There are many possible alternatives to the transparency difficulty. SHAP tried to resolve the transparency difficulties by visualising the contribution of every element to the output.[191] LIME can locally approximate a model with an easier, interpretable design.[192] Multitask learning supplies a