Ethical Limits of Artificial Intelligence in the Online Shop

Artificial intelligence (AI) is omnipresent. In e-commerce, it can help to increase the turnover of an online shop through personalization by generating added value for the customer through the targeted use of the collected data. For example, walking shoes can be displayed to a user in an online shop if the trip to the Alps has not even been planned.

Now, a data ethics committee is to find an answer to ethical questions about the use of AI in online shops. An overview.

Which mechanisms are used to make predictions in online shops?

Webshops are less concerned with making explicit predictions about buying, but rather with quickly guiding customers to products that may be of interest to them, and helping them make their purchasing decisions.

Of course, AI is used here. Artificial intelligence has existed for a long time and has become more powerful in recent years, mainly due to the improved hardware.


In principle, two approaches should be distinguished: The interest of the so-called “strong AI” is to create a general, human-like intelligence. This is the direction in which big data processes are used to collect as many detailed user data as possible. These are sourced from all possible channels to model the user and anticipate what he needs. This approach is often found when placing advertisements.

So-called “weak AI systems”, on the other hand, use self-learning methods and semantic methods to solve specific detail problems. In the field of online shopping, this may be such that the AI ​​system is limited to the behavioral traces a user leaves in a specific webshop; Based on extensive product data, these AI systems try to find out what the user is looking for.

If you develop this approach further, you will reach a “dialog-based AI”. This dialog-based AI learns in real-time from the situational reactions of a client to suggestions that the system has made.

It actually mimics the behavior of a human seller in a real business, so it is a kind of “digital seller”. This type of AI is fed mainly with industry knowledge. The processing of knowledge of this kind also uses semantic methods. But also in the placement of advertising, dialog-based AI could be used successfully.

What are the ethical concerns of using AI in the online shop?

For me, the categorical imperative is a good guide when dealing with AI. In other words: “take your counterpart as a person seriously”. The ethical decision-making actually dictates the choice of approach. If I use a dialog-based approach, the privacy of the customer is automatically preserved.

Because I reduce the collected customer data to a minimum.

The operator of dialog-based recommendation services, which are usually offered as “software as a service”, generally does not know the identity of a respective customer at all and does not need to know it: it is the service as a random, artificially generated string ( hashkey).

The user behavior is not tracked cross-shop. As already mentioned, conversely, dialog-based approaches require very differentiated product data that does not affect the customer’s privacy.

Of course, the already mentioned “offer of hiking boots” does not work with dialog-based AI. For this I have to refer to collected customer data. And then the ethical basis quickly becomes precarious. Because this approach, in my opinion, does not take the customer seriously as a person with appropriate personal rights.

Incidentally, I think it is a myth that recommendations generated by extensive use of customer data are more useful than dialog-based recommendations. Corresponding comparisons that we conducted showed rather the opposite. Dialogue-based methods are not exotic in e-commerce, but quite common.

Teaserbild New

Minimal online or technical enthusiast?

Tinder replaces the bar, Alexa the secretary. The digital world has long since become one with the real world. But everyone does not want to believe reality. How confidently do you move in networked life?

Do the test here

What contribution can the privacy ethics committee make to the government?

In my opinion, it does not help to introduce more and more individual rules on data handling; Rather, one should take a look at the roots of the whole event.

The Gretchen question is whether we want to promote a humanistic image of mankind or whether we follow a deranged technology and scientific belief, namely transhumanism, as some thinkers and companies in Silicon Valley represent it.

This mindset sees technological progress as an extension of human possibilities. In my opinion, this is philosophically naive, but promises enormous profits. If you ask “for the mother of all problems”, then for me that is the unbridled triumphal march of transhumanism, which comes in the form of the “strong AI” in a new escalation counteracts.

The “Social Score” recently introduced in China, for example, should be a haunting warning as to where unbridled technology can lead. In contrast, dialogue-based AI, as we use it in online shops, for example, can serve as an example that AI can also be used ethically.

As these AI systems evolve, we can create an identity that is not only a competitive advantage, but also compatible with our humanist ideals.

About the author: Michael Bernhard is Managing Director of the Karlsruhe-based company epoq internet services GmbH, which he founded together with Thorsten Mühling. After studying physics and philosophy, he focused on the subject of artificial intelligence. Over the last 25 years, he has developed many new algorithms and algorithms in this area, which are also used by the software as a service manufacturer epoq. He advocates dealing with artificial intelligence that respects the privacy of the user.

Leave a Reply

Your email address will not be published. Required fields are marked *