What ethical rules AI needs

Intelligent algorithms can already be found in large parts of our everyday lives. As so-called artificial intelligence (AI), they assist us in almost all activities in our private and public lives, as well as in business. Yet their era has only just begun. It is high time we thought about what ethical rules AI needs so that we are not defencelessly at the mercy of intelligent algorithms in the near future.

Artificial intelligence (AI) is the panacea for all areas of society and industry! At least that seems to me to be the tenor of many publications on the subject. Autonomous driving, more efficient work processes, resource-saving production, energy savings, humanoid robots in care, faster access to relevant information: Intelligent algorithms are gradually penetrating every area of life – they are also being used in sensitive and critical segments, such as medicine or the power supply. And the more they do this, the more important the question of the ethical rules that artificial intelligence follows becomes. Otherwise, there is a risk that this technological progress, which makes sense in principle, will turn into an unregulated juggernaut that makes it impossible for users to decide freely.

The serving function of AI

To prevent this, the serving function of AI must be at the heart of every development. Its use must not lead to constraints that override established ethical norms of coexistence or restrict positive developments. Self-determination, justice and privacy are fundamental values whose integrity must not be restricted by an AI system at any time. The subjugation of an individual or even a company must be ruled out.

[quote]

Worrisome developments

However, many of the applications that are now used on a daily basis are heading in this direction. Be it because corporations such as Amazon, Alphabet or Meta have long since become masters of an enormous amount of highly private and economically relevant data. Or because state-controlled Chinese companies are taking a leading position worldwide in the development of new AI systems. Both facts should worry us: The first, because hardly anyone can understand how the AIs of US corporations use this information. The second, because the Chinese government is not necessarily known for sharing our values; the social scoring system, which will probably soon be part of everyday life for most Chinese people, is inconceivable without the use of AI, to name just one example. In comparison, the Schufa rating is a nice toy at best. And quite a few countries are now flirting with these surveillance options – and not just authoritarian states.

But we don’t even have to look to China to find ethically questionable issues. For example, Amazon’s AI-based job application tool has discriminated against women for years. Discriminatory behavior by algorithms can also be observed time and again in the automatic allocation of loan commitments or insurance policies. I would like to emphasize one thing at this point: AI systems do not tend towards dominant or discriminatory behavior on their own. If their results or predictions lead to individuals being prevented from establishing their rightful personality, negatively influencing them and thus affecting or impairing their ability to access resources, this is due to incorrectly set parameters – in other words, to the people who designed the system.

Initial framework for regulation

As early as 2019, Microsoft and other industry giants therefore pointed out that the misuse of AI algorithms or their poor design poses an ethical and legal risk for the companies using them – and can therefore also have an economic impact. In the same year, the EU published ethical guidelines for trustworthy AI and identified topics that will dominate the design of ethical AI. The key requirements are:

• Human capacity to act and supervision must be maintained when implementing the systems.
• AI systems should enable just societies by supporting human agency and fundamental rights.
• Human autonomy must not be reduced, restricted or misled.
• Transparency and information autonomy must be ensured.
• Private individuals or companies that use AI-based systems must be able to assume that the results and suggestions obtained are benevolent, target-oriented and designed for the intended purpose.
• Algorithms must be secure, reliable and robust enough to avoid errors or inconsistencies in these systems during all phases of their life cycle.

In my opinion, when deciding to use an AI, entrepreneurs should make sure that the system meets these criteria without compromising. Ideally, the desired system should also bear the quality seal of the German AI Association. Its certification rules are based on the aforementioned EU guidelines, among other things. And they should use their power to ensure that binding ethical regulations are codified – ideally with global reach.