[KIELive#51] TrustyAI: Ensuring the Fairness and Transparency of Decision Models

See pitfalls found within machine learning development cycle and how TrustyAI can help identifying and correcting problems as they arise.

About this event

[KIELive#51] TrustyAI: Ensuring the Fairness and Transparency of Decision Models

Machine learning has come to rapid prominence in recent years as a means of leveraging valuable and actionable insights from massive datasets, and its unification with traditional decision management is crucial for building highly intelligent applications. Red Hat seeks to achieve this integration by embracing open source and open standards: the Red Hat Decision Manager platform offers the DMN (Decision Model and Notation) and PMML (Predictive Modeling Markup Language) standards to unify decision and predictive models.

However, the integration of machine learning into a decision pipeline is not without its perils. The prominence of machine learning has also highlighted the dangers and the pitfalls of its use: models are vulnerable to biased data, problematic learning, fragile or unfair decisions, and deceptive results. Without prudent development and deployment, machine learning models can make decisions that stand at odds with corporate or personal values. Even if a model is perfect at deployment time, changing user behaviors and data distributions can introduce new and unforeseen issues. Furthermore, modern regulations like GDPR demand accountability and transparency in automated decision making. Explainable AI or XAI is a research field that seeks to provide solutions to these issues and requirements, and the techniques produced can be generalized to the greater practice of decision automation. Red Hat’s TrustyAI initiative seeks to connect the cutting edge of XAI research with enterprise applications and deployment, such as to facilitate the development of powerful decision models that are simultaneously fair and transparent.

In this talk, we’ll look at some practical examples of the pitfalls found within machine learning development and deployment, and explore how TrustyAI’s suite of model monitoring, auditing, and explanation tools can help us identify and correct problems as they arise.

Keywords: eXplainable Artificial Intelligence, Machine Learning, TrustyAI, Red Hat Decision Manager

About the invited speaker:

Rob is a Senior Software Engineer at Red Hat, where he works on researching and implementing explainable AI algorithms as part of the TrustyAI initiative. Rob studied Physics at the University of Chicago and is a PhD candidate at Newcastle University, researching the automated design of deep convolutional neural networks.

About the KIE Lives:

The KIE Live Series is composed of live streamings that bring technical information and updates about business automation delivered by the projects under the KIE umbrella: Drools, jBPM, OptaPlanner, and Kogito.

Problems like process automation, decision automation, resource planning solution are the main topics, and of course, we always have in mind recent technology concepts like cloud-native application target for any type of cloud (private/public/hybrid/edge). You can expect to hear from business automation experts who code or/and deliver business automation within big enterprises across the world.

Join us to share, learn, and grow together.

Knowledge is everything