Fin-X - Towards a meaningful explanation of AI systems

Explainability of the outcomes and operation of artificial intelligence (AI) applications is an important condition for safeguarding consumer and society confidence in AI applications, especially in the financial sector. In this project we develop tools to convert the outcomes of complex AI applications into an understandable explanation for employees such underwriters and claims officers. This is important because these employees communicate with the customer and must be able to provide an explanation, for example if a customer's claim is rejected because it has been labelled as fraudulent by an AI application.

Objective

The project aims to make a concrete contribution to the implementation of effective and human-oriented AI applications by developing tools that give internal users of those AI applications more and better insight into the operation and outcomes.

Results

Guidelines (in the form of tools and instruments) that provide direction and support in generating, communicating and evaluating explanations. The primary target group of the guidelines are designers and developers of AI and XAI applications.

Duration

31 January 2023 - 31 March 2025

Approach

The project follows a design science approach in which artifacts are developed based on practical needs and already available knowledge. The artifacts in this project are the guidelines described under results.

HU researchers involved in the research

Collaboration with knowledge partners

Amsterdam University of Applied Sciences, Floryn, FRISS, De Volksbank, EMS, Sygno, MavenBlue, Mogelijk, Deeploy, Datacation, BBO, Amsterdam Data Collective, BG.legal, De Nederlandsche Bank, Jheronimus Academy of Data Science, Verbond van Verzekeraars, Holland FinTech, ROM Utrecht, Vereniging Financieel Adviseurs, Human & Tech Institute

Subsidy

FIN-X is a project that receives a subsidy from the National Organisation for Practice-Oriented Research SIA in the context of the RAAK-SME scheme September 2022 with file number RAAK.MKB17.003.

 

Any questions or want to collaborate?

Henry Maathuis

  • Researcher
  • Research group: Artificial Intelligence