Actionable companion · Assessment & reasoning
Learning with agency: An Alternative Approach for Detecting Problematic Alcohol Use: Developing the Student Alcohol Risk Assessment Scale‐15 Using AI‐Assisted Machine Learning
AI-assisted editorial implementation guide, separate from the original research. Proposed actions are not claims or recommendations attributed to the source authors. No independent peer review of this guide is asserted.
Evidence and context
Title and abstract screening; stage labels describe reported settings, not verified numerical ages. Design labels require full-text appraisal. Reported stage(s): Higher education. Verify the participant population and task before transferring the findings.
Original authors’ abstract, reproduced under the linked source license:
ABSTRACT Background/Aim Risky alcohol use is common among university students and negatively impacts physical and psychosocial health. Current screening instruments focus only on behavioral indicators, neglecting psychosocial factors. This study aimed to develop an AI‐supported, brief student‐focused tool for assessing alcohol‐related risk and to evaluate its preliminary reliability and validity among university students. Materials and Methods A total of 599 university students participated in this cross‐sectional study. From a pool of 59 items covering behavioral, psychological, social, and academic questions related to alcohol use, a new risk model was created using 15 items selected with ChatGPT‐supported AI algorithms (SARAS‐15). Risk scores ranged from 0 to 27, categorizing participants into low, medium, and high‐risk groups. The model's performance was evaluated using machine learning methods such as Random Forest, Logistic Regression, SVM, and KNN, along with cross‐validation and ROC analysis. We also analyzed its internal consistency and its correlations with screening instruments (AUDIT, RAPS4‐QF, CAGE). Results In machine learning analyses, the logistic regression model achieved the highest performance (93.5% accuracy, F1‐score = 0.864, sensitivity = 0.826, specificity = 0.959). ROC analysis demonstrated excellent discrimination between low‐risk (AUC = 0.96) and high‐risk (AUC = 0.93) groups, with strong discrimination for the intermediate‐risk group (AUC = 0.88). The Random Forest model achieved an overall accuracy of 87%, successfully differentiating between the low‐risk group (F1 score = 0.91) and the high‐risk group (F1 score = 1.00). The new model's Cronbach's alpha was 0.811, with strong convergent validity correlations with screening instruments (AUDIT, r = 0.861; RAPS4‐QF, r = 0.793; CAGE, r = 0.631). Conclusion The developed artificial intelligence‐supported 15‐item new risk score model is valid and reliable in assessing risky alcohol use among university students. This scale demonstrates a high level of agreement with traditional tests and can accurately detect three risk levels. Due to its multi‐domain content structure, which addresses both behavioral and psychosocial effects, it serves as a complementary screening tool for early risk identification and intervention planning.
What kind of evidence is this?
Supporting AI learning science · Survey / observational study
Stage: Higher education. Screening labels describe reported context and study-design language. They are not a risk-of-bias rating or proof of effectiveness for an age band.
Study methods · source passage
A cross‐sectional and descriptive study was conducted with university students at xxx, xxx, during the 2023‐2024 academic year (June 2024) and the 2024‐2025 academic year (August and September 2024). The survey, which consisted of 81 questions and was administered online, was accessed via a link provided to students by student representatives and deanships, or by scanning the QR code on a poster describing the study displayed on university boards. Informed consent was obtained from all participants at the beginning of the study, after they were provided with information about the research and asked if they wished to proceed. Only participants who chose “yes” were taken to the questionnaire page, where they could quit at any time. Furthermore, a consent form was provided before the survey to ensure anonymity. The study was open to all students enrolled at the university. Participants did not receive any financial compensation, academic credit, gift, or other incentive for completing the survey. All participants completed the survey within 20–30 minutes. All 599 participants completed the mandatory survey items, and no participants were excluded because of missing data.
Limitations and uncertainty · source passage
Internal consistency was evaluated according to each instrument's response format. For the SARAS‐15, Cronbach's alpha was reported for comparability with previous alcohol‐screening research, and McDonald's omega was calculated as a complementary reliability coefficient for the total risk score because the instrument includes both dichotomous and ordinal items. For the dichotomously scored screening instruments, namely CAGE and RAPS4‐QF, internal consistency was reported using the Kuder–Richardson Formula 20 (KR‐20). KR‐20 is mathematically equivalent to Cronbach's alpha when all items are dichotomously scored. Corrected item‐total correlations were also examined for SARAS‐15 items, with values above 0.30 considered acceptable. The construct validity of the new risk model was evaluated by computing Pearson correlation coefficients between the model's total score and scores from the AUDIT, CAGE, and RAPS4‐QF (convergent validity). To determine the model's generalizability, 10‐fold stratified cross‐validation was conducted using attributes derived from the 15‐item scale.
Design the system before the pilot
Use this study to inform a learning mechanism or implementation question. It is not direct evidence that autonomous agents improve learning. Keep the proposed agent behavior separate from what the source actually tested.
- Use common rubrics, blind scoring where feasible and a fresh unaided reasoning task. Document permissible assistance and avoid treating AI-detector scores as proof of misconduct.
A 90-day district pilot
- Days 1–15 · Appraise. Have the curriculum lead and a teacher read the methods and limitations. Record the study design, sample, comparator, outcomes and evidence gaps. Define a local question that the study can help inform.
- Days 16–30 · Design. Choose one disciplinary task. Require a first attempt, a documented AI consultation, a critique of the output and an independent explanation. Align the comparison group and scoring rubric with the same objective. Name an accountable owner and set the baseline, success threshold and budget before implementation.
- Days 31–60 · Test. Start with a small voluntary cohort, preserve an accessible non-AI route and require teacher review. Use synthetic or de-identified examples; approve data handling and tool terms through district processes.
- Days 61–90 · Decide. Blind-score reasoning and accuracy on a new task without AI. Distinguish model examination scores, student perceptions and actual learner gains; only the last directly tests educational benefit. Compare with baseline, document uncertainty and stop or revise if learning worsens, access is unequal or critical errors persist.
The board decision brief
Report the learning problem, evidence quality, participating population, total cost including review time, measured benefit, errors, access differences and remaining uncertainties. Make a dated recommendation to stop, revise or expand. A positive perception survey alone does not establish learning improvement.
Attribution and reuse
Şükrü Alperen Korkmaz; Pınar Mutlu; Sibel Oymak; Gamze Çan. An Alternative Approach for Detecting Problematic Alcohol Use: Developing the Student Alcohol Risk Assessment Scale‐15 Using AI‐Assisted Machine Learning. International Journal of Methods in Psychiatric Research (2026-09-01). https://doi.org/10.1002/mpr.70096
Companion prepared 2026-10-30. Original authorship and DOI are preserved. The abstract is an attributed source extract; the implementation plan is new AI-assisted editorial material, offered under the same CC BY 4.0 license. This guide has not undergone independent peer review.