Relevance, reliability, and accountability – these are SAP's guidelines for artificial intelligence built into solutions and products. The “responsible” part is overseen by Dr. Sebastian Wieczorek and his AI ethics department at the software company.
“At SAP, ethics has been part of our artificial intelligence research and development from the beginning,” said Wieczorek, SAP’s head of AI ethics. “All developments in the AI space are deeply aligned with SAP values.”
The beginnings of artificial intelligence at SAP
Mr. Wieczorek was part of SAP's original AI division, founded in 2014. “In addition to technical and product tasks, we have always considered the ethical aspects of our work from the beginning,” he says.
SAP is the first European company to define guidelines for dealing with AI and set up a corresponding advisory board. Although Wieczorek's work has always been technical, he is also a member of the SAP AI Global Ethics Steering Committee, a member of his AI research committee in the German Bundestag, and has reported on the use of artificial intelligence in the EU. Ta. Parliament.
SAP began an internal process early on to develop an ethically objectionable approach to AI, which ultimately led to the SAP AI Global Ethics Policy.
To address ethical issues, experts must have deep knowledge of AI technology and be willing to address not only the legal aspects of the technology, but also philosophical and moral issues. must be able to do so.
“Our work in AI ethics is similar to that of a translator,” Wieczorek says. “The realities and possibilities of technology must be “translated” into the language of philosophy, sociology, and law. The results must then be “translated” into technical requirements to enable continuous exchange between the two areas. ”
role of humans
Currently, AI systems cannot develop their own motivations or concepts of themselves or the world. Much less can we take our own initiative to consider how to best optimize this world.
“Their purpose and mission are determined by humans,” Wiesorek emphasizes. “What humans no longer do is define the exact implementation of the task.”
As always, whether a task is delegated to a machine or to another person, ensure that certain rules regarding fairness, transparency, and the right to human participation are observed in its execution. is needed.
“When it comes to AI, we don't know much about how decisions are ultimately made compared to traditional software,” Wieczorek says. “Therefore, we must always leave open the possibility of intervention if this automation does not work as expected in a particular case.”
What such an intervention looks like in individual cases can vary greatly, as the entire software needs to be considered and not just the individual components.
The most well-known example of discrimination through intelligent software is the exclusion of groups that have historically been underrepresented in the job application process. The historical data used to train the AI may reflect biased selection criteria from the past. Therefore, it is theoretically possible for AI to adopt and reproduce these biases in its own selection process.
“Due to the high potential for automation in AI software, these types of side effects can occur at scale and relatively quickly,” Wieczorek says. “Therefore, we must set high standards for the types and methods of automation to limit side effects and have the ability to effectively reverse side effects.”
Training data set guidelines do not uniquely exploit or guarantee maximum fairness.
“There's a whole host of things to consider,” Wieczorek said. “The system as a whole must be able to provide assurance that the evaluation is fair. The system's behavior must be perfect overall.”
Dedicated ethics review for every AI use case
“The SAP AI Global Ethics policy states that all of our products and solutions that use AI must be monitored from an ethical perspective, both during the development phase and subsequently after they are launched on the market.” says Wieczorek.
Therefore, each AI use case is subject to an individual review, including a declaration from the product team on how the use case complies with the guidelines of the SAP AI Global Ethics Policy.
After definition, all use cases undergo a classification process. However, if your use case involves making automated decisions that affect people or processing personal data, for example, they are automatically considered sensitive and classified as high risk.
“Such use cases undergo a mandatory review process, which is continually accompanied by experts, such as my team, for example,” says Wieczorek. “In this way, the risks in each individual case are systematically checked and, if necessary, it is determined what measures need to be taken to implement the ethical standards stipulated by SAP.”
Does the use of AI limit human responsibility?
Routine tasks that are taken over by AI are often already automated to some degree and do not need to be formulated anew each time. But AI is also taking over increasingly personalized tasks, such as through chat interactions, increasing the responsibility of individual users.
Wieczorek believes there is a shared responsibility between AI developers and users. “We do not intend to shift responsibility to the product, especially for work that affects people and affects the human body,” he emphasizes.
Companies that provide AI applications have an obligation to provide transparency about how they work and make it clear what they were designed for and what they were meant to be. do not have was designed for.
This allows users to take responsibility for themselves. Assigned tasks must comply with ethical principles, and results must be verified rather than simply accepted.
It is especially important that people be able to intervene in the functioning of the system in case undesirable side effects, such as unfair system behavior, become apparent over time.
“You always need to ensure that decisions made by the AI can be reviewed, questioned, and even reversed by humans,” Wieczorek says. “This is the responsibility of the manufacturer of the AI system, as set out in the SAP AI Global Ethics Policy.”
Learn more about SAP's approach to ethical AI below.