Recommender System Type
Test recommendation slates for repetition, novelty drift, cold start, trust collapse, and abandonment.
Risks Determina tests
Recommender Behavior Trials focus on release risks that aggregate metrics often hide:
- slate repetition
- novelty drift
- cold start weakness
- trust collapse
- abandonment by impatient or exploratory users
- candidate behavior changes relative to a baseline
Supported workflows
Recommender is the deepest public System Type today. It supports native HTTP and schema-mapped HTTP setup, hosted audit and compare workflows, and platform/core-owned generated test case workflows.
shell
determina run-swarm --project-id <project-id> --system-version-id <system-version-id> --system-type recommender --brief "test trust collapse and weak first-slate behavior"
Evidence language
Recommender evidence should describe cohorts, slates, click/skip/abandon behavior, repetition, novelty, cold start, and trust risks in the language release teams use to decide whether a candidate is safer than the baseline.