Description of the project
The research project focuses on the methodological improvement of the Center's surveys, based on the development, evaluation and application of new methodologies linked to experimental designs, machine learning and artificial intelligence.
1. Evaluation of the recruitment strategies of the Citizen Panel of Catalonia through the use of experimental methodology (RCTs) to study the effect of different strategies (reminders, incentives, framing of letters, multi-channel approaches, etc.) on the rate of response and in the representativeness of the samples. Use of machine learning methods to estimate the expected effects of treatments for different profiles of individuals. Development and optimization of algorithms. Development of new adaptive survey designs to optimize recruitment strategies based on the results of RCTs and prediction algorithms. Evaluation of adaptive design models.
2. Application of the extended language models (LLM) of artificial intelligence to the analysis and coding of open questions. Prompting, evaluation and development of solutions. Comparison with alternative strategies (human coding and topic models of natural language processing).
3. Development and use of machine learning models for small area estimation problems, to generate estimates of public opinion at local or regional level.
4. Analysis and development of the longitudinal aspect of the Citizen Panel of Catalonia, with special attention to the attrition rate and the consequences on representativeness
