AI interaction and cognitive effects

Dataset: AI interaction behaviors and unaided reasoning and ideational fluency in secondary school students Anonymized student-level data for the study AI interaction behaviors and unaided reasoning and ideational fluency in secondary school students (exploratory cross-sectional study, N = 150). File: observed_150_analysis.csv (150 rows, 22 columns, UTF-8, comma-separated, one row per student, no missing values) Citation: [Author]. ([Year]). Data for: AI interaction behaviors and unaided reasoning and ideational fluency in secondary school students [Data set]. [Repository]. [DOI] License: [CC BY 4.0 / restricted access: choose one] Contact: [corresponding author email] Participants and collection Students aged 14–18 (M = 15.30, SD = 1.03) from several classrooms of a private school in Madrid, Spain: 3º ESO (n = 56), 4º ESO (n = 52) and 1º Bachillerato (n = 42). Recruitment was by convenience sampling. Students completed a self-report questionnaire and written tasks under supervision, without AI assistance. Parents or legal guardians gave signed informed consent, and the study had institutional ethics approval. Anonymization Participants are identified only by a sequential code (P001–P150). The file contains no names, contact details, school or classroom identifiers, dates, or free-text responses. Variables Column Description Values participant_code Anonymous sequential code P001–P150 Edad Age in years 14–18 Curso School year 3º ESO, 4º ESO, 1º Bachillerato Frecuencia_IA Self-reported frequency of AI use for schoolwork Nunca (never), Una vez a la semana (weekly), Cada dos días (every two days), Diariamente (daily) ai_frequency_rank Numeric coding of Frecuencia_IA 0 = never (n = 14), 1 = weekly (n = 26), 2 = every two days (n = 62), 3 = daily (n = 48) interaction_index_exploratory Exploratory interaction index = mean of P4, P5, P6. Study-specific, not a validated scale 1–5 P1–P8 Questionnaire items (see below) 1 = strongly disagree … 5 = strongly agree T1_IA_aprendizaje_pts Claim-evaluation prompt about AI and learning (reasoning) 0–5 T2_botella_pts Alternative uses of a plastic bottle: number of credited uses 0–9 T3_mundo_sin_palabras_pts Hypothetical prompt about communication in a world without words 0–5 T4_ladrillo_pts Alternative uses of a brick: number of credited uses 0–11 T5_recursos_digitales_pts Claim-evaluation prompt about digital resources (reasoning) 0–5 fluency_total_supplied Combined ideational fluency = T2_botella_pts + T4_ladrillo_pts 0–20 reasoning_two_prompt_mean_supplied Focal reasoning outcome = mean of T1_IA_aprendizaje_pts and T5_recursos_digitales_pts 0–5 three_task_mean_supplied Broader composite = mean of T1, T3 and T5 (used only in sensitivity analyses) 0–5 Questionnaire items (1–5): P1 perceived efficiency; P2 generation of complete answers; P3 dependence on AI for organizing ideas; P4 verification of information; P5 comparison with other sources; P6 revision of generated answers; P7 perceived learning benefits; P8 concern about reduced independent thinking. Column names are in Spanish because they match the original instruments; the Spanish values in Frecuencia_IA and Curso are the original response options. Scoring notes Task responses were rated manually in line with psychological test-design principles. Scores are the ratings used in the analysis; the original written responses are not included. Scoring anchors and inter-rater reliability are not documented (see the article's limitations). The _supplied columns are the scores as provided for the analysis, not re-derived. Zero scores are real values (no credited uses or no creditable response) and were retained in all analyses. Reproducing the main results Spearman correlations (N = 150) between the focal variables: fluency_total_supplied reasoning_two_prompt_mean_supplied interaction_index_exploratory ρ = .255 ρ = .397 ai_frequency_rank ρ = .014 ρ = −.028 The article reports two-sided permutation tests (9,999 permutations, plus-one correction), Holm adjustment across the four focal tests, and percentile bootstrap confidence intervals (5,000 paired participant-level resamples). Use and limitations These are exploratory, cross-sectional, self-report-based data from a convenience sample in one school. Classroom identifiers are not available, so clustering cannot be modeled. The data do not support causal conclusions about AI use. Users must not attempt to re-identify participants.