concept

Recommendation stability

recommendation-stabilitystatus · Active research2026-07-24

Whether the same inputs produce the same recommendation, and whether unstable outcomes trace to unstable requirement association.

Question
Are recommendation outcomes stable across repeated runs, and when they are not, is the instability in the recommendation or upstream in how requirements are associated?

Research components

Answers business questions

  • Why did we disappear after follow-up questions?
  • What must become true to survive evaluation?
  • How do we measure whether our position improved?
  • How do AI systems evaluate and recommend vendors?

Recommendation stability studies whether the same question and the same requirements produce the same recommendation across repeated runs. When outcomes move, the instability may sit in the recommendation itself or upstream in how the evaluator associates a requirement with a vendor.

The core question is how stable recommendation outcomes are, and where any instability originates. What remains open is how much variation is normal and what separates unstable requirement association from an unstable recommendation.


This object references

Referenced by

Authors

Upstream Zero

Machine rendering

recommendation-stability.json