ESRA 2027 sessions by theme
Back to Overview of Sessions
From Respondents’ Own Words to Valid Survey Measures: AI in Data Collection and Coding |
| Coordinator 1 | Ms Lovisa Moller Vallgarda (National Centre for Social Research) |
| Coordinator 2 | Professor Julia Tuppat (Leipzig University) |
| Coordinator 3 | Dr Jacqueline Kroh (Leibniz-Institut für Bildungsverläufe (LIfBi) ) |
| Coordinator 4 | Dr Andreas Niekler (Leipzig University) |
| Coordinator 5 | Dr Poppe Stephan (Leipzig University) |
Session Details
Open or naturally expressed responses can capture experiences and perspectives that predefined categories may miss, but integrating them into large-scale surveys has traditionally been constrained by standardisation and scalable processing. Recent advances in AI may change this trade-off: conversational tools can support flexible, respondent-centered data collection, while AI-assisted coding can translate complex verbal responses into structured survey data at scale.
Together, these developments raise a methodological question: can greater flexibility in eliciting and processing responses produce valid, reliable, comparable measures underpinning large-scale social research?
This session examines both sides of this challenge. Moving beyond proof-of-concept, it asks whether, when, and for whom AI-enabled approaches improve survey measurement. It also considers potential risks, gains and losses when standardisation shifts from fixed questions and response categories to AI-mediated elicitation and interpretation.
On the data collection side, contributions may examine:
Whether dynamic, AI-driven questioning improves accuracy of complex categorical variables by enabling respondents to describe their responses rather than categorizing themselves;
How do conversational approaches to attitudinal measurement compare with fixed scales;
How adaptive questioning affects respondent experience, including cognitive burden, relevance, trust, and willingness to disclose;
Whether open-response and AI-moderated approaches can broaden participation among groups for whom standardised formats are particularly demanding.
On the analysis side, contributions may address:
Methods for coding complex, multi-dimensional, ambiguous, or context-dependent responses;
The design, validation, and evaluation of AI-assisted and semi-automated coding pipelines, including accuracy, reliability, and comparability to human coding;
Bias, fairness, and implementation within the operational, ethical, and quality constraints of large-scale survey infrastructures.
We welcome empirical, methodological, and applied contributions that move beyond feasibility to examine implications for validity, comparability, bias, respondent experience, inclusion, research efficiency, and survey practice.
