RESEARCH OUTLOOK · PLANNED STUDY · NOVEMBER 2026
Experiment A — Stability is not recoverability
Can declining recoverability be measured independently of conventional early-warning signals?
Experiment A is a prospective research design within BenchEWS Duck Lab. We intend to compare predefined recovery outcomes after controlled disturbances with established early-warning baselines and model-based methods.
Planned approach
- Define disturbances, observation windows and recoverability outcomes before confirmatory data collection.
- Use controlled simulation and paired comparisons against a nominal condition.
- Compare indicators with established baselines such as variance, lag-1 autocorrelation, system identification and control-theoretic estimates where feasible.
- Check measurement artifacts, uncertainty, false alarms and out-of-sample performance.
Mali and the research environment
The local Mac is intended to run simulation, Python analyses and time-critical control. A private iPhone control center is planned for monitoring and non-time-critical commands. Virtual Mali comes first; physical MicroDuck tests will follow only after separate technical and scientific approval.
Scientific status and limitations
The prospective study protocol has been published, but publication of a protocol is not evidence of experimental success. The frozen internal protocol B-RG5a.4 remains subject to PUBLICATION HOLD and DATA-GENERATION HOLD; EXECUTION_ENABLED remains false. No confirmatory Experiment A results or incremental predictive benefit have been established. Simulations are not physical robot measurements. No inference to social or ecological systems follows from this design.