Do you have an opinion on what kind of milk, coffee or meat you prefer? Organic or conventional? Fairtrade or as cheap as possible? Free-range or caged? You probably have a clear idea of which products are in line with your personal ethics! But do you also have an opinion on which algorithms you prefer? None of your business? Are they too complex for you? Form an opinion, because ethics are also crucial when it comes to algorithms!

The good news: transparency is possible! If you can have a say in cows and milk, you can also have a say in AI, because cows and AI have a lot in common:

1. The end product depends on what goes into the system: food or data. This input can be of good or bad quality and may or may not fit in with our personal ethical ideas.
2. The production process can be very different – cows are in the barn or out in the open, farmers are paid fairly or not. The architecture and rules of AI systems favor discrimination or are fair, the underlying assumptions favor one group or exclude another.
3. We all consume what comes out: the milk or the decisions made by algorithms. Both influence us, our bodies, our well-being, our purchasing behavior and our society.
4. We can hardly tell from the result how it was produced, whether it was produced according to our ideas or not. Try tasting milk to see if the farmers were paid fairly.
5. They are complex and the underlying mechanisms are difficult for laypeople to understand – this is certainly true of cows. Very few of us would be able to keep a cow in a species-appropriate manner, feed it healthily or milk it. On the other hand, we all develop and use algorithms every day, for example when we cross the road or think about what we want to eat for breakfast. Algorithms are simply instructions for solving a problem.
6. We rely on transparency in order to be able to make consumption decisions that meet our ethical standards. However, ingredients and production processes are transparent to a considerable extent when it comes to food, but hardly at all when it comes to algorithms – or do you know why a suggestion is displayed at the top on Amazon?

Would it be possible for the level of transparency in the food sector to be transferred to algorithms and artificial intelligence? I think that must be the goal!

Yes, but…

I hear two objections from manufacturers of algorithms and AI.

On the one hand: “That’s our trade secret.” In the food industry, consumer interest weighs heavily, but why should it be any different for algorithms whose results affect us directly? I’m not saying that the source code should be disclosed. But it must be possible to understand what criteria were used to develop it, what quality and ethical rules it is based on – so that we can all decide whether it meets our wishes and values.

The second objection is: “You can’t explain that to a layperson.”If we can make something as complex as milk production understandable to practically anyone, we can do the same with algorithms. Anyone who cannot make their IT systems transparent and explain them to a non-specialist either does not understand what they are doing themselves or has a reason for not wanting to explain them.

… when AI develops itself further

Yes, artificial intelligence becomes difficult to understand when it “learns”, i.e. when it recognizes patterns and probabilities based on input data or feedback and changes its rules independently. But if something can no longer be explained, is it still controllable? We want to be sure that an AI system produces qualitatively and ethically appropriate results! Therefore, there is no way around explainability. The same applies to cows and AI: Transparency is the basic prerequisite for quality and for ethical decisions. And it is possible.