Probabilistic human-machine cooperation in product personalization

When personalizing products, AI algorithms relieve customers of the burden of choice. However configuration recommendations by AI are probabilistic in nature. Users need to understand this to make well informed decisions.
We therefore propose a user interaction paradigm for recommender and configuration systems which is based on Single Pass Bayesian Reasoning and on Suitability Probability Tables. Personalizing shoes is used as a use case for demonstration.
This interaction paradigm can be maintained even with modified algorithms. Generalizability to other classes of algorithms remains to be proven as well as correctness of interpretation by users and user acceptance.