How IBM Watson is helping clients build trust in data, models and processes

By | 4 minute read | April 21, 2021


Through my role as Chief AI Officer at IBM, I have the privilege of working with organizations to ensure they have a holistic approach to AI. We do that by helping businesses design AI strategies, set guiding principles, and infuse responsible AI products with intent and purpose throughout their business.

Today, IBM announced new capabilities for IBM Watson products that are designed to help businesses build trust in their data and AI models. Below, I will put those product updates into the context of our overarching approach to building trustworthy AI throughout the AI lifecycle.

IBM’s governed data and AI technology is structured on the application of our fundamental principles for ethical AI — transparency, explainability, fairness, robustness and privacy — to three key elements of the AI lifecycle: data, models and process. By offering tools to instill trust in these critical facets of the AI lifecycle, we’re helping businesses build trust in their AI and the outcomes it drives.

Trust in data

When it comes to data, trust means delivering a comprehensive view of quality data that is governed and ready for analysis. I’ve seen firsthand how important it is for businesses to have timely, trusted, and quality data to help ensure the data and its lineage are well-structured, understood and maintained over time. No amount of algorithmic sophistication can overcome poor data.

Our goal of infusing trust into data is twofold: first and foremost, it’s about the governance of data, such as lineage tracking and policy enforcement, which are both required to manage data to provide a solid data foundation. Second, it’s about helping companies make that data available to the people who need to use it and the AI models that require it through self-service mechanisms.

That’s why we offer IBM Watson Knowledge Catalog, a hybrid, multicloud enterprise metadata repository designed to allow organizations to access, curate, categorize and share data assets and their relationships, wherever they reside, providing end-to-end data governance capabilities. Watson Knowledge Catalog is our nervous system for everything we do in data and AI — we designed it to help users understand data, govern and protect data and AI estates, and improve data quality. This solution is designed to not only help businesses build the enterprise catalog of all data from different sources across multiple public and private clouds. It also allows for varying views and access to data based on personas and roles.

Trust in models

Trust in models means increasing confidence in model accuracy. How do we do that? At IBM we have integrated the tools and capabilities we believe are needed to more efficiently run and manage AI models, simplify AI lifecycle management, and empower data scientists to help optimize their data-driven decision-making into one industry-leading tool: IBM Watson Studio.

IBM Watson Studio is designed to help organizations to build and scale AI across their organization with trust and transparency by automating AI lifecycle management. Deploying AI with continuous model governance helps enable users to accelerate time to discovery, prediction, and outcomes while keeping AI explainable and tuned to an organization’s business demands. We’ve also infused intelligent automation into the product, which is designed to augment human skills to build and manage models, identifying and mitigating problems like bias or drift.

IBM Watson Studio now includes a tech preview of federated learning capabilities to help businesses apply machine learning techniques to situations where data cannot or should not be moved due to reasons such as data privacy, secrecy, regulatory compliance, or simply the size of data involved. With IBM Watson Studio, businesses will be able to train AI models on previously siloed data sources.

Building trust in models also means providing enhanced explainability for models and their resulting predictions. That’s why we’re planning to bring a new statistical details page to IBM Planning Analytics with Watson in Q2 2021 to provide more transparent and easy-to-understand facts about how a forecasting prediction was generated.

Trust in process

Our approach to trust in process is built upon helping to maintain compliance, repeatability and overall explainability for AI at scale. That’s where a highly scalable governance, risk and compliance (GRC) solution like IBM OpenPages with Watson comes in.

IBM OpenPages with Watson centralizes siloed risk management functions within a single environment designed to help businesses meet their obligations to identify, manage, monitor and report risk and regulatory compliance, especially in today’s changing business landscape. The new Data Privacy Management module in IBM OpenPages with Watson is designed to help businesses meet evolving data privacy challenges. By integrating OpenPages with Watson Knowledge Catalog, we’re also able to help businesses break down silos and get a more holistic view of how private data is being used, from applications to AI models.

Building trust is at the core of our AI for business strategy at IBM and is an important factor for businesses on their journey to AI.

By working closely with IBM Research to continually bring better and more innovative tools and approaches to IBM Watson, we’ve designed a set of products, underpinned by a commitment to ethical principles and an open ecosystem, to help businesses build trust at every point of the AI lifecycle so they can confidently operationalize AI across hybrid multicloud environments.

Forward-Looking and Cautionary Statements: IBM’s statements regarding its plans, directions, and intent are subject to change or withdrawal without notice at IBM’s sole discretion. Information regarding potential future products is intended to outline our general product direction and it should not be relied on in making a purchasing decision. The information mentioned regarding potential future products is not a commitment, promise, or legal obligation to deliver any material, code or functionality. Information about potential future products may not be incorporated into any contract. The development, release, and timing of any future features or functionality described for our products remains at our sole discretion.