Hello. Have you tried to create a training set for machine-learning-based MMI detection. For example you can display a deviation graph as PEAR did, and make operator influence it to go up. The exact timing of operators intent can be detected using emotiv EEG, since its capable of detecting when someone thinks about certain direction in space. In this exact time, bits from QRNG can be saved and used for ML-training.
The other approach, we tried recently, is replacing unstable human element with plant-based intention source. You can see it in this topic Psionic Generator Experiment
Also if you want to know more detailed information about Randonautica mechanics, it can be found here Mind Matter Interaction | Randonauts Wiki | Fandom