Request an online demo
Get a personalized online demo of Kount's trust and safety technology at a time and date of your choosing.
Decoding the Black Box: New Explainable AI by Equifax
Artificial intelligence has revolutionized fraud detection, offering unparalleled speed and accuracy. However, as AI models grow in complexity, a significant challenge has emerged — the "black box" problem. We feed data into these advanced algorithms, and they deliver results, but the reasoning behind those results often remains shrouded in mystery. This lack of transparency can erode trust and hinder effective decision-making, particularly in the critical realm of payment fraud.
For Equifax customers who use Payments Fraud on Kount 360, Omniscore provides valuable insights. But sometimes users struggle to understand why a particular score was assigned and what specific factors influenced the decision. This ambiguity creates friction, leading to time-consuming manual reviews and management of complex policies.
The Need for Clarity: Why Explainable AI Matters
Model explainability isn't just a nice-to-have — it's a strategic imperative. For Equifax, it's about building stronger customer confidence and trust. By leading with AI that is explainable, we empower users to make more informed decisions, ultimately driving greater value through enhanced understanding and trust.
Introducing Explainable AI
To address this critical need, Equifax has introduced a powerful new feature for Payments Fraud designed to shed light on the inner workings of the Omniscore. This feature provides unprecedented transparency, allowing users to understand precisely what factors contributed to a transaction's risk assessment.

This new functionality is broken into four key areas, all focused on making the Omniscore more understandable:
- Expansion Action in the Omniscore Widget: This allows users to quickly access a detailed breakdown of the factors influencing the Omniscore directly from the transaction details view.
- Omniscore Explainability Display: A dedicated modal provides a comprehensive view of the Omniscore's calculation, offering a clear and concise explanation of the contributing factors.
- Waterfall Chart: This visual representation breaks down the Omniscore into its constituent parts, showing the positive and negative influences of each factor. This allows users to easily identify the most significant drivers of risk and safety.
- Risk and Safety Factors: This feature clearly delineates risk factors (those that decrease the Omniscore) and safety factors (those that increase it), providing a nuanced, yet comprehensible, understanding of the transaction's profile.

Benefits of Enhanced Explainability
The benefits of this new feature are significant:
Provides Insights to Manual Reviews
By providing clear explanations of key risk factors, users now have additional data points to use during investigation.
Improves Accuracy and Reduces Time Spent on Reviews
Users can quickly identify the key risk factors, reducing the need for extensive investigations, saving valuable time and resources.
Reduces Complexity in Policies
The increased transparency reduces the need for overly complex rules. With an increased trust and adoption of the Omniscore, users simplify the management of the system.
How Do We Do It?
With a combination of ML, game theory, and a patent pending method for sub-millisecond inference over our vast amount of related transactions and variables, we can find the real impact of features that translates signals from our fraud network into an on-demand explanation for every transaction.
Empowering Users with Transparency
With this new feature, Equifax is moving beyond simply providing results. We're empowering users with the knowledge and understanding they need to make confident decisions. By illuminating the "black box" of AI, we're fostering trust, driving efficiency, and ultimately, creating a safer and more secure payment environment.
This new feature is available on the Kount 360 platform. Check out our platform page to learn about all our product offerings.