Explainable AI (XAI) Interfaces for Edge–Cloud Forecasting Platforms: Enhancing Operations Manager Trust in Deep Learning Predictive Output

• Asher Noah

  covenant University

  Author

operational prediction, yet operations managers remain reluctant to act on outputs they cannot interpret. This study addresses the gap between predictive accuracy and managerial trust by designing, implementing, and validating an explainable AI (XAI) interface layer for a collaborative hybrid CNN–LSTM edge–cloud forecasting architecture. Using a design-based research methodology combining retrospective retail demand data analysis with a controlled user evaluation (N=48 operations managers), the study tested an XAI dashboard integrating SHAP attributions and counterfactual explanations against a standard prediction-only interface. Results indicate that the XAI-enhanced interface achieved 89.4% managerial trust endorsement compared to 61.2% for the control condition (p < 0.001), while preserving the underlying model's 94.7% forecasting accuracy. Feature attribution transparency significantly reduced decision latency and increased appropriate reliance on model outputs. The study concludes that trust-calibrated XAI interfaces constitute a necessary socio-technical layer for edge–cloud forecasting adoption, and provides a replicable framework for designing explanation dashboards that align with operations managers' decision workflows.

• pdf

Copyright (c) 2026 Asher Noah (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.