VZ4: Modern measurement and data | PRINS

Modern measurement and data

Quality of questionnaire data Customised population measurement AI detection of social risks

VZ4.1

Methodological research to improve the quality of data from questionnaire surveys

Principal investigator: prof. Martin Kreidl, Ph.D., KSoc FSS MU

Contemporary society is changing rapidly – people are digitally connected, yet increasingly rely on themselves and emphasise individual choice. For social-science research, this presents a major challenge: how can we capture shifting behaviours, attitudes, and values in such a dynamic environment? This research activity focuses on improving data-collection methods and developing new research tools that better reflect the conditions of an individualised society. The aim is to examine how different modes of surveying (e.g., online vs. face-to-face interviews) influence the findings and to propose more effective ways of measuring key phenomena. Attention is also directed toward topics that have so far remained at the margins of research – such as strategies of family cohesion (e.g., kin-keeping), experiences of uncertainty in partnerships or parenthood, or mental health. The activity also involves the use of AI-based methods to map online risks, such as the spread of misinformation or hate speech, which may threaten social cohesion.

VZ4.2

Personalisation of measurement: Creation and verification of methodological procedures for cost-effective collection of representative population data

Principal investigator: Mgr. Hynek Cígler, Ph.D., INPSY FSS MU

How can we obtain high-quality data on mental health without unnecessary costs or overburdening respondents? This research activity focuses precisely on this challenge. Its aim is to develop efficient and accurate data-collection methods using computerised adaptive testing and planned missing-data designs. These approaches make it possible to significantly shorten questionnaires without compromising measurement quality, while ensuring comparability of results across studies. The team will design and validate procedures that allow researchers to reliably estimate the distribution of outcomes in the population from shortened tests administered to a representative sample – for example when assessing the prevalence of mental-health difficulties. In the same way, methods will be created to enable the interpretation of results from non-representative surveys using a small set of carefully selected questions. The validity of these approaches will be evaluated using publicly available datasets and methodological studies. The research will also generate concrete insights into the prevalence of mental-health problems within the Czech population, including differences by education and gender. An additional outcome will be a publicly accessible database of psychological items with verified psychometric properties, available for further research in this field.

VZ4.3

Using artificial intelligence as a method for early detection of online risks threatening society

Principal investigator: doc. RNDr. Aleš Horák, Ph.D., KSUZD FI MU

In this age of information overload, we need tools that can reveal the essence of a message as well as hidden manipulation. This research activity focuses on the use of modern artificial intelligence methods to understand and process texts. Using large language models based on deep neural networks, experts from the Natural Language Processing Center are investigating how to automatically identify manipulative techniques, summarize key messages, and recognize semantic relationships in large language data sets. Particular attention is paid to the Czech language and other Slavic languages, which are often neglected in the development of AI tools. The aim is to develop tools that will be useful not only for scientific research, but also in the media, education, and in detecting disinformation and other forms of linguistic manipulation.