User-Centric Explainability in Fintech Applications
S Deo, N Sontakke - HCI International 2021-Posters: 23rd HCI …, 2021 - Springer
HCI International 2021-Posters: 23rd HCI International Conference, HCII 2021 …, 2021•Springer
Fintech applications such as robo-financial advisors (RAs) are complex algorithmic decision
making systems, which gained prominence with their claim to democratize finance. Lack of
transparency and explanations for these automated decisions leads to a trust deficit for
users, limiting the potential of these applications. Our research aims to analyse the
effectiveness of user-centric explanations in conveying the decision-making logic of complex
algorithmic systems. Our user study tests techniques from explainable AI, varying in …
making systems, which gained prominence with their claim to democratize finance. Lack of
transparency and explanations for these automated decisions leads to a trust deficit for
users, limiting the potential of these applications. Our research aims to analyse the
effectiveness of user-centric explanations in conveying the decision-making logic of complex
algorithmic systems. Our user study tests techniques from explainable AI, varying in …
Abstract
Fintech applications such as robo-financial advisors (RAs) are complex algorithmic decision making systems, which gained prominence with their claim to democratize finance. Lack of transparency and explanations for these automated decisions leads to a trust deficit for users, limiting the potential of these applications. Our research aims to analyse the effectiveness of user-centric explanations in conveying the decision-making logic of complex algorithmic systems. Our user study tests techniques from explainable AI, varying in complexity and transparency. The quantitative aspects of our study determine the efficacy and usability of explanations and the qualitative aspects measure the effect of explanations on users and system usability. Our study finds trust and confidence of users in the system is positively correlated with comprehension and transparency provided by the presence of an explanation. There is a notable reduction in comprehension and trust between transparent white and opaque black box explanations of algorithms. This study is designed to aid policymakers and regulators in order to understand user needs which are crucial to designing better policies around algorithmic explainability for RAs.
Springer
以上显示的是最相近的搜索结果。 查看全部搜索结果