Home › Conflict Resolution › Interest › Interest Scoring Methodology
Interest Scoring Methodology
People often assume the worst about the other side's motives, then treat the assumption as settled. The ISE tries to automate the principles of conflict resolution, as taught in Applying "Getting to Yes" Principles.
In conflict resolution, we are taught to "focus on interests, not positions," which lets us come up with elegant solutions that meet the valid interests of each side. The ISE automates this process by using the ReasonRank algorithm to identify the most likely interests of those who agree and disagree with each belief, rank the validity of each interest or motivation, and identify the side with the most valid interests that would be benefited, using a cost-benefit analysis.
An interest is written as a need with a subject ("renters need housing they can afford", "current homeowners need to protect the value of their homes"), so one sentence can be argued two ways: is this really what drives the side, and should it be honored? Those are different questions with different evidence, and they are never allowed to borrow from each other.
| Score | The question | What settles it |
|---|
| Linkage Accuracy (0 to 100) |
Is this really why they hold the position, or a cover story? |
Behavior. The candidate motive that best predicts what the side actually does wins, argued pro and con like any claim. |
| Interest Validity (0 to 100) |
Even if it is real, should it be honored, and how far? |
Objective criteria (reciprocity, the universal test, and the rest below), argued pro and con. Never the holder's power. |
| Drive (0 to 100) |
What is actually powering this side? |
Linkage Accuracy × Validity ÷ 100. Computed, never typed. The engine uses it to name the one conflict that matters. |
Linkage Accuracy: the alibi test
A neighborhood association opposes a 60-unit apartment building. Its letter says traffic. Check the record. Last year it supported a supermarket that generates more car trips than the apartments would. Two years ago it opposed a townhouse project that came with a parking garage. "Traffic" predicts one of those three positions. "Current homeowners need to protect the value of their homes" predicts all three. The second hypothesis wins, not because anyone read the members' minds, but because it explains the behavior with fewer exceptions.
Run the same test on the other side. The developer's brochure says "housing for families." The permit application is 80 percent studios. The brochure loses to "the builder needs the project to pencil out at current land prices." Neither finding is an insult. Both are better predictions, and a prediction that fails is a motive you have misread.
That is the whole method. Several candidate interests are proposed for each side, each a claim, and the debate under each is about behavior:
| Raises it | Lowers it |
|---|
| Votes, spending, donations and effort go where this interest says they should |
The side acts against what this interest would predict |
| Members name it themselves, in surveys and on the record |
A rival interest predicts the same record with fewer exceptions |
| The same group takes the same side in other debates where only this interest is at stake |
The behavior has another documented cause, so it is not evidence for this one |
| Analysts with no stake in the outcome find the same pattern |
The attribution rests on rhetoric alone, with no behavioral record either way |
Within a side, interests rank by Linkage Accuracy, so the top row is the best explanation of the side's behavior, not the most flattering one. The rule applies with equal suspicion in both directions. The share of a side that each interest actually drives (Prevalence) is recorded beside it from surveys, with a source; it is a reading, not a score, and it multiplies into nothing.
Interest Validity: the mirror test
"I need the street in front of my house kept free of other people's cars." Real? Certainly; people fight over it every day. Valid? Hold it up to a mirror. Would you accept the same rule enforced on you by your neighbor? And if everyone claimed the street in front of their house, where would anyone park? The interest is real and fails anyway. "I need to get to work in under an hour" passes both tests. Only one of these is a need a resolution should be built around, and it is not the louder one.
| Criterion | The question | An interest fails it when |
|---|
| Reciprocity |
Would the side accept this interest honored for its opponents? |
It wants a rule for others it would reject for itself |
| Universal test |
What happens if everyone pursues it? |
Universal pursuit makes it impossible for anyone, or guarantees perpetual conflict |
| Asks no loss from others |
Can it be met without someone else losing? |
It is satisfied only at another party's expense, or it is a claim on other people's property or choices |
| Alternative satisfaction |
Can the underlying need be met another way? |
Substitutes exist and the side rejects them, which usually means the stated need is not the real one |
| Scope and reversibility |
How many people, how badly, and can it be undone? |
Ignoring it costs little, spreads thin, or is easily reversed |
| Unmet and shared |
Is the need actually unmet, and does anyone outside the asking group recognize it? |
The status quo already satisfies it, or it is held by few and recognized nowhere else |
Validity is never set by who holds the interest, how many of them there are, or how much power they have. And it never absorbs feasibility: how hard an interest is to satisfy, and what each side can do without a deal, are scored on their own track. Cross either line and the engine rebuilds "might makes right" inside the machine that was designed to replace it.
