Active Learning Method for Experience Sampling

I was wondering, if there is current work being done about the integration of active learning methods to help the experience sampling. Currently the experience sampling in the various benchmarks is handled automatically. I was wondering if there would be a way to enable approaches such as "naive (uniform sampling)" and "margin based" to actively/intelligently select which samples should be included in the next experience

Various Sampling Methods: https://github.com/google/active-learning/tree/master/sampling_methods