AI Dependency, Cognitive Offloading, and Academic Stress among University Students: A Survey Dataset of 148 Responses in Bangladesh perspective.

Abstract

This dataset contains 148 questionnaire response records on AI use, cognitive offloading, and academic stress among university students. Prepared for a thesis on AI dependency and academic stress, it provides 43 variables for analysis and exploratory machine learning. Each row is one response. Responses were collected through an English-language Google Form shared with friends via personal email. Recorded submission dates span 21 July to 3 September 2026; the timestamp timezone is unknown. Recruitment followed a personal network, not a reported probability-sampling procedure. Geographic eligibility, invitation counts, and response rate were not supplied; representativeness should not be assumed. Variables include a generalized self-reported CGPA band, daily internet-use category, seven binary AI-tool indicators, academic AI-use frequency, and 27 agreement items coded from 1 (Strongly Disagree) to 5 (Strongly Agree), without reverse scoring. Five demographic fields describe age band, gender, university, department, and year of study. Reassigned respondent IDs support record management and should not be predictors. All 148 records are retained. The owner confirms that quantitative answers came from actual participants. No synthetic observations or missing-value imputations were introduced. Participant emails, exact timestamps, two open-ended answer fields, and eight empty columns are excluded. The two lower CGPA categories are combined into “Below 3.00.” University and department spelling variants were harmonized. Demographic details were generalized or withheld to reduce identification risk: each released five-field combination occurs in at least five records. Eighty-three records contain at least one “Not disclosed (privacy)” value. This denotes withheld information, not a participant response or an imputed value. The check covers only these demographic fields and does not establish anonymity of complete questionnaire patterns. The owner supplied an approval statement naming the Institutional Review Board / Ethics Committee of Bangladesh Army University of Engineering & Technology (BAUET), reference BAUET/REC/2026/13-06, dated 13 June 2026 and endorsed by Ananya Sarker. Its publication scope covers fully anonymized individual-level data. The original signed approval was not independently inspected. Participant consent, including consent for the under-18 participant, and completion of human verification are reported by the owner. The package includes an Excel workbook with Data, Codebook, README, and Authors sheets, an equivalent UTF-8 CSV, and documentation. Researchers should define outcomes, account for withheld demographics, and fit preprocessing within training data to avoid leakage. The small sample limits evaluation and generalization. No trained model, performance result, or causal conclusion is supplied. The owner authorizes release under CC BY 4.0.