Interests on Maslow's ladder
Maslow sets the opening price; the arguments do the trading. Every interest starts from the band of the need it serves, and the argument tree moves it from there. The same need can appear on several rungs depending on how it is written: "my family needs to eat" is physiological; "my family needs to eat at the restaurant our neighbors can't afford" is esteem, and the mirror test treats them very differently. Each row links to the pages where that kind of interest is already being argued.
| Rung | Opening prior | Interests written as needs | Where they are argued |
|---|
| Physiological |
85 to 100 |
People need food, water, shelter, sleep and medical care. Workers need wages that cover them. Patients need treatment they can reach. |
Health · Housing |
| Safety |
70 to 85 |
People need protection from violence, crime and war. Households need an income they can count on, savings that hold their value, and a legal order that applies tomorrow the way it did today. Financial interests sit here: "homeowners need to protect the value of their homes" is a safety need, and so is "renters need a lease that cannot be cancelled on a whim." |
Security · Safety · Privacy · Rule of law |
| Belonging |
50 to 70 |
People need a family, a community, a congregation, a country and a team to belong to. Every form of group solidarity lives on this rung: national, religious, cultural, racial, ethnic, regional, partisan, and the sports team. The need to belong is valid; what the group does to outsiders is a separate interest, scored separately, and often below the floor (see the last row). |
Family · Faith · Religion · Culture · Patriotism · American pride as a valid interest |
| Esteem |
40 to 60 |
People need to be respected, heard and credited for what they do. Officials need to be seen as honest; voters need to be answered; workers need their effort recognized. Status needs sit here too, and the mirror test is where they usually fail: status is scarce by definition, so "I need to rank above them" cannot be universal. |
Accountability · Transparency |
| Self-actualization |
30 to 50 |
People need to grow, create, learn, worship and choose for themselves. Low prior, not low value: these needs lose to survival when the two collide and win on their own arguments when they do not. |
Freedom · Freedom of religion · Education |
| Curiosity and play |
30 to 50 |
Maslow added these late: people need to understand, to be entertained, to feel excitement. This rung explains motives the political ones miss. Some people share a conspiracy theory because it is thrilling and makes them one of the few who see, which is belonging and esteem riding on a manufactured premise. Some people vote for disruption because watching the system break is more interesting than watching it work. The need is real and scores on this rung; the vehicle is scored on the belief page, where a false premise scores false. |
No pages yet. Candidates: entertainment as a motive, disruption as a motive, the excitement of conspiracy theories. |
| Below the floor |
0 to 20 |
Not a rung. These are interests that fail the criteria outright: domination (our group needs the other group kept down), tribal winning (we need them to lose), pretext (a belonging interest worn as a disguise for one of these), and manufactured needs that exist only if a false claim is true. Racial solidarity as "we belong together" sits on the Belonging rung; racial solidarity as "they must not" sits here, and the engine keeps the two rows apart so the valid half cannot smuggle the invalid half across. |
Outgroup and tribal bias · Fanaticism · Patriotism as the last refuge of a scoundrel · Status-quo bias |
The bands are priors, not verdicts, and not a lookup table. A level reflects how widely and urgently a need is shared, which is one argument for validity. A famine is physiological. So is a demand for the last seat on the lifeboat, and that one fails reciprocity the moment someone else wants the seat. An esteem need finishes above a safety need when its arguments are stronger.
Bad motives are scored, not deleted
The most useful row in any interest table is the one that is real and illegitimate. In the seeded government-ethics corpus (the stock-trading ban for members of Congress and the president), the opponents' list includes "some officeholders want to keep using their position and their families to make money." The engine agrees it is real: its Drive ties for highest on that side (57 on the wiki's 0 to 100 scale, 0.57 in the workbook). Its validity is the lowest on the page (38). So it ranks first as an explanation of behavior and last as something a resolution should honor. That is not a contradiction. That is the information: it tells you what the side will actually fight for, and that giving it to them is not how the conflict ends.
The same move must be made against your own side, or it is propaganda. "Our side needs to win the next election" belongs on both lists of any partisan belief, with the same low validity on both, because it fails reciprocity identically in both directions. A motive analysis that finds pretexts only among the people you disagree with has not analyzed anything. The outgroup bias page lists the detection signals.
Where the scores go: one orange, two sisters
Fisher and Ury's oldest example still does the most work. Two sisters want the same orange. The fair-minded parent cuts it in half. One sister squeezes her half for juice and throws away the peel; the other grates her peel into a cake and throws away the fruit. Both got half of what they needed when both could have had all of it. Positions ("I want the orange") collided. Interests ("I need juice", "I need peel") never did.
