The partner#

The Institut zur Erforschung von Mission und Kirche (IMK) researches missional church, community development and evangelism for the German-speaking world. It is based in Attersee, Austria, and belongs to the Werk für Evangelisation und Gemeindeaufbau of the Evangelische Kirche A.B. in Austria. The project was led by Maike Ritzer, social science researcher at IMK.

The challenge#

IMK wanted to hear from churchgoers about their own wellbeing and the wellbeing of their congregations, in their own words and at a scale that one-to-one interviews could not reach. Respondents came from congregations in Austria, Germany and Switzerland, from Vienna and St. Gallen to Konstanz, Annaberg and Bautzen, and about half of the congregations were rural.

The Causal Map solution#

We ran the interviews with QualiaInterviews and analysed them with Causal Map.

Our approach#

1. Designing the interview together. Between December 2023 and spring 2024 we wrote the interview guide with Maike and took it through three versions. We tested each one first in English, including runs against an AI-simulated respondent, and then in live test sessions between April and June 2024.
2. Changing what did not work in conversation. The first design asked respondents to name up to five areas of life, rate each from 0 to 6 and then spend 10 points across them. That proved awkward in a chat, so we dropped it. The live interview opened with a few short questions (denomination, congregation size, urban or rural, how often the person attends) and then moved into an open conversation about personal and congregational wellbeing. We added progress messages ("Wir sind jetzt zur Hälfte durch das Interview") so respondents knew how much was left.
3. Fieldwork in German. Fieldwork ran from June to November 2024. The interviewer ran on GPT-4 through the OpenAI API, configured by a system prompt and an interview prompt only, at temperature zero. Respondents never gave names or contact details: each interview carried a random session id and a region key, and the only other data were region, urban or rural, and congregation size. The interviewer was instructed not to accept personal identifying information. Gabriele monitored interviews as they came in. By the end of November 2024 IMK had 208 interviews.
4. Causal mapping. Between January and June 2025 we coded the transcripts in Causal Map with our "radical zero-shot" approach: the AI (GPT-4.1) was given no codebook and labelled each cause and effect in German. It found 2,649 causal links, and 191 of the 208 transcripts contained at least one. We produced a top-level map plus maps by area, region and congregation size. Maike then reviewed the clusters of causal factors and proposed the final labels, grouped into factors that foster flourishing and factors that hinder it.

Results#

• What helps a congregation flourish. The starting points respondents mentioned most were activities for every age and interest group (Angebote für alle Alters- und Interessengruppen), exchange in small groups, authenticity and openness, and good leadership from the pastor. The busiest factors on the map were cohesion in the congregation (Zusammenhalt in der Gemeinde) and personal faith (Persönlicher Glaube/Spiritualität), and both fed a lively, supportive community.
• What gets in the way. Impaired wellbeing (Das Wohlbefinden ist beeinträchtigt) was the strongest outcome on the map. Respondents traced it to too few resources, staff or activities, to overloaded volunteers and to conflict, from disputes between pastors to members who were quick to judge.
• Differences between groups. The urban and rural maps, and those for small, medium and large congregations, told much the same story. The regional samples were uneven (57 interviews from Vienna, 4 each from Bremervörde and Dresden), so we compared regions with care.
• IMK also analysed the transcripts by hand with a content analysis, alongside the causal maps.
• IMK has written a paper on the AI interviews themselves, now [in revision after peer review / accepted in JOURNAL].

What Maike says: