Expert Opinion on Ethics of AI in Evaluations

The projectZukunftswerkstätten Evaluation und Qualitätssicherung in der Extremismusprävention, Demokratieförderung und politischen Bildung:Analysis, Monitoring, Dialog (PrEval)is a research project that develops formats and structures to strengthen evaluation in extremism prevention, democracy promotion and political education in Germany. The project is funded by the Federal Ministry of the Interior and Homeland (BMI) and will run from October 2022 to December 2024. PrEval is supported by a network of relevant research institutes and practitioners and is designed as a broad network, which is reflected in the involvement of other relevant actors in analysis and knowledge transfer formats. As part of the "Monitoring and Data Analysis" work package, expert reports will be written to systematically analyze and process new cross-disciplinary and cross-discourse developments in evaluation research, with a particular focus on the fields of preventing extremism, promoting democracy, shaping diversity and political education. The expertise to be awarded here is to be prepared on the topic of "Ethics of Artificial Intelligence in Evaluations". Artificial Intelligence (AI), including Large Language Models (LLMs) such as ChatGPT, use deep learning techniques and huge text datasets to understand, generate and predict content. They offer a wide range of applications, including text summarization, key information extraction, text analysis, classification, generation and translation. Despite their potential, LLMs pose challenges for democracy and social participation such as reinforcing biases and distorting facts, potentially limiting their accuracy for evaluation purposes. To mitigate these risks, close collaboration between the evaluation and data science communities is needed to reflect tools and processes and ensure quality and ethical standards. Expertise - Fachbeitrag - Fachreferent