Evaluating the Usability of an RF–LSTM Learning Platform with Explainable AI among University Students

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Educational AI platforms require usable interfaces that enable students to interpret complex predictive and explainability outputs. This study evaluated the usability of an RF-LSTM learning platform incorporating explainable artificial intelligence among university students. A multidimensional usability evaluation involved 40 Informatics Engineering and Computer Science students who completed six structured tasks covering data exploration, preprocessing, model prediction, performance comparison, SHAP interpretation, and navigation. Usability evidence was obtained from task performance, structured observations, the System Usability Scale (SUS), and open-ended feedback. Task completion ranged from 85% to 100%, while the mean SUS score of 76.7 indicated positive overall usability. Greater difficulties emerged when students compared performance metrics and interpreted SHAP visualizations, particularly regarding technical terminology, visual presentation, navigation, and system feedback. The findings suggest that algorithmic explainability does not automatically ensure comprehensibility during student interaction. Contextual explanations, clearer navigation cues, scalable SHAP visualizations, and processing feedback emerged as design priorities, although the findings remain limited to students within a higher education context.Platform pendidikan berbasis AI memerlukan antarmuka yang mudah digunakan agar mahasiswa mampu menginterpretasikan keluaran prediktif dan explainability yang kompleks. Penelitian ini mengevaluasi usabilitas platform pembelajaran RF-LSTM yang mengintegrasikan explainable artificial intelligence pada mahasiswa. Evaluasi usabilitas multidimensional melibatkan 40 mahasiswa Teknik Informatika dan Ilmu Komputer yang menyelesaikan enam tugas terstruktur meliputi eksplorasi data, preprocessing, prediksi model, perbandingan performa, interpretasi SHAP, dan navigasi. Bukti usabilitas diperoleh melalui performa tugas, observasi terstruktur, System Usability Scale (SUS), dan umpan balik terbuka. Tingkat penyelesaian tugas berkisar antara 85% hingga 100%, sedangkan rerata skor SUS sebesar 76,7 menunjukkan usabilitas keseluruhan yang positif. Kesulitan lebih terlihat ketika mahasiswa membandingkan metrik performa dan menginterpretasikan visualisasi SHAP, terutama terkait terminologi teknis, penyajian visual, navigasi, dan umpan balik sistem. Temuan menunjukkan bahwa explainability algoritmik tidak secara otomatis menjamin keterpahaman dalam interaksi mahasiswa. Penjelasan kontekstual, penanda navigasi yang lebih jelas, visualisasi SHAP yang dapat disesuaikan, dan umpan balik pemrosesan menjadi prioritas pengembangan, meskipun temuan terbatas pada mahasiswa dalam konteks perguruan tinggi.