Paper Number
1028
Paper Type
Complete Research Paper
Abstract
Cross-company federated learning (FL) represents a privacy-preserving approach towards collaborative artificial intelligence. FL approaches in a consortium may thus be used to establish platforms for value co-creation among the participating organizations. However, like other types of digital platforms, FL platforms can benefit from adopting control mechanisms to handle potentially harmful actors that may not positively contribute to the predictive power of the shared artificial intelligence model. Nevertheless, there is little research regarding the employment of control mechanisms for FL platforms, particularly considering the promising yet mostly overlooked input control. To fill this gap, we propose an explainable IT artifact. It serves as FL input control and provides an objective and traceable decision about participation requests of actors willing to join the platform. Finally, we demonstrate the utility of our input control artifact for the case of customer churn prediction in e-commerce
Recommended Citation
Düsing, Christoph; Röder, Marco; Thiesse, Frédéric; and Cimiano, Philipp, "Design of an Explainable Input Control for Cross-Company Federated Learning" (2024). ECIS 2024 Proceedings. 4.
https://aisel.aisnet.org/ecis2024/track10_dmds_ecosystems/track10_dmds_ecosystems/4
Design of an Explainable Input Control for Cross-Company Federated Learning
Cross-company federated learning (FL) represents a privacy-preserving approach towards collaborative artificial intelligence. FL approaches in a consortium may thus be used to establish platforms for value co-creation among the participating organizations. However, like other types of digital platforms, FL platforms can benefit from adopting control mechanisms to handle potentially harmful actors that may not positively contribute to the predictive power of the shared artificial intelligence model. Nevertheless, there is little research regarding the employment of control mechanisms for FL platforms, particularly considering the promising yet mostly overlooked input control. To fill this gap, we propose an explainable IT artifact. It serves as FL input control and provides an objective and traceable decision about participation requests of actors willing to join the platform. Finally, we demonstrate the utility of our input control artifact for the case of customer churn prediction in e-commerce
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.