Towards Explainable Artificial Intelligence and Machine Learning in Credit Risk Management
Auf einen Blick
- Projektleiter/in : Dr. Branka Hadji Misheva
- Co-Projektleiter/in : Prof. Jörg Osterrieder, Dr. Jan-Alexander Posth
- Projektteam : Florian Bozhdaraj, Dr. Helmut Grabner, Piotr Kamil Kotlarz, Dr. Sandra Stupar, Prof. Marc Wildi
- Projektvolumen : CHF 280'000
- Projektstatus : laufend
- Drittmittelgeber : Innosuisse (Innovationsprojekt / Projekt Nr. 41084.1 IP-SBM)
- Kontaktperson : Branka Hadji Misheva
Beschreibung
For Switzerland, the reputation of stable long-standing
financial institutions has always represented a crucial advantage
for the acquisition and retention of customers. However, as the
financial service industry becomes more global, clients expect
personalized offerings and inclusiveness of service which in turn
requires lenders to rely extensively on alternative data and
advanced modelling techniques.
Swiss financial intermediaries are falling behind in using
alternative data and novel Artificial Intelligence (AI)-driven
credit scoring methodologies. Partially, this is due to the large
legacy IT landscape which might not be accommodating for advanced
analytics. An even more relevant barrier for wider adoption of AI
in the context of the Swiss finance sector is related to the
concept of explainability. Namely, AI solutions are often referred
to as “black boxes” because typically it is difficult to trace the
steps the algorithms took to arrive at its solution and understand
how they have made their decisions. This challenge is particularly
relevant for Swiss financial intermediaries: both traditional and
FinTech companies face similar challenges in adhering to
data-related regulations such as the General Data Protection
Regulation (GDPR) which provides a “right to explanation”, enabling
users to ask for an explanation as to the decision-making processes
affecting them. This impacts their ability to develop AI-based
solutions.
We propose building a visual analytics (VA) tool, specifically
tailored to the context of credit risk evaluation, useable for both
model developers (i.e. Swiss financial intermediaries operating in
the commercial or consumer credit space) as well as model
evaluators (Swiss regulatory bodies that have to validate the
models). More importantly, such a visual tool will enable model
evaluators, a non-technical audience, to gain some insight into how
AI models applied to credit scoring work and identify the reasons
behind the decisions taken.