Company News

ExplainX Rolls Out Enhancements to its Open-Source Explainable AI Platform Designed to Unlock Black-Box AI

Post by
Executive Office
explainX - a powerful explainable AI platform for data scientists

July 16th, 2020 — New York City —, today announced enhancements to its open-source explainable AI platform designed to take explainability of black-box AI models to a new level. These enhancements include Region-based Explanations using SQL, Multi-Level EDA and Similar Prototypes.

Gartner research indicates that by 2020, "85% of CIOs will be piloting artificial intelligence programs through a combination of buy, build, and outsource efforts." Ultimately, companies have embraced the value of artificial intelligence and machine learning. But to make a disruptive impact in organizations, AI has to be trusted. As we delve more into trust, ethics and biases, it is critical that organizations have the support they need to build trust in their AI systems. ExplainX’s new enhancements will empower data scientists to remove biases, build trustworthy and ethical AI. 

In the latest release of the platform, explainX has introduced: 

  • Region-Based Explanations using SQL: It is especially useful to understand the regions or subsets in the data that caused the model to make a particular decision. By using explainX’s in-built SQL module, data scientists and researchers can easily drill-down into their data and get region-based attributions. With explainX’s region-based explanations, users can set specific conditions to trigger real-time how the model performs in different subsets of the data. This builds on explainX’s explainability capabilities by allowing users to increase transparency and lower risk for predictions, not just on local but also on a regional level. 
  • Multi-Level EDA: Inherent biases can be a result of a poorly annotated or balanced dataset. In light of that, we have added a multi-level EDA that allows users to compare their model predictions and actuals on multiple subsets of data visualized in a single view. 
  • Similar Prototypes: In this release, we have added an optimized & model-agnostic-version of ProtoDash, an algorithm that efficiently represents local prediction data-point by selecting prototypes based on its importance weights or feature attributions. With explainX’s similar prototypes module, users can summarize and identify similar data points to communicate meaningful insights to business users in domains where simple explanations are hard to extract.

“We believe that explainability and fairness in AI are becoming extremely crucial as more companies adopt AI - with this release, we are enabling individual data scientists and researchers to dive into how their AI models perform. The purpose is to ultimately help them build a narrative that decision makers can easily understand and trust - leading to a jump in accurate decision making” said Raheel Ahmad, co-founder and author of “With our open-source model and enhancements, we are making it extremely seamless for literally anyone to explain any black-box AI model they have” said Muddassar Sharif, co-founder and author of

Users can get started with explainX by either installing through PIP (pip install explainx) or cloning the repo from explainX’s Github repository.

About ExplainX

ExplainX is a front-runner in open-source explainable AI, delivering packaged and easy to use explainable AI algorithms to thousands of data scientists around the globe. ExplainX’s open-source python library democratizes explainable AI with end-to-end explainability techniques for explaining, debugging and monitoring machine learning models. This platform maximizes business value by enabling trustworthy, responsible and ethical AI. 

With the mission of bringing trust and transparency in AI systems, explainX is calling for fair and responsible AI systems. For more information, visit, and join the conversation on Slack and LinkedIn

Share this with your friends.

Ready to dive in?
Create your free account now.