The Values You Can't Measure With a Slider: How to Test Alignment on Money, Family, and Time

Nobody has ever ended a relationship because of an Openness score.

People end relationships because one of them wanted to send $800 a month to their mother and the other one thought that was a loan, not a gift. Because one of them assumed Saturday was for the two of them and the other assumed Saturday was for the group chat. Because "we'll figure out the holidays later" turned out to mean "I will be in Ohio for nine days every December and you will accommodate that."

Those aren't trait mismatches. Traits are how you move through the world — how fast, how loudly, how tidily. Values are what you move toward when you can't have everything. And almost every meaningful relationship decision is a situation where you can't have everything.

This is the gap most compatibility tooling walks straight past. We're very good at measuring the adverbs and very bad at measuring the nouns.

The evidence for taking values seriously

Start with the uncomfortable finding. In the largest effort of its kind, researchers pooled 43 longitudinal couples datasets and used machine learning to see what actually predicts relationship quality. Relationship-specific variables — things like perceived partner commitment, appreciation, sexual satisfaction — were roughly "two to three times as predictive as individual differences," in the lead author's words.[1] More striking: one partner's individual characteristics explained only about 5% of the variance in the other partner's relationship satisfaction.[1]

Read that carefully, because it's easy to over-read. It does not say personality is useless. It says that who you are, measured in isolation, is a weak predictor of how your partner feels about the relationship. What's happening between you matters far more than what's listed on your profile.

So what is "between you"? Partly conflict behavior, which we've covered elsewhere. And partly the thing conflict is usually about: the allocation of scarce resources.

Money is the loudest example. Analyzing longitudinal data from more than 4,500 couples in the National Survey of Families and Households, Dew, Britt and Huston found that financial disagreements were the strongest type of disagreement predicting divorce — stronger than disagreements about household tasks, time together, or in-laws. Notably, when financial disagreements were in the model, financial well-being was no longer associated with divorce. It isn't how much money you have. It's whether you fight about it.

There's a quieter finding worth sitting with too. Couples tend to be strongly similar on demographics and attitudes, only moderately similar on values and interests, and weakly similar on personality. We don't actually pair up by trait profile. We pair up by worldview and circumstance — and then discover, months in, that "moderately similar on values" leaves an enormous amount of room for surprise.

And value similarity isn't decorative. Research measuring perceived value similarity across different relationships found that people perceive the highest value similarity with their intimate partner, and that similarity with a partner predicts life satisfaction.[2]

Why the slider fails

Here is the standard way apps and inventories approach values:

How important is family to you? (1 = not at all, 7 = extremely)

Everyone picks 6 or 7. The person who hosts a weekly dinner picks 7. The person who calls twice a year and means it sincerely picks 7. The person who would decline a promotion to stay near their sister picks 7, and so does the person who moved 3,000 miles away and feels fine about it.

Three problems stack up at once.

Sliders measure endorsement, not ranking. Values only become informative under scarcity. Asking "do you value family?" is like asking "do you value oxygen?" Asking "would you move cities for your parents if it cost your partner their job?" is a measurement.

Sliders invite the flattering answer. We've written about the honesty problem before — people answer inventories in the direction of the person they'd like to be. Nobody rates themselves low on generosity.

Sliders don't carry units. "I'm generous with family" is a sentiment. "$800 a month, indefinitely, no discussion" is a fact. The gap between those two is where most of the damage lives — and it's close cousin to the vocabulary problem: two people pick the same number and mean completely different lives.

Scenario-based assessment exists precisely because of this. In work and selection psychology, situational judgment tests present concrete dilemmas rather than self-descriptions, and research developing SJT measures of personality notes they are less prone to faking than classical self-report measures.[3] You can tell a questionnaire you're flexible. It's harder to tell a specific dilemma you're flexible, because the dilemma makes you pick.

Three resources, not twenty topics

You don't need a sixty-item values inventory. Almost everything that matters routes through three scarce resources:

• Money — what it's for, who decides, what's shared, what secrecy is permitted.
• Obligation — who has a claim on you that you did not choose, and how strong that claim is.
• Time — who gets your weekends, your evenings, your energy at the end of a bad week.

Most "value conflicts" in relationships are really allocation conflicts in one of these three, wearing a costume.

How to build a scenario question that actually works

Four rules.

1. Force a trade-off. A good scenario has no free lunch. If both options are fine, it's a preference, not a value.

2. Attach numbers and dates. "A lot of money" is unmeasurable. "Four thousand dollars" is measurable. "Often" is unmeasurable. "Every other Sunday" is measurable.

3. Ask for the default, then the exception. Values live in defaults; character lives in exceptions. "What normally happens?" followed by "what would have to be true for that to change?" tells you more than either question alone.

