Is Artificial Intelligence able to Produce Content Appropriate for Education Level? A Review on ChatGPT and Gemini
Yazarlar (1)
Dr. Öğr. Üyesi Mehmet Fatih KARACA
Tokat Gaziosmanpaşa Üniversitesi, Türkiye
| Bildiri Türü | Tebliğ/Bildiri |
| Bildiri Alt Türü | Tam Metin Olarak Yayınlanan Tebliğ (Uluslararası Kongre/Sempozyum) |
| Bildiri Niteliği | Alanında Hakemli Uluslararası Kongre/Sempozyum |
| DOI Numarası | 10.1145/3660853.3660915 |
| Bildiri Dili | İngilizce |
| Kongre Adı | AICCONF '24: Proceedings of the Cognitive Models and Artificial Intelligence Conference |
| Kongre Tarihi | 25-05-2024 / 26-05-2024 |
| Basıldığı Ülke | Türkiye |
| Basıldığı Şehir | İstanbul |
| Bildiri Linki | http://dx.doi.org/10.1145/3660853.3660915 |
| Özet |
| This study examined 120 Turkish stories written for primary, secondary, high school, and undergraduate education levels by ChatGPT-3.5, ChatGPT-4, and Gemini1.5 Pro. The data was processed by software created with Natural Language Processing methods in mind. The general characteristics, quantitative characteristics, and readability of the stories were all reviewed within the scope of the study. Using the Ateşman and Bezirci-Yllmaz formulas, which are widely used for Turkish texts, the readability of the stories was calculated. As a result of the analysis, it was determined that AI is able to produce distinct stories, and it generates cohesive stories by utilizing subjects and themes that are suitable for a specific educational audience. When the average readabilities are taken into account, it has been found that ChatGPT-3.5 generates better stories suited for the education level based on the Ateşman formula and Gemini based on the Bezirci-Yllmaz formula, and the difficulty level of ChatGPT-3.5 stories rises in tandem with education level in both formulas. Also, the stories at the undergraduate level were found to be the hardest to read and with primary schools having the easiest readability. When the number of stories at readability levels is taken into account, it has been found that GPT-3.5 and ChatGPT-4 in the Ateşman formula and ChatGPT-3.5, ChatGPT-4, and Gemini in the Bezirci-Yllmaz formula generate appropriate stories for the education level; the levels range from easy to difficult as the education level increases. Additionally, it has been found that the number of stories included is gradually increasing. It was concluded that AIs produced stories above their educational level; while ChatGPT-3.5 and Gemini were more successful in story production, the Bezirci-Yllmaz formula was better in determining readability. |
| Anahtar Kelimeler |
| Artificial Intelligence | ChatGPT | Gemini | Natural Language Processing | Readability |
BM Sürdürülebilir Kalkınma Amaçları
| Atıf Sayıları | |
| SCOPUS | 7 |
| Google Scholar | 10 |