a deep learning framework to model the moderating effect of ethical leadership on emerging technology's impact on auditors' professional judgment

This study introduced a deep moderated neural network (dmnn) to model the non-linear interaction between emerging technology adoption and ethical leadership in shaping auditors' professional judgment. drawing on survey data from 151 auditors, were normalized five technology usage items and five ethical leadership items to [0,1] and constructed the target judgment score as the average of five professional-judgment items. we benchmarked the dmnn against four alternative methods ordinary least squares regression (ols), ols with explicit pairwise interactions (ols+int), support vector regression (svr), and random forest (rf) using rmse and r² on a held-out test set. the dmnn obtained an rmse of 0.1014 and an r² of 0.7562 better than those by ols (rmse = 0.1035, r² = 0.703), ols+int (rmse = 0.1333, r² = 0.508), svr (rmse = 0.1818, r² = 0.084) and rf (rmse = 0.1006, r² =0.719). these results showed that the dmnn learns complex moderation effects much better and also achieved higher predictive accuracy and explained variance than linear- and traditional machine learning approaches. the dmnn demonstrated enhanced performance as well as theoretical interpretability to shed light on auditors' judgment processes and represents a promising new analytical tool for auditing research moving forward.