Sürdürülebilir Kalkınma Amaçları
Yazarlar
Yayınlar
- MakaleAçık Erişim2026Scopus SJRQ1How accurately do large Language models interpret sport safeguarding principles: an evaluation using the International Olympic Committee framework
Gökmen, Mehmet Yiğit; Belibağlı, Mehmet Cenk; Karıncaoğlu, Ergin; Nazlıcan, Mahmut Çağatay; Türkmen, Eren; Yönal, Mehmet; Uluöz, Eren
Background: Safeguarding in sport, defined as the prevention of abuse, harassment, and exploitation, has become a core ethical responsibility for sport organizations. With the growing influence of artificial intelligence, large language models are increasingly being used to interpret, summarize, and operationalize policy documents. However, their reasoning fidelity in value-laden contexts such as athlete protection remains uncertain. Methods: This document-based comparative study evaluated two advanced large language models, ChatGPT (OpenAI) and NotebookLM (Google), against twenty-five decision points derived from the 2024 International Olympic Committee Consensus on Interpersonal Violence and Safeguarding in Sport. Responses were independently scored by two evaluators across four dimensions: accuracy, ethical reasoning, applicability, and responsibility awareness. Quantitative and qualitative analyses assessed concordance with IOC recommendations, inter-rater reliability, and reasoning patterns. Structured PICO-style prompts were used to ensure standardized and comparable model interrogation across all decision points. Results: ChatGPT achieved a higher mean composite score (3.84 ± 0.41) than NotebookLM (3.32 ± 0.77) (Wilcoxon Z = 3.41, p = 0.001, r = 0.68; n = 25 paired decision points), with full concordance in 21/25 (84.0%) versus 14/25 (56.0%) decision points. Inter-rater agreement was excellent (Cohen’s kappa = 0.89). ChatGPT demonstrated stronger procedural reasoning and clearer alignment with IOC-defined responsibilities, while NotebookLM emphasized empathy and cultural nuance. Both models underperformed in survivor-centred reasoning and outcome evaluation metrics. Conclusion: LLMs can partially reproduce the ethical and procedural logic embedded in sport safeguarding frameworks, but their reasoning remains incomplete without human oversight. Their integration into governance, education, and policy translation should follow clear ethical guardrails to prevent misinterpretation or amplification of harm. These systems may support policy drafting, education, and communication, yet they should function strictly as augmentative tools rather than independent decision-makers. To our knowledge, this is the first study to systematically benchmark LLM reasoning against an official international safeguarding framework in sport. Future research should evaluate interactive and multilingual applications to enhance contextual accuracy and equitable safeguarding outcomes.
AtıfScopus1WoS1
İncele - MakaleAçık Erişim2026Scopus SJRQ3From black box to pedagogical partner: Students’ sense-making of LLM-based automated assessment
Kara, Abdulkadir; Yıldırım, Serkan
While current educational research on automated assessment heavily emphasizes technical validity, a significant gap remains in understanding students’ sustained, real-world experiences with these systems in authentic learning environments. This study examines students’ experiences with a large language model-based automated assessment system embedded in the regular flow of a university course. The study employed a mixed-methods design with 47 university students over a seven-week period. Quantitative data were obtained from system interaction logs and student feedback ratings, while qualitative data were collected from focus group interviews with 24 students and 175 written feedback responses. The results reveal that students perceive the LLM-based assessment system as a learning assistant, an impartial evaluator, and a self-assessment tool. The transparency of explanations was identified as a decisive factor in building trust in the algorithmic system by helping students understand the rationale behind scores and feedback. Sustained interaction with the system triggered a shift from high-frequency trial and error to more efficient and strategic participation, indicating that the assessment criteria were gradually internalized. The system appeared to create an environment free from perceived social judgment, providing favorable conditions for productive failure, repeated attempts, and self-directed revision. Overall, the study demonstrates that artificial intelligence can be positioned as a tool that scales pedagogical intent without replacing the teacher and can be effectively integrated within the framework of human–AI complementarity in higher education.
- MakaleYayıncıda açık erişim2026Scopus SJRQ2A pilot pre–post study of a scaffolded AI chatbot for financial literacy education among university students
Ayberkin, Doruk; Yirik, Emirhan
This pilot study examines pre- to post-test changes in university students’ financial literacy knowledge following a one-week student-directed learning period using a pedagogically scaffolded GPT‑4o Application Programming Interface (API) chatbot. Guided by constructivist learning theory [29], the immediate feedback model [10], and self-determination theory [24], the study employed a single-group pre-test–post-test quantitative design with 32 undergraduate students at a Turkish public university. The primary outcome was financial literacy knowledge, measured with 32 factual statements rated on a six-point agreement scale and keyed against correct responses, across six subscales (Basic Economics and Finance, Individual Banking, Retirement and Insurance, Financial Statements, Investment, and Tax and Legislation). Five objective multiple-choice items served as a limited, complementary performance check. Paired-samples t-tests were used for normally distributed difference scores, and Wilcoxon signed-rank tests for non-normal difference scores; six subscale comparisons were interpreted using Holm-adjusted p values. The overall score increased from pre-test to post-test, with a large standardized mean change (d = 1.09, 95% CI [0.64, 1.52]), and all six exploratory subscale comparisons remained statistically significant after Holm adjustment. Objective item performance also increased on several items, including the compound-interest item (3.1% to 81.2%). However, because the design included no control or comparison group, identical objective items were repeated, and chatbot interaction logs were not systematically recorded, the observed changes cannot be causally attributed to the chatbot. The findings therefore provide preliminary, uncontrolled evidence that warrants testing in controlled studies rather than evidence of intervention effectiveness.