Four things that characterize responsible AI

More and more companies are relying on artificial intelligence for their offerings. Their algorithms suggest products and services, breathe life into chatbots and translate texts. As a result, companies are under considerable competitive pressure, because the competition has long since launched its AI systems. Nevertheless, they should beware of rash decisions and ensure that they use responsible AI. Otherwise, they could be at risk of significant damage. Pegasystems lists the four central characteristics of such an AI.

Fairness

Companies should ensure that their AI treats all genders, ethnicities, age groups, social classes, income groups, religious communities, and sexual orientations equally. To do this, they must build AI models that are free of bias, proactively monitor them, and continuously analyze their results.

The aspect of fairness is often not the top priority when using AI. However, if algorithms make discriminatory decisions, this can have extremely damaging consequences for a company's reputation and business.

Transparency

When in doubt, companies should be able to show how their AI arrived at a decision. This is especially true for highly regulated industries such as financial services or insurance. Recently, for example, a financial services provider came under fire for apparently offering women with the same income background as men worse terms for credit cards. Since it could not explain how its AI arrived at the relevant decisions, it was seen as hostile to women. For regulated and high-risk use cases, companies should therefore use transparent and explainable algorithms.

Empathy

The need for companies to be empathetic to their customers has never been greater than in the stressful times of the Corona pandemic. That's why they should deploy empathic AI. Such AI makes decisions that are relevant and helpful and puts the needs of customers first.

To do this, it must be able to take a customer's full context into account. Only then can it know what he needs right now. This doesn't necessarily have to be a specific offer, but can also mean leaving him alone for a certain amount of time.

Robustness

A few years ago, the Twitter bot "Tay" made headlines. Twitter users were invited to interact with it - within 24 hours, it turned into a nightmare. The underlying AI model became misogynistic and racist because of the data it picked up from Twitter conversations. This case highlights the need for enterprises to have robust AI; AI that cannot be easily influenced by built-in protections, rules and guidelines.

"Fairness, transparency, empathy and robustness are the cornerstones of responsible artificial intelligence. With them, companies can ensure that their AI reflects the values of our society and focuses on customer needs," explains Kay Knoche, Principal Solution Consultant at Pegasystems. "This is not only ethically imperative, but also has a positive impact on companies' revenue in the long term."