17 April 2023

"Discrimination by AI can basically affect anybody."

Artificial intelligence (AI) has the potential to fundamentally change and improve our professional world. But while some see the technology as a trailblazer for a more equitable society, others warn that it will exacerbate existing inequalities. In this debate, the research project "Artificial Intelligence in the Service of Diversity" (KIDD) is a pioneer for inclusive technology adoption.

Together with eight companies and partner organizations, KIDD is working to ensure that AI is used in an equitable and inclusive way. With support from the German Federal Ministry of Labor and Social Affairs (BMAS) and as part of the New Quality of Work Initiative (INQA), the KIDD consortium developed an innovative process to help companies integrate AI into their workflows in a sensitive way. The goal: to make the transition to a digitalized workforce fair and to ensure that no one is left behind. We spoke with Katja Anclam of KIDD to learn more about the innovative process and the challenges faced in developing AI in the service of diversity. In our interview, Ms. Anclam highlights how AI can help make our workplaces more inclusive and equitable, and what steps are needed to drive inclusive digitalization.

Hello Ms. Anclam, thank you for taking the time to answer our questions. Please introduce yourself:

Thank you very much for inviting me. My name is Katja Anclam. I am a communications expert, TV producer and, together with the computer scientist Christoph Henseler, managing director of difgl - the "German Institute for Good Living". There, I am responsible for the area of science communication. In 2019, I founded the non-profit association female.vision e.V. in Berlin with Diversity Expert Annette von Wedel. The vision: more equality for all people. My goal in doing so: to create ways to implement this vision in concrete terms. To this end, I co-initiated the BMAS research project "KIDD - Artificial Intelligence in the Service of Diversity".

What is the KIDD project in more detail?

The research project "KIDD - Artificial Intelligence in the Service of Diversity" started at the end of 2020 and is funded by the Federal Ministry of Labor and Social Affairs under the umbrella of the New Quality of Work Initiative (INQA). In the three-year project period, we have developed quality criteria, training and a concrete procedure, the so-called KIDD process. The KIDD process, with its four stages, is based on a classic software development or implementation process. At the heart of the process is the so-called "Panel of Diversity" (PdV) as an internal committee of the company that wants to introduce an AI system. It should bring together as many relevant stakeholders as possible in order to uncover any distortions that the system might exhibit as a result of their different perspectives. On this basis, companies can then introduce algorithmic decision-making systems such as artificial intelligence in a transparent and diversity-sensitive manner as part of a self-audit.

How was the idea for the project KIDD - AI in the service of diversity born, what problem did you want to solve with it?

The idea was born at the very first female.vision summit, in the workshop "Artificial Intelligence - without human prejudices" by KIDD co-initiator Rosmarie Steininger. There, it quickly became clear to the participants that with AI, unlike with conventional products, it is not feasible to check the quality of the final product only after it has been completed, because by then it is too late. This is because unintentional biases and discrimination have already been implemented by then. The resulting considerations of how "pre-programmed" unequal treatment and unconsciously distorting algorithms could be detected and prevented at an early stage by a suitable procedure, especially in companies, gave rise to the idea for the KIDD process and the research project.

How does discrimination through AI manifest itself in concrete terms?

Discrimination by AI can in principle affect anybody. Since the underlying data is often biased and disadvantages certain groups, there is a risk that algorithmic decision-making systems such as AI reproduce or even reinforce this discrimination. This can have undesirable consequences in companies, for example in HR decisions or recruiting. In order to uncover further potential patterns of discrimination, our partners have brought very different approaches for the use of algorithmic decision-making processes to the project. These include the use of algorithms in HR management, sales intelligence on customers, and the automated evaluation of services. In the process, we have discovered that the four stages that the Diversity Panel goes through as part of the KIDD process provide very positive impulses for the entire corporate culture in terms of diversity and non-discrimination beyond the application.

The project brings together a large number of project partners. How has that worked so far?

The collaboration on the project is working extremely well and is extremely targeted because it is so practice-oriented. A consortium of eight partners is involved nationwide. The Berlin-based nexus Institut für Kooperationsmanagement und interdisziplinäre Forschung GmbH is coordinating the project, the TU Berlin is responsible for the scientific evaluation, and female.vision is managing the areas of communication and sustainability and contributing the relevant expertise on data ethics and diversity. Five companies from the business sector are involved. These include Heraeus Medical GmbH, software manufacturer msg systems AG, Munich-based CHEMISTREE GmbH together with Q_PERIOR AG, and Hamburg-based epsum GmbH. These companies represent companies of different sizes and industries, SME`s as well as internationally active large corporations. They have all tested the KIDD process directly in practice as experimental spaces, have been involved in the development from the beginning and thus make a valuable contribution to bringing the joint research results into the field at the end of the project period.

How is the project being received by companies and what successes have been achieved since its launch in fall 2020?

Fortunately, there is an ever-increasing interest. On the one hand, this is related to the current discussion about chat GPT, in which many companies have recognized how close AI applications already are to everyday business life. At the same time, it is becoming apparent that the EU will soon lay down a binding legal framework for the application of artificial intelligence in the form of the AI Act. This will affect both the providers and the users of AI systems. German companies will then be obliged, depending on the level of criticality of the application, to introduce binding procedures for the introduction of human-centered AI in order to avoid high penalties under certain circumstances. We have developed such a procedure with the self-audit based on the KIDD process and successfully tested it in practice during the course of the project. High-profile members of our Advisory Board, such as Lothar Schröder, a member of the Supervisory Board of Deutsche Telekom, and Kenza Ait Si Abbou of IBM, have pointed to the KIDD process in their recent publications as a solution approach for companies that want to live up to their digital responsibilities. And the DIN Deutsche Institut für Normung e.V. (German Institute for Standardization) has also mentioned the KIDD process several times as a best practice example in the current edition of the Standardization Roadmap Artificial Intelligence (AI). This puts us a big step ahead of current developments.

How do you see Berlin as a location and the AI ecosystem?

Berlin is an ideal location for our research project. Without the vibrant ecosystem of science and research, civil society initiatives and associations, and a broad corporate network, it would hardly have been possible to implement our research project within the relatively short planning period. Throughout the entire duration, we were able to benefit again and again from important impulses from the Berlin AI landscape. We were able to present the results of our research work in the Ludwig Erhard House at a central point of Berlin's corporate landscape and are now looking forward to the traveling exhibition, which will make stops at the Berlin Chamber of Industry and Commerce and the Berlin Digital Agency, among others.  Since the next step is to translate the research results into business practice, we are already thinking about founding a KIDD Academy and certifying the process with some of our partners. I am sure that we are also on the right track with a start-up on this topic in Berlin!

Thank you very much, Ms. Anclam, for the exciting insights into the work of KIDD and the importance of inclusive digitalization.

For more information on the project, please visit: www.kidd-prozess.de