Analisis Komparatif Performa Large Language Model dalam Perancangan User Interface Berbasis Framework CO-STAR
DOI:
https://doi.org/10.47111/jointecoms.v6i2.25659Keywords:
Large Language Model, Prompt Engineering, CO-STAR, User Interface, Masyarakat AwamAbstract
Penelitian ini mengeksplