TL;DR
- 01Fisher and Ury split positions from interests in 1981. Forty years later, almost no software implements the distinction.
- 02In the ACO, Position is a
Claim(type=Position); Interest is a separate primitive with a mandatoryderivationchain. - 03Interest inference is interpretive; the kernel tiers it: stated / inferred-with-citation / analyst-reconstructed.
- 04Better calibrated than wrong with false precision.
Fisher and Ury split positions from interests in 1981. Forty years later, almost no software implements the distinction. Here is what happens when you do — and why the kernel tiers interest inference rather than guessing.
The distinction, with examples
- ▸Position: "I want a 20% raise." Interest: compensation that reflects contribution.
- ▸Position: "I want a private office." Interest: workspace where focus is possible.
- ▸Position: "The corridor must be closed." Interest: a guarantee that supply lines cannot be cut.
- ▸Positions are narrow and oppositional. Interests are broader and, often, compatible across parties.
Why the graph matters
The same text that produces a position also contains the interest, almost always. But the interest has to be inferred — from surrounding context, from what the party did not say, from what the other side's position would mean for this party's situation.
A system that only extracts explicit claims will miss every interest, and every resolution that lives underneath the positions.
How the ACO handles it
- ▸
HOLDS_INTERESTis a typed edge from Actor to Interest. - ▸Interest carries
derivation(source claims, behavioural evidence, declared positions). - ▸Confidence tier exposed explicitly to the user.
- ▸Graph queries look for interest overlaps across parties — where resolution lives.
SOURCES
