After doing this, SHAP computes an importance weight of 0.5 for both the loan amount and the annual income, and there are two counterfactual explanations: the applicant can either reduce the loan amount or increase the annual income to get the loan approved (columns 1 and 2 in Table 2). Now, consider a different scenario. Suppose the bank were stricter with the loans it approves and used a decision threshold 2.5 percentage points lower. Now, in order to get approved the applicant of Loan 4 would need both to reduce the loan amount and to increase his/her annual income. In other words, with this different decision system, there is only one counterfactual explanation consisting of both features. However, SHAP would still show an importance weight of 0.5 for each feature, heedless of the fact that the decision-making process changed.