We have all for some time now read about the opportunities and risks of bias when using more advanced AI models automating decisions. It's time for us business people, developers and general society to move forward and become aware of how this can be done in practice. Join us at Goto 10.

Register to this event by sending an email to Gabriel Delerin, gabriel.delerin@ibm.com.

So we invite you to an afternoon of insight and we picked one of the most mature areas as our scenario - credit risk modeling in retail and corporate banking.

The scenario

Traditional lenders are under pressure to expand their digital portfolio of financial services to a larger and more diverse audience, which requires a new approach to credit risk modeling. Current standard modeling techniques - like decision trees and logistic regression - work well for moderate datasets, making recommendations easily explainable. This satisfies regulatory requirements of lending decisions being transparent and explainable.
New services providing credit access to a wider and riskier population, applicant credit histories must expand beyond traditional credit input, like mortgages and car loans, to alternate sources of input, like utility and cell phone plan payment histories, education and job titles. These new data sources offer promise, but introduce risk by increasing likelihood of unexpected correlations - thus introducing bias based on an applicant’s age, gender, or other personal traits.
Popular data science techniques for these diverse datasets, e.g. gradient boosted trees and neural networks, can generate highly accurate risk models, but at a cost - transparency. Such "black box" models generate opaque predictions that must somehow become transparent, to ensure regulatory approval (e.g. Article 22 of the GDPR) as well as maintaining trust and credibility in the eyes of clients and stakeholders.

Talks & demos

* Theory introduction to explaining ML model outcomes and identification & management of bias, Wiktor Mazin, Chief Data Scientist Europe@IBM
* Perspective from Fintech/solution provider, speaker TBC
* IBM, AI fairness framework, introduction and demo how to manage bias in credit risk modeling, Wiktor Mazin, Chief Data Scientist Europe@IBM
- how to detect-, mitigate and monitor bias in training data & models from any source and explain decision recommendations.

Register to this event by sending an email to Gabriel Delerin, gabriel.delerin@ibm.com.