How AI helps to better understand children's health: AI-SIC project develops new methods for data analysis
How do children and parents think about children's health – and how can artificial intelligence (AI) analyze their attitudes? In the AI-SIC project, researchers from Bamberg and Leipzig are combining qualitative interviews with machine learning. The goal is to develop a semi-automatic process that makes the evaluation of interviews significantly more efficient – and thus sets new standards in health research.
In qualitative interviews, participants answer open-ended questions on a specific topic. In order to evaluate their individual answers scientifically, they must be systematically sorted into categories – a process known as coding. It is precisely this step that is considered a bottleneck in the evaluation of qualitative data – it costs a lot of time and human resources. And it is precisely this step that is to be supported by artificial intelligence (AI) in the future: In the AI-SIC (Artificial Intelligence in Survey Interview Coding) project, researchers from the Leibniz Institute for Educational Trajectories (LIfBi) and the University of Leipzig want to develop a semi-automatic process that uses artificial intelligence to evaluate interviews more efficiently. At the same time, AI-SIC should provide deeper insights into how children and parents perceive children's health.
“Qualitative interviews allow us to gain a deep understanding of what people think and feel,” says Dr. Jacqueline Kroh, who heads the project at LIfBi. “But evaluating such conversations is laborious, expensive, and time-consuming. That's why only a few people are often interviewed for a study. If we can use AI to make some of the work easier, we will be able to include many more people in the future. For health research in particular, this means that we can get closer to people's real lives—and help them even better.”
The special feature of the process lies in the combination of human expertise and machine learning: First, some of the interviews are coded by human experts. In the next step, this is used to train an AI model, which then codes the remaining interviews semi-automatically on its own. Semi-automated means that when uncertainties arise, the AI asks the experts to make the correct assignment. This, in turn, allows the model to continue learning and refine its accuracy.
While the development and implementation of the semi-automatic coding process is mainly based at the University of Leipzig, an important part of the content analysis of the data obtained is carried out at LIfBi. The focus here is on how children and parents assess their own general state of health. Differences between self-assessments and external assessments are also examined, as well as the influence of age and gender.
AI-SIC is funded by the German Research Foundation (DFG) as part of the infrastructure priority program SPP 2431 “New Data Spaces for the Social Sciences.”
Wissenschaftlicher Ansprechpartner:
https://www.lifbi.de/en-us/Start/Institute/People/Person/account/1373?name=Kroh%2cJacqueline Dr. Jacqueline Kroh
Weitere Informationen:
https://www.lifbi.de/en-us/Start/Research/Projects/AI-SIC Project information
Die semantisch ähnlichsten Pressemitteilungen im idw
