adding noise

Examples of places where adding noise makes a thing better. Better can mean more accurate, more fault-tolerant, less pathological, or anything else. In modeling this can mean noticing uncertainty or noticing variance. One good workflow for building out a model is to define parameters and vary or fuzz them to see if a) it fites the data better and b) to protect against overfitting.

adjust your seat is an example of assuming and responding to high variance. beating, cascading, and other ways distributed systems respond to synchronicity (and lack thereof) discusses how multiagent systems and distributed systems work with noise.

Related: Perturbation Theory.