Abstract

Strengthening critical thinking abilities of citizens in the face of news published on the web represents a key challenge for education. Young citizens appear to be vulnerable in the face of poor quality news or those containing nonexplicit ideologies. In the field of data science, computational and statistical techniques have been developed to automatically collect and characterize online news media in real time. Nevertheless, there is still not a lot of interdisciplinary research on how to design data exploration platforms supporting an educational process of critical citizenship. This article explores this opportunity through a case study analyzing critical thinking ability of students when facing news dealing with the social mobilization “No+APF”. From data collected through 4 online exercises conducted by 75 secondary school students, 55 university students and 25 communication specialists, we investigate to what extent young citizens are able to classify news headlines and ideological orientation of news media outlets. We also question the influence of the media’s brand name and the subjectivity of each participant in regards to the social mobilization “No+APF”. The results underline the importance of group work, the influence of the brand name and the correlation between criticalthinking abilities and having a defined opinion.

Abstract

Abstract

Abstract

The Internet is a public source of information, where information can be found at minimum search cost. Social media are becoming increasingly popular among web users trying to find and analyze information about the current economic situation. Web users get the opportunity to exchange views or discuss various issues in the news communities of social networks. This information can be used by economic agents to make decisions. Thus, the study of user behavior in social networks makes it possible to identify the expectations and preferences of economic agents.

The goal of this study is to assess the expectations and sentiments of economic agents based on textual analysis of social media data.

The study addresses the following objectives:

Analysis of the mechanisms of influence of the information dissemination and networking effects on the behavior of economic agents;

Systematization of the results of theoretical and empirical analysis of the economic agents’ expectations;

An overview of machine learning methods used in text processing;

Development of an algorithm for identifying sources of information for web scraping and rules for selecting text information to create a body of posts and comments;

Collecting a database and preparing posts and comments for text analysis;

Application of topic modeling to the identification of topics and keywords in social media data;

Assessment of high-frequency indicators of the public sentiment.

The subject of the research is a quantitative assessment of the sentiment of web users based on Russian data.

The novelty of the study is the assessment of inflation expectations, sentiments in the foreign exchange market and indices of economic conditions using structured and unstructured internet data. Methods: topic modeling; machine learning methods and econometric methods of time series analysis.

The study is based on data for Russia in 2014-2021. The study shows that social media posts, search queries and online news articles can be good proxy variables for the economic agents’ expectations. We construct three types of public confidence indicators based on internet data: inflation expectations; sentiment in the foreign exchange market and index of economic conditions. The results of econometric analysis indicate that the quality of macroeconomic performance models with sentiment indicators is higher than without these indicators. Additionally, indicators based on VK posts, RBC news articles and Google Trends search queries are more informative compared to comments. The main conclusion of the study is that internet data can improve the quality of macroeconomic performance models. In a further study, we plan to expand the list of indicators of the sentiment of economic agents and to evaluate advanced time series models