A lender tested two models for approving small-business loans. A gradient-boosted ensemble is 1% more accurate than a logistic regression model, but regulators require a clear reason for every decision, and the business agreed that the simpler model's accuracy is acceptable. Which model should the lender deploy?