Yapay Zekâ Modelleriyle Çocuk Diş Hekimliğinde Endodontik Tedavi Hakkında Ebeveyn Sorularının Değerlendirilmesi


Creative Commons License

Bardakçı E., Karadeniz H. B.

32.İzmir Dişhekimleri Odası Uluslararası Bilimsel Kongre ve Sergisi & 1. Ege Bölgesi Dişhekimleri Odaları Bilimsel Kongre ve Sergisi, İzmir, Türkiye, 20 - 23 Kasım 2025, ss.361-362, (Özet Bildiri)

  • Yayın Türü: Bildiri / Özet Bildiri
  • Basıldığı Şehir: İzmir
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.361-362
  • Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu
  • Atatürk Üniversitesi Adresli: Evet

Özet

Amaç: Yapay zekâ tabanlı sohbet robotları, günümüzde klinik karar destek sistemleri ve bilimsel araştırmaların yanı sıra, çocuk diş hekimliği kliniklerinde ebeveynlerin tedaviyle ilgili sorularını yanıtlamak için de kullanılmaktadır. Bu çalışma, süt dişlerine uygulanan endodontik tedavilere ilişkin ebeveynler tarafından sıkça sorulan sorulara (SSS) Grok 3, Microsoft Copilot ve ChatGPT-5’in verdiği yanıtların doğruluk, kalite ve okunabilirlik düzeylerini karşılaştırmayı amaçlamaktadır.

Gereçler ve Yöntemler: Ebeveynlerin sıkça sorduğu sorular (SSS), süt dişi endodontik tedavileri olan pulpektomi ve pulpotomi konularından seçilmiştir. Google arama motorunda ilgili anahtar kelimelerle yapılan İngilizce aramalar sonucunda elde edilen 20 web sitesinden, her tedavi için 10’ar soru seçilerek toplam 20 soru oluşturulmuştur. Sorular Grok 3, Microsoft Copilot ve ChatGPT-5 adlı üç yapay zekâ tabanlı sohbet robotuna yöneltilmiştir. Elde edilen yanıtların doğruluğu 5 puanlık Likert ölçeğiyle, kalitesi Global Kalite Skalası (GQS) ile ve okunabilirliği Flesch Kolay Okunabilirlik Skoru (FRES) yöntemiyle değerlendirilmiştir. Veriler IBM SPSS Statistics 27 yazılımında analiz edilmiş, anlamlılık düzeyi p<0,05 olarak kabul edilmiştir.

Bulgular: Likert ölçeğine göre yapılan değerlendirmelerde, üç yapay zekâ tabanlı sohbet robotunun yanıtlarının doğruluk puanları arasında istatistiksel olarak anlamlı bir fark saptanmamıştır (p>0,05). Tüm modellerin yanıtları genel olarak “çoğunlukla doğru” ile “tamamen doğru” aralığında değerlendirilmiştir. Benzer şekilde, GQS puanları açısından da anlamlı bir fark bulunmamış (p>0.05) ve yanıtların genel kalite düzeylerinin birbirine yakın olduğu belirlenmiştir. Buna karşın, okunabilirlik açısından yapılan analizlerde Microsoft Copilot’un diğer gruplara oranla daha kolay okunabilir olduğu saptanmıştır (p<0,05).

Sonuç: Yapay zekâ tabanlı sohbet robotları ebeveynlere doğru ve güvenilir bilgi sunma potansiyeline sahiptir; ancak bu bilgilerin anlaşılır biçimde iletilmesi, kullanıcıya sağlanan bilginin niteliğini güçlendirecektir.

Aim: Artificial intelligence-based chatbots are currently used not only in clinical decision support systems and scientific research but also in pediatric dentistry clinics to address parents’ questions regarding treatment procedures. This study aims to compare the accuracy, quality, and readability of responses provided by Grok 3, Microsoft Copilot, and ChatGPT-5 to frequently asked questions (FAQs) posed by parents concerning endodontic treatments of primary teeth.

Materials and Methods: Twenty FAQs on pulpectomy and pulpotomy in primary teeth, frequently asked by parents, were selected from 20 (10 per treatment) websites based on English-language Google searches. The questions were posed to three AI-based chatbots: Grok 3, Microsoft Copilot, and ChatGPT-5. The accuracy of the obtained responses was evaluated using a 5-point Likert scale, their quality was assessed with the Global Quality Scale (GQS), and readability was determined using the Flesch Reading Ease Score (FRES). The data were analyzed using IBM SPSS Statistics 27 software, with a significance level set at p < 0.05.

Results: According to the evaluations based on the Likert scale, no statistically significant differences were observed in the accuracy scores of the responses provided by the three AIbased chatbots (p > 0.05). Overall, the responses from all models were rated within the range of "mostly correct" to "completely correct. Similarly, no significant differences were found in terms of GQS scores (p > 0.05), and the overall quality levels of the responses were observed to be comparable across the models. In contrast, analysis of readability revealed that Microsoft Copilot responses were significantly easier to read compared to those of the other models (p < 0.05).

Conclusion: AI-based chatbots have the potential to provide parents with accurate and reliable information; however, presenting this information in a clear and comprehensible manner will enhance the quality of the knowledge delivered to the user.