Internet-based surveys have expanded public-opinion data collection but have weakened researchers’ ability to monitor respondent attentiveness, creating a growing need for ex-post evaluation of data quality.
⏱️ How attentiveness is measured
Introduces response-time attentiveness clustering (RTAC), a new proxy that applies dimension reduction and an unsupervised clustering algorithm to response-time data. RTAC leverages variation in response time both between respondents and across questions to classify attentiveness without adding items to the survey instrument.
Key features:
- Uses dimension reduction to summarize response-time patterns.
- Applies unsupervised clustering to those summaries to identify attentiveness groups.
- Operates unobtrusively, requiring no extra survey space or attention-check items.
🧭 What the theory changes
Argues that the existing dichotomous classification of respondents as simply "fast" or "attentive" is insufficient. Theoretical development shows that this binary approach overlooks important profiles—most notably slow but inattentive respondents—and therefore can mischaracterize data quality.
✅ How RTAC was validated
Validates both the theoretical classification and the empirical RTAC strategy by comparing them to commonly used proxies for survey attentiveness. The validation demonstrates that RTAC captures attentiveness in ways that differ meaningfully from existing methods.
🔎 Why this matters
RTAC provides a practical, scalable tool for improving survey data quality: it enables researchers to detect a wider range of inattentive behavior using only response-time data, and it does so unobtrusively without sacrificing survey space or adding explicit attention checks.
