IBM Watson OpenScale wins 451 Firestarter Award for making AI outcomes fair and explainable

Offering Manager, Watson Data & AI, IBM

Businesses from every industry are significantly investing in AI, ramping up platforms and skills. But although they can see the business value that AI will deliver, some companies are hesitating to move their AI experiments from pilot to production.

A key problem is that AI can seem like a black box,. It is hard to explain an AI model’s predictions, and it is difficult to ensure that outcomes are not being affected by unwanted biases in the model. You need to be confident that model predictions result in compliant, fair and explainable outcomes. And you need to ensure you can test and manage model risk before, and during, a model’s production lifecycle.

 Released late last year, IBM Watson OpenScale can provide all of these capabilities. It brings bias detection, explainability, and governance to AI and machine learning models. It presents these tools, tests and metrics in a way that is easy to understand for business managers and model validation teams who are the prime stakeholders of the AI applications. It also helps data scientists to focus more time on building and improving models instead of monitoring them in production environments.

These are key reasons why analyst firm 451 Research just awarded Watson OpenScale one of its Q2 451 Firestarter awards, which recognize organizations for exceptional innovation and disruption in their markets.

Awardees are nominated by the 451 Research analyst team based on their insights and expert opinion of long-term trends and competitive landscape, and their in-depth conversations with players in the industry. Nominations are based on a combination of factors, including uniqueness, strategic and technology vision and the disruptive potential of the organization's technology.

IBM Watson OpenScale was chosen for the award, in the words of the 451 Research announcement, for “giving business users the power to contextually examine these models in a way that is not bottlenecked with a few expert users, typically on the data science team. IBM Watson OpenScale allows business users to visually drill down into the effects of variables and control their impact on models. Because perfect model accuracy and perfect model fairness are often at odds, these capabilities provide the 'knobs' to make adjustments so that automation can be implemented ethically and defensibly at scale. 

OpenScale assumes that AI models are indeed a black box, but it can examine their behavior in production, isolating variables. The 451 Research announcement notes that this allows “the business to control for bias and ensure explainability and fairness for both regulatory compliance and operationalization. Monitoring capabilities help organizations increase the performance and scalability of AI and ML initiatives, providing proactive use cases as well as reactive compliance ones.”

Interested in exploring how you can strengthen AI and ML models at your business? Get a quick overview of Watson OpenScale. View the Watson OpenScale capabilities that helped win the Firestarter award. And get started on Watson OpenScale at no cost.