared to couples who just "talked it out." The structure itself improved the outcome, independent of what they actually decided.
Forty percent is not marginal. That's the difference between "I guess this is fine" and "we made a good decision.
I think the mechanism is that structured elicitation forces explicit communication about values. You can't just nod along and then resent the outcome later. You wrote down your weights. You agreed on the dealbreakers. The process creates accountability to your own stated preferences.
Which is a knock-on effect worth naming: the process of building a joint utility model actually improves the relationship. Not just the housing outcome — the friendship or the partnership. You learn how the other person thinks about trade-offs. You learn what they value that you didn't know they valued. The apartment is almost a side effect.
Heuristic five: the Third Option. When two people are stuck on a binary choice — "your apartment or my apartment" — force a third option that neither proposed. This breaks the anchoring effect and often reveals a superior solution.
The binary trap is real. You each did your Zillow research independently and each found "the one." Now you're locked into a choice between Apartment A and Apartment B, and the conversation becomes adversarial because advocating for your apartment means arguing against the other person's judgment.
The third option doesn't even have to be a real apartment yet. It can be a hypothetical — "what would an apartment that gives Person A the commute and Person B the office look like?" — and then you search for that. The act of defining the third option forces you to articulate what you actually want, not just defend what you already found.
The spotlight effect makes people overvalue the options they've already considered. You found Apartment A, you've mentally moved in, you've pictured your furniture in the living room. Letting go of that feels like a loss, even if Apartment C is objectively better for both of you.
I want to address something that's been implicit in all of this. What do you do when preferences are genuinely incompatible? Not just "I prefer east-facing windows and you prefer west" — but "I need quiet to function and you host dinner parties twice a week." These aren't negotiable preferences; they're incompatible lifestyles.
This is where the frameworks earn their keep, because they surface the incompatibility early. If you do the weighted ranking and Person A's number one criterion is "quiet" and Person B's number one is "entertaining space," the joint optimization might reveal that no apartment can satisfy both. And that's actually a successful outcome of the process — you've learned that sharing housing isn't viable before you signed a lease.
The disagreement point — finding separate places — is sometimes the Nash solution. If the product of gains from any possible shared arrangement is lower than the product of gains from living separately, the math says don't share housing. And that's not a failure of compromise. That's the framework correctly identifying that compromise shouldn't happen here.
Most people never get to that clarity because they never formalize their preferences. They have a vague sense that "this might not work" but they push forward because housing is expensive and sharing seems like the responsible thing to do. Six months later, the friendship is strained and someone's sleeping on a friend's couch while they find a new place.
Another hard case: power imbalances. Person A has a three-thousand-dollar budget, Person B has fifteen hundred. The naive approach is to split the difference — twenty-two fifty each. But that ignores the fact that Person A's marginal utility for a nicer apartment is probably lower than Person B's marginal utility for saving money. Person A can afford an extra two hundred a month much more easily than Person B can.
The fair split isn't fifty-fifty. It's proportional to something — income, utility, the disagreement point.
The Nash framework would suggest Person A pays more than half — say, eighteen hundred versus twelve hundred — because Person A's gain from the nicer shared apartment, relative to what they'd get alone, is smaller in utility terms than Person B's gain from saving money. The product of gains is maximized when the richer person pays more.
Which feels unfair to a lot of people because we're trained to think of fairness as equality. Equal split, equal sacrifice, equal everything. But equality of input is not the same as fairness of outcome.
This is where I think the misconception about equal sacrifice really does damage. If Person B is stretching to pay fifteen hundred while Person A is comfortable at three thousand, an equal split of twenty-two fifty means Person B is stressed about rent every month while Person A barely notices. That's not fair — it's just symmetrical.
Proportional sacrifice, not equal sacrifice. The person for whom the marginal dollar matters less should bear more of the cost.
Let me bring in one more concept that ties this together. In computer science, a constraint satisfaction problem is one where you have variables, domains of possible values, and constraints that limit which combinations are allowed. A solution is any assignment of values that satisfies all constraints — not necessarily optimally, but feasibly.
Housing search as a SAT solver.
And the insight from constraint satisfaction is that some constraints are harder than others. A hard constraint is binary — satisfied or not. A soft constraint is continuous — you can be more or less satisfied. The art of compromise is converting as many hard constraints as possible into soft constraints, because hard constraints shrink the feasible space dramatically.
"I must have a dishwasher" is a hard constraint. "I strongly prefer a dishwasher but could live without one if the kitchen has good counter space" is a soft constraint. The second version keeps more apartments in play.
Here's the counterintuitive part. People often state preferences as hard constraints because they think it gives them negotiating leverage. "If I say I absolutely need this, I'm more likely to get it." But in a joint optimization, hard constraints are actually self-defeating. They shrink the solution space to the point where no apartment satisfies both people's hard constraints, and you end up at the disagreement point — separate apartments — which is worse for everyone.
The iron fist approach produces the worst outcome for the person using it.
Soft constraints with high weights get you more of what you want than hard constraints that break the search.
Let's distill this into actionable insights. If you're about to start a shared housing search — or any joint decision with real stakes — what do you actually do on Saturday morning?
Step one: before any negotiation, spend thirty minutes independently generating your utility functions. Rank your top five criteria, assign weights, label your three dealbreakers. Write them down. Don't share them until both of you are done, to avoid anchoring.
