TRUST: Measurement and Explanation (TRUSTME)

Research question/goal: 

The project explored different methodological innovations to evaluate trust measures as well as to explain why some people’s trust is lower than others. We investigated both measures of social and political trust. Landesvatter & Bauer (2024a) compared the validity of older and newer social trust measures using open-ended probing questions. One of the main insights of this study is that more refined and precise questions may not necessarily lead to measurements that better reflect generalised social trust (defined as trust in strangers). Landesvatter et al. (2024b) explored the accuracy of different transcription algorithms for open-ended audio data and showed that Whisper (OpenAI) is the most performant one. Landesvatter and Bauer (2023) investigated whether the amount of information people provide in open-ended questions is higher in text response formats or in voice response formats. We found that spoken answers tend to be longer and slightly more informative than written responses. Landesvatter and Bauer (2024c) explored to what extent political trust judgments are based on emotions. In this study we aimed to measure emotions in an innovative way by analysing the textual content of open-ended audio responses and trying to directly pick up emotional cues in the audio recording using machine learning techniques.

Fact sheet

Funding: 
DFG
Duration: 
2018 to 2024
Status: 
completed
Data Sources: 
Survey Data
Geographic Space: 
Europe

Publications