Abstract
Recent advantages in artificial intelligence (AI) research allow building sophisticated models to advise users in various scenarios (e.g., in financial planning, medical diagnosis). For companies, this development is relevant since it allows scaling of services that were not scalable before. Nonetheless, in the end, the user decides whether he/she uses a service or not. Therefore, we conducted a survey with 226 participants to measure the relative advantage of AI-based advisory over human experts in the context of financial planning. The results show that the most important advantage users perceive is convenience, since they get easy and instant satisfaction of their informational needs. Furthermore, the effectivity of eleven measures to increase trust in AI-based advisory systems was evaluated. Findings show that the ability to test the service noncommittal is superior while the implementation of human traits is negligible.
Recommended Citation
Mesbah, Neda; Tauchert, Christoph; Olt, Christian Michael; and Buxmann, Peter, "Promoting Trust in AI-based Expert Systems" (2019). AMCIS 2019 Proceedings. 6.
https://aisel.aisnet.org/amcis2019/ai_semantic_for_intelligent_info_systems/ai_semantic_for_intelligent_info_systems/6
Promoting Trust in AI-based Expert Systems
Recent advantages in artificial intelligence (AI) research allow building sophisticated models to advise users in various scenarios (e.g., in financial planning, medical diagnosis). For companies, this development is relevant since it allows scaling of services that were not scalable before. Nonetheless, in the end, the user decides whether he/she uses a service or not. Therefore, we conducted a survey with 226 participants to measure the relative advantage of AI-based advisory over human experts in the context of financial planning. The results show that the most important advantage users perceive is convenience, since they get easy and instant satisfaction of their informational needs. Furthermore, the effectivity of eleven measures to increase trust in AI-based advisory systems was evaluated. Findings show that the ability to test the service noncommittal is superior while the implementation of human traits is negligible.