Chatbots help us in many situations – they generate text for us, answer our questions, and we can also just talk to them. Such an (often helpful) conversation can lead to an one-sided emotional bond with expectations that a chatbot can’t meet. But what happens when we humanize chatbots like that? A media report published by Mitteldeutscher Rundfunk (MDR) on February 13, 2026 looks into this question. The experts interviewed for this media report include:

• Prof. Michael Färber, Professor for Scalable Software Architectures for Data Analytics at TU Dresden and ScaDS.AI Dresden/Leipzig
• Janosch Delcker, journalist and author
• F. Gerrik Verhees, Psychiatrist at University Hospital Carl Gustav Carus Dresden
• Prof. Stephen Gilbert, Else Kröner Professor for Medical Device Regulatory Science at Else Kröner Fresenius Center for Digital Health and TU Dresden

Prof. Michael Färber explained that modern chatbots are usually based on large language models that generate text based on context and previous words (tokens). Answers are therefore not already stored somewhere, but are generated by the model and its parameters. For this technology, “methods were developed that were crucial in determining what the user actually wanted”, says Prof. Michael Färber. At the base of this technology lays statistics – and yet there were cases where people attributed feelings to chatbots and developed feelings for them.

According to F. Gerrik Verhees, it is deeply human to assume that the other person is like ourselves, even if the other person is a chatbot or a fictional character. Unfortunately, forming such an emotional bond can have fatal consequences in real life. Besides legal regulations, built-in technologies can help to protect people. Read the full media article and watch the accompanying video for more information.