Abstract

With the rapid evolvement of new technologies in recent years, such as ChatGPT and other AI systems, knowledge can be more easily and faster acquired. This new technological development both impacts everyday life and work. The use of AI in the field of survey methodology can enhance various aspects of the survey process and has already been used to some extent in various parts of the survey process. Experiments of implementing elements of AI have been done both before, during and after the data collection process of a survey. The use of AI before the survey is conducted, such as questionnaire design and sample design, have been studied by Statistics Norway and other institutions. Experiments have been done how well chatbots can improve question formulations, summarize test person results, or create synthetic test persons. During the data collection process, chatbots may assist or replace the survey itself for collecting information from respondents and creating a more natural and humanlike answer process. Other possibilities are to use machine learning algorithms to adjust the questionnaire content or design to the respondent’s needs and preferences. After the data are collected, post data processing procedures are performed. In Statistics Norway and other institutions, machine learning algorithms have been used to both structure and edit data.
A strong paradigm in survey design is to create respondent centred surveys that attend to the respondent’s needs and abilities. In a respondent centred survey design, survey methodologists test survey questions on test persons to gain insight into if the questions are understandable and easy to answer, if they are relevant and engaging for the respondent, or if they are difficult to answer or comprehend, or sensitive and may therefore require adjustments to create a more respondent friendly survey. However, with the rise of new technologies, data may be more efficiently analysed and used than before, and this may open up new opportunities to adjust the questionnaire design even during the data collection process itself.
In this paper, I want to examine if and how new technologies can be incorporated to the data collection process. More specifically, I will present how new technologies can be harnessed to improve the respondent experience and data quality shown on two examples. First, I will outline the use of machine learning algorithms based on paradata and sample data to detect high response burden and how the results can be utilized to adjust better to the respondent’s needs. For this, I will use the Adult Education Survey as an example. Second, I will draft the steps for implementing a chatbot in the Labor Force Survey.
When using new technologies, many challenges and limitations, such as time, resources and ethical issues need to be considered as well as the public’s trust of using these new technologies. In this paper I will present some of the concerns, challenges and limitations and how one may handle them. Yet, some of the challenges remain to be solved which will also be addressed.