Informative Semi-Factuals for XAI: The Elaborated Explanations that People Prefer

Abstract

Recently, in eXplainable AI (XAI), $\textit{even if}$ explanations -- so-called semi-factuals -- have emerged as a popular strategy that explains how a predicted outcome $\textit{can remain the same}$ even when certain input-features are altered. For example, in the commonly-used banking app scenario, a semi-factual explanation could inform customers about better options, other alternatives for their successful application, by saying "$\textit{Even if}$ you asked for double the loan amount, you would still be accepted". Most semi-factuals XAI algorithms focus on finding maximal value-changes to a single key-feature that do $\textit{not}$ alter the outcome (unlike counterfactual explanations that often find minimal value-changes to several features that alter the outcome). However, no current semi-factual method explains $\textit{why}$ these extreme value-changes do not alter outcomes; for example, a more informative semi-factual could tell the customer that it is their good credit score that allows them to borrow double their requested loan. In this work, we advance a new algorithm -- the $\textit{informative semi-factuals}$ (ISF) method -- that generates more elaborated explanations supplementing semi-factuals with information about additional $\textit{hidden features}$ that influence an automated decision. Experimental results on benchmark datasets show that this ISF method computes semi-factuals that are both informative and of high-quality on key metrics. Furthermore, a user study shows that people prefer these elaborated explanations over the simpler semi-factual explanations generated by current methods.

Figures (5)

• Figure 1: The decision space for three loans, with two features shown (i.e., Loan Amount in $s and Credit Score as CS), for three applicants: John, Mark and Mary. John and Mark have their loans accepted but Mary has been rejected. Mark asks for an explanation about how to get a better deal. The semi-factual (SF) tells him that with his 550 credit score, he can actually get a $65k loan. Mark previously thought that if he asked for more, he would end up being rejected like Mary. The semi-factual shows Mark the limits on his loan aplication given his credit score (n.b., if he asked for $70k, he would be rejected like Mary).
• Figure 2: The rightmost graph shows a decision space for Mark and his semi-factual explanation (SF), with a path between them based on two perturbation steps (Q' and Q") in which the key-feature loan amount is systematically increased from $20k to $65K, without changing credit-score (which stays at 550). The leftmost graphic shows the relative changes in the marginal contributions of these two features across these perturbed instances as they remain in the loan-accept class. As loan amount increases its marginal contribution to keeping instances in the loan-accept class decreases (see purple plot) and even though credit-score's value does not change, it's marginal contribution increases (see orange plot) revealing this seesaw pattern between the two features.
• Figure 3: From Expt.1, the percentage of semi-factuals, for five datasets, generated by the ISF and Ensemble-methods (N=38,233 in total), that manifested the seesaw pattern in key-feature versus hidden-feature contributions.
• Figure 4: From Expt.2, the mean goodness scores for semi-factuals generated by the ISF and Ensemble-methods for the five datasets (N=38,233 in total), in each of the four evaluation metrics (a-d).
• Figure 5: The percentage of people choosing the (a) Bare Semi-factual or Informative Semi-Factual in User Study 1, and (b) the Good or Bad Semi-factuals in User Study 2, for scenarios with the loan-accepted or loan-rejected outcomes

Theorems & Definitions (2)

• definition 1: Semi-factual Explanation
• definition 2: Informative Semi-factual Explanation