The writing-down part matters. If you just talk about preferences, you'll influence each other before either of you has articulated your own position clearly.
Step two: use the Expand the Pie heuristic as your first joint activity. Before you compare rankings, before you look at specific apartments, spend thirty minutes generating at least ten search criteria or housing types neither of you initially considered. Sublets, different neighborhoods, different unit types, different lease structures. This expands the feasible set and often reveals a Pareto-superior option before you've had a single argument.
Step three: build in temporal escrow. Agree to a six-month review. This reduces the perceived permanence of the decision and makes it easier to say yes to an apartment that's good-but-not-perfect. Most of those reviews will result in staying put, but the option to revisit matters psychologically.
Step four: if you're stuck between two options, force a third. Define what a solution that satisfies both people's top criteria would look like, then search for that. Don't let the binary trap turn a joint optimization into an adversarial negotiation.
Underlying all of this: treat it as a discovery process, not a concession process. You're not giving things up. You're finding a configuration that neither of you could have specified in advance.
There's a broader point here about AI and the future of this problem. Automated roommate-matching platforms are already using multi-attribute utility optimization — basically the mathematical frameworks we've been describing, implemented in software. Zillow and Apartments.com are incorporating preference elicitation tools. As these get more sophisticated, the machines will get better at finding the Pareto frontier — the set of apartments where no improvement for one person is possible without hurting the other.
The machines can't define your utility function for you. They can't tell you whether you actually care about natural light or you just think you should care about natural light. They can't distinguish between a preference you absorbed from Instagram and a preference that affects your daily happiness.
The human skill becomes more valuable, not less, as the optimization improves. The bottleneck isn't finding apartments — it's honestly articulating what you want. And most people are terrible at that. They don't know their own utility functions. They haven't done the introspection to distinguish dealbreakers from preferences from things they listed because they felt like they should.
The weighted ranking exercise is as much about self-discovery as it is about negotiation. You might sit down to rank your criteria and realize you don't actually care about having a gym in the building — you just assumed you did because every apartment listing mentions it.
That kind of self-clarity makes you a better roommate, a better partner, a better decision-maker in every domain. The housing search is just the training ground.
Before we wrap up, there's one uncomfortable question we haven't addressed. What happens when the disagreement point — no deal, separate apartments — is better for one person than any possible compromise? What if Person A would be happier living alone in a studio than in any shared arrangement Person B would accept?
Then the Nash Bargaining Solution says: don't share housing. The product of gains is negative or zero. The framework doesn't force a compromise where none exists — it identifies when compromise is counterproductive. And that's a feature, not a bug.
Culturally, we treat "we couldn't compromise" as a failure. A moral failing, even. If you were reasonable and mature, you'd find a way to make it work.
That cultural pressure produces terrible housing situations. People force compromises that leave both parties worse off than if they'd just admitted the incompatibility upfront. The frameworks we've discussed are valuable precisely because they can tell you when compromise is the wrong answer.
The hardest sentence in a housing search might be "I think we want different things, and that's okay, and we should probably not live together." But that sentence, said early, preserves friendships. The alternative is a year of resentment followed by a messy move-out.
The frameworks make that conversation easier because it's not personal. You're not saying "I don't want to live with you." You're saying "our weighted priority rankings have a cosine similarity of point three, and the Nash solution converges to the disagreement point." The math takes the sting out.
Nothing says friendship like citing cosine similarity to explain why you can't share a bathroom.
I'm serious, though. Depersonalizing the conflict is one of the underrated benefits of formal frameworks. It's not about who's being unreasonable. It's about whether the preference vectors are compatible.
One more thing I want to name. We've been talking about compromise between two people. But there's a whole other domain here — compromise between a human and an AI system. What does it mean to reach a fair compromise with something that doesn't have preferences in the human sense? That's a question for another episode.
Which we'll actually be tackling in Episode two hundred two. The human-AI compromise problem is fascinating — when your AI assistant proposes something you don't want, what's the right framework for resolving that? Is it negotiation, calibration, something else entirely?
For now, though, the housing search. Two people, one apartment, incompatible wishlists. The answer isn't "be reasonable." The answer is: write down your utility function, expand the pie, weight your priorities, build in a review date, and if you're stuck, invent a third option. And if none of that works, the fairest outcome might be separate leases.
The math doesn't guarantee a happy ending. But it does guarantee that if a happy ending exists, you're more likely to find it.
And now: Hilbert's daily fun fact.
Hilbert: In the seventeen twenties, a naturalist in the Azores claimed butterfly wing colors came from pigments secreted by the insects' diet of nectar. It wasn't until electron microscopy arrived in the twentieth century that scientists discovered the colors are structural — microscopic ridges and layers that refract light — and have nothing to do with pigments or diet at all.
Butterflies have been lying to naturalists for three hundred years.
Structural color is everywhere once you look. Peacock feathers, soap bubbles, the blue in blue jays. None of it is pigment. But thanks, Hilbert.
This has been My Weird Prompts. Thanks to our producer Hilbert Flumingtop. You can find every episode at myweirdprompts.com, and if you want to support the show, leave us a review wherever you listen — it helps other people find us. We're back next week with Episode two hundred two, where we tackle what it means to compromise with an AI system that doesn't have preferences but sure acts like it does. Until then.
Don't split the difference.