Ensuring AI that is Relevant, Reliable, and Responsible

Many people think that artificial intelligence (AI) is primarily a technical affair. Nothing could be further from the truth. AI projects have not only technical aspects but also ethical, legal, and business dimensions that require thorough consideration beforehand. In this blog, AI expert Niels van der Kam explains how to ensure AI that is not only relevant but also reliable and responsible.

The AI revolution is well underway. Every week brings the launch of new, groundbreaking AI applications. Sometimes these tools are launched too hastily and behave in ways their creators had not anticipated. Consider the chatbot that tried to break up the marriage of a tech journalist, the airline bot recommending competitor flights, or image generators that preferentially portray white men as doctors or surgeons.

Artificial intelligence has been a subject of university research since the 1950s. Today, it’s not just scientists and engineers who are interested; AI has become a societal topic. This has led to ethical and legal debates that could significantly impact how AI will be used in the future. So be very aware that an AI project is much more than just a technical exercise. When thinking about your AI project, consider these other aspects and view your project through various lenses.

Start with why

Want to launch an AI project? Don’t start with what is technically possible (almost anything is possible), but rather with answering other essential questions: What are we doing? And most importantly: Why are we doing this? What problem are we solving? And for whom? Are there internal policy agreements about using AI? Include these in your considerations. Then delve into the potential risks from ethical and legal perspectives.

At SAP, we have thought through all these aspects thoroughly. We follow the rule that AI must always be explainable. Under the hood of our Business AI solutions, there never is an inexplicable black box. We always want to know how AI reaches a conclusion. Additionally, our Business AI tools must never make autonomous decisions. We believe that humans should always be ultimately responsible for making choices.

Is the use case clear? Are the preconditions clear? Then you can start looking at the technology perspective. What algorithm do you use to make decisions? Which datasets do you use to train this algorithm? And how do you ensure that these datasets do not contain biased information? Document all the choices you make.

Governance over coolness

The umbrella of all measures you take to ensure that AI stays within the lines is called AI Governance. My motto is “Governance over coolness.” Always prioritize setting up governance over rushing to launch your new, trendy app.

A good tip for setting up your AI Governance is to establish a committee with people who can objectively look at the project through different lenses (HR, Legal, Ethical). When you properly set up AI Governance, you achieve AI that is Relevant, Reliable, and Responsible. You develop an AI tool that is relevant for users, offers trustworthy output, and operates within the correct ethical and legal frameworks.

Maintaining control and overview

As you can see, there are quite a few aspects to keep track of. At the same time, the demand for AI is higher than ever. Before you know it, AI projects can proliferate throughout your organization. Maintaining control can then become a challenge. That’s why SAP has established a strategic partnership with Collibra. Are you building AI solutions with SAP or with data from your SAP environment? With Collibra, you can easily set up an organization-wide registry of all those AI projects. Collibra also helps you fully configure the governance around your AI projects. From Collibra, you can easily link to underlying datasets, such as those within Datasphere.

If you set up AI Governance from day one with Collibra, you will always have a clear view of all your AI projects. If the law changes or a dataset turns out to be faulty, you can immediately determine which AI projects in your organization are affected.

AI Governance

Interested in learning more about AI Governance? Explore our website or pre-register for the SAP BTP Innovation Day on October 8th, where AI Governance will be among the key topics discussed.