The engine scores that story, and it scores the split as the worse outcome. A candidate solution earns, for each interest it satisfies, the fraction of the interest it meets (0 to 1) times the interest's validity (as a fraction), summed per side. The overall score is 2 × S × O ÷ (S + O), the harmonic pairing of the two side totals. Illustrative numbers, with both needs at validity 80: half an orange each gives S = 0.5 × 0.8 = 0.40 and O = 0.40, total 0.40. Peel to the baker and juice to the drinker gives S = 0.80 and O = 0.80, total 0.80. Twice as good, and the engine can see it without anyone being clever. Give the whole orange to the baker and the total is zero, because a solution that satisfies only one side has resolved nothing. Splitting the difference is what people do when they have not looked underneath the positions; the score says so.
| Output | How the scores set it |
|---|
| Shared interests |
Interests on opposite sides whose statements say the same thing (paired by similarity of the sentences, then confirmed or disputed by people), with validity at or above the Resolution Floor of 70 on both sides. The pair's combined validity is the harmonic mean, so a lopsided pair sinks. In the stock-ban example the shared interest reads: both sides want officials who are not, and do not appear to be, for sale. Compromise is built here first. |
| Primary Conflict Pair |
The highest-Drive interest on each side that is not shared, staged head to head. This replaces the flat list of "conflicting interests" with the one disagreement that is actually driving the debate. In the seeded corpus the pair is "citizens need government decisions made on the merits rather than on what enriches the decider" against "officials and their families need the same property rights and privacy as other citizens" (both at Drive 32). Drive uses validity on purpose: a pretext cannot be the primary conflict, and neither can a saintly interest nobody is moved by. |
| Solution scores |
The orange arithmetic above. Interests below the floor earn a solution nothing even when it satisfies them, which is why buying off the "keep making money" interest scores zero. No interest counts extra because its holders are powerful. This is the arithmetic under inventing options for mutual gain. |
| Argument importance |
A true fact that matters to nobody weighs nothing. Every reason on a belief page speaks to one or more interests, and its importance is set by the most valid interest it genuinely bears on: validity × how directly the reason bears on it, taking the largest. "The bill misspells the sponsor's name on the cover" can be perfectly true and still carry no weight, because no valid interest turns on it. |
| Who gains, who pays |
Every cost and benefit in a cost-benefit analysis names the interest that collects it or pays it. A cost always has an address. Once it does, "is it worth it?" can be asked of each party separately instead of in the aggregate, where the losers disappear. |
The full pipeline, including how the engine classifies a dispute as factual, linkage, values or mixed, is on Conflict Resolution Scoring. It reads interest scores and never writes them.
Where the table lives
Every belief page carries an interest table for each side, filled in with the Belief Analysis template: the interest as a need with a subject, its value tag from the Values Framework, what it is measured by, Prevalence with its source, and the three scores read from the interest's own debates. An interest that is contested, or shared by several beliefs, gets its own page from the Interest Evaluation template; American pride is a worked one. One interest, one page, and every belief that lists it reads the same scores. A row nobody has argued sinks, and its cells stay blank rather than showing a guess.
To contribute: propose an interest for either side, written as a need with a subject, with the same suspicion for your own side as for the other; add a reading or a measure; argue the alibi with behavioral evidence or a rival interest that predicts better; argue the validity against one of the criteria; confirm or dispute a cross-side pairing, and propose a solution that satisfies valid interests on both sides. No step asks you to type a score.
See also
| Interest |
What an interest is, and how it differs from a position, a value and a belief |
| Motivation |
Stated and unstated interests, and why a man's salary shapes what he can understand |
| Focusing on Interests, Not Positions |
Why interests, not positions, are where the overlap hides |
| Getting to Yes and Applying its principles |
The book, scored, and how the platform enforces its four pillars |
| Conflict Resolution Framework |
Interests, not positions, as the platform's foundation |
| Conflict Resolution Scoring |
The full pipeline: shared ground, the conflict pair, value conflicts, compromise candidates |
| Interest Evaluation template |
A page for one contested or shared interest |
| Values Framework |
How values tag and rank interests |
| Linkage Scores |
The general method for scoring whether one claim bears on another |
| Insisting on Objective Criteria |
The criteria validity arguments are judged against |
| Cost-Benefit Analysis |
Where each cost and benefit names who gains and who pays |
Contact me to map the interests behind a belief you care about. | GitHub for the engine.
Comments (0)
You don't have permission to comment on this page.