
Interactive reminder: Ludwig-Maximilians-Universität München
1. Initial Situation
Shortly after World War II, and especially in the 1980s, efforts were made to document and preserve the memories of Holocaust survivors through various media forms (e.g., audio interviews and audiovisual oral history interviews). Due to the length of many of these interviews, they are often not ideal for classroom use. Around the intellectual legacy of Holocaust survivor Ernst Otto Krakenberger, an interactive media format was developed to engage younger students, who are already familiar with everyday communication via digital platforms. The decision was made to use an AI chatbot. In this medium, users can now ask topic-related questions, allowing them to engage with the content actively rather than passively watching the entire interview.
2. Challenge & Implementation
First, an interview was conducted with Holocaust survivor Ernst Otto Krakenberger. The interview was then transcribed, fact-checked, and conceptually adapted for use with generative AI. In this process, incomplete speech was supplemented, names and events were corrected where necessary, and grammatical and semantic adjustments and additions were made. The interview was then split in the aiSuite, with answers stored in a topic catalog and supplemented with corresponding audio files. In addition, a list of questions with thematic fields was created, providing contextual information about the stories shared and enabling a more comprehensive engagement with the Holocaust memories. The prototype chatbot, powered by generative AI, was already very impressive in terms of content. However, it was also noted that the response behavior changes when the language model is updated, requiring regular checks and adjustments.
3. Results & Outlook
The hypothesis that user behavior changes through independent engagement with content in the form of written questions was confirmed in the first test groups. There is evidence of individual and multifaceted engagement with the topics – ranging from emotional to factual interest. Personal attitudes toward political and historical events can also be inferred from the user dialogues, often with references to current events. Given the widespread use of ChatGPT and similar tools, there are also indications of a fundamental societal shift in the way media content is approached and interacted with.

Florian Duda
Research Assistant LMU Munich
“Thanks to aiSuite, we were able to work on our AI application very independently right from the start and dynamically design valuable eyewitness interviews in combination with generative AI for posterity.”
Discover more AI use cases
Chatbots for flexible trade fair communication
Find out how the trade fair provider was able to automate recurring requests for different trade fair events.
Generative AI in the health portal
Find out how the integration of generative AI improves the health portal and optimizes the use of information by customers.
AI for efficient citizen communication
Find out how the Warendorf district administration was able to improve the customer experience by making an AI chatbot available 24/7 in its service portal.
© 2024 Kauz GmbH. All rights reserved.