Explaining black box AI machine learning models and using interpretable models
Question for written answer E-005390/2020
to the Commission
Patrick Breyer (Verts/ALE)
There has been criticism of the fact that the private sector has an incentive to produce non-transparent black box algorithms in order to be able to commercialise them. Simple interpretable models with the same or a higher level of performance would be free for everybody to use.
- 1.What is the Commission’s opinion of the proposal that for certain high-stakes decisions, black boxes should not be allowed if there is already an interpretable model with the same level of performance?
- 2.What is the Commission’s opinion of the proposal that organisations that introduce black box models should also be obliged to report the accuracy of interpretable modelling methods?