4. Ask who decides, not just what's decided. Two people can agree on an outcome and completely disagree on who had the authority to choose it. Decision rights are a value.

The scenarios

Use these as raw material. Not all at once — you are not conducting a deposition.

Money

1. The windfall. You get $20,000 unexpectedly. Walk me through where it goes, in order. (Listen for: debt vs. experience vs. security vs. family. Order matters more than content.)
2. The asymmetry. One of you earns twice what the other does. What does dinner out look like? What does rent look like? (This is the proportional-vs-equal split question. There's no right answer; there's a right match.)
3. The loan that isn't. Your sibling asks for $3,000 and has not repaid the last $1,500. What happens, and who gets a say? (Boundaries, and whether a partner's opinion is input or veto.)
4. The private account. Is there an amount you could spend without mentioning it? What's the number? (Everyone has a number. People who claim they don't are either unusually merged or not answering.)

Family and obligation

5. The December problem. Describe the next five Decembers. Where are you, with whom, for how many days? (Forces a concrete answer out of "we'll alternate.")
6. The caregiving question. A parent needs daily help for two years. What actually happens — your house, their house, paid care, a sibling? (Among the highest-stakes value questions almost nobody asks before year three.)
7. The disapproval test. Your family dislikes a major choice you've made. What changes about your behavior? (Measures the weight of inherited approval.)
8. The relocation. A job offer doubles one income but moves you four hours from one person's family. Who makes that call, and how?

Time

9. The weekend default. Describe a normal Saturday if nothing special is planned. Then describe the ideal one. (The gap between those two is the negotiation you'll be having for years.)
10. The 60-hour month. Work demands 60-hour weeks for one month. What gets protected no matter what? (Reveals the true top of the stack.)
11. The friendship budget. How many evenings a month belong to people who aren't me? (Say the number out loud. Watch the flinch.)
12. The recovery question. After a draining week, do you refill alone, together, or in a crowd? (The energy-budget question — not a value exactly, but it governs how all the other time gets spent.)

Why your avatar should be arguing about these

Here's the practical problem: you cannot run twelve scenarios on a first date without becoming memorable for the wrong reasons. And you shouldn't have to. Most of these conversations are only worth having with someone who's already plausible.

This is exactly the work a psychometric avatar should be doing. Not comparing Agreeableness percentiles — comparing answers to dilemmas. Given the Joel et al. finding that individual traits predict so little of a partner's satisfaction,[1] a matching system that stops at trait correlation is optimizing a weak signal. The stronger signal is behavioral and situational, and it's generated by making two models argue about the windfall, the December problem, and the 60-hour month — across hundreds of variations, including the ones you'd never think to raise.

An avatar can have that fight two hundred times before breakfast. You get the summary: you two agree about money in general and disagree sharply about who has a say in family lending. That's a far more useful thing to know on date two than a similarity percentage. If you want to see what that looks like in practice, you can create your avatar and let it do the unglamorous part.

The date-three script

Short, usable, non-interrogative. Four moves.

Frame it. "Can I ask you a weirdly specific question? I've decided the vague ones are useless."

Ask one money scenario. The windfall is the friendliest: nobody feels judged describing where $20,000 would go. Listen to the order.

Ask one obligation scenario. The December problem works because it sounds logistical and is actually structural. "Where are you next Christmas?" is small talk. "Where are you the next five Christmases?" is an audit.

Ask one time scenario. The weekend default. Then the ideal weekend. Then stop.

Close the loop. "Okay, say the honest version — where do you think we'd clash?" This is the move most people skip, and it's the one that works. You're not looking for agreement. You're looking for whether they can name a disagreement without getting defensive, because that's the skill that determines whether the next twenty disagreements are survivable.

Three questions and a closing line. Ten minutes. It will tell you more than any slider you've ever dragged.

One caution, because this is where values work goes wrong: a mismatch on a single scenario is data, not a verdict. Plenty of durable couples disagree about money and have simply built a working treaty around it. What you're actually testing is whether the two of you can locate the disagreement, say it in plain language, and stay interested in each other afterward. The scenarios are a flashlight, not a scorecard.

The slider asks what kind of person you'd like to be. The scenario asks what you do on a Tuesday when there isn't enough to go around. Only one of those will still be true in four years.

Sources

1. Machine learning predicts satisfaction in romantic relationships (coverage of Joel et al., PNAS 2020) — TechXplore
2. Value similarity across relationship contexts and life satisfaction — Frontiers in Psychology
3. Situational Judgment Tests as a method for measuring personality: Development and validity evidence for a test of Dependability — PLOS ONE