Political-Value Survey Responses from a Kazakhstan-Based Respondent Sample and AI Chatbots: A Cleaned Combined Dataset

Description This dataset contains cleaned and recoded data for a comparative study of political-value responses provided by human respondents in Kazakhstan and by conversational artificial intelligence systems. It was prepared within the research project AP26104035, The Influence of New Digital Media on the Political Behaviour of Kazakhstanis in the Context of Artificial Intelligence Development. The deposited package includes a combined dataset integrating two types of observations: responses from a Kazakhstan-based human respondent sample and repeated questionnaire completions generated by three conversational AI systems: ChatGPT, Copilot, and Alice. The current version contains 1,315 records in total, including 1,015 human respondent records and 300 AI-generated response records. The human respondent data included in this dataset were collected in accordance with the ethical requirements applicable to the research project. The study protocol and research instruments were reviewed and approved by the Ethics Committee of the Eurasian National University named after Gumilev, approval/protocol no. 2, dated 29/09/2025. The publicly deposited dataset contains anonymised and recoded data only. Direct identifiers and potentially identifying free-text information have been removed prior to deposition. Both human respondents and AI systems were assessed using the same short political-values questionnaire. The common analytical block consists of twelve 1–10 scale indicators covering: the importance of free elections; attitudes toward strong leadership without parliamentary or electoral constraints; preference for democracy; views on income inequality and personal responsibility; the balance between public order and individual freedoms; justification of breaking an unjust law; attitudes toward traditions and progress; national pride; and trust in scientists, government institutions, and mass media. For human respondents, the dataset additionally includes anonymised and recoded socio-demographic variables, settlement type, aggregated region information, political information-source indicators, and indicators of AI-tool use. For AI-generated records, the dataset includes the chatbot model and run number. Derived indices are included for exploratory comparison of democratic orientation, egalitarian orientation, state responsibility, civil liberties and law challenge, attitudes toward tradition and progress, and institutional trust. The dataset was prepared for methodological analysis of the comparability and non-equivalence of human survey responses and AI-generated questionnaire answers. AI-generated responses should not be interpreted as public opinion or as responses of human-like respondents. Rather, they represent generated outputs of specific conversational AI systems under the questionnaire conditions used in the study. The human respondent sample is a pilot, non-representative sample and should not be used to estimate the distribution of political attitudes in the population of Kazakhstan. It is predominantly composed of younger, urban and student respondents. The dataset is suitable for exploratory comparison of response profiles, methodological discussion of AI-generated survey data, and analysis of how conversational AI systems reproduce normative political-value patterns in comparison with a human respondent sample. For public-data protection, exact timestamps, original source-row identifiers, exact age, raw regional text, and raw free-text responses concerning AI-tool use were removed or recoded. Potential data-quality issues, including straightlining and contradictory information-source responses, were flagged rather than automatically deleted. Users should refer to the accompanying README, codebook and quality log before conducting secondary analysis.