Smart surveys use the features of smart devices in collecting and processing data. Smart features are real-time in-device application of pre-trained AI-ML, the use of mobile device sensors, the linkage to (wearable and/or fixed) sensor systems, the linkage to online public data and the donation of private online data. Doing so, smart surveys combine traditional survey questionnaire data with new forms of data. Given the investment in methodology, IT and infrastructure, smart surveys are especially promising for survey topics that are cognitively burdensome, require knowledge that is non-central to most respondents, or for which questions provide weak proxies of the concepts of interest. In other words, smart surveys aim both at reduction of drop-out from studies and at better measurement data quality.

 

Smart surveys bring new challenges to methodology in all stages from design to analysis and estimation. Some challenges are modern versions of research questions that have been central to survey research for many years, others come as specific to smart surveys. Many of the research questions are only modestly country-dependent. For this reason, Eurostat-funded project Smart Survey Implementation (SSI) set out to develop, test and evaluate smart survey case studies. SSI is a consortium of seven national statistical institutes, three universities and a spin-off of one of the universities.

This invited paper session focusses on the four main methodology research areas in smart surveys:

• Strategies to involve and engage respondents in smart surveys

• Application of AI-ML techniques to smart data to improve the quality of predictions 

• Trade-offs in active – passive involvement of respondents in checking and editing smart data

• Estimation and evaluation of method effects when going smart

 

In SSI, cross-country case studies have been conducted throughout 2024. Findings of these studies, along with those of other studies, are presented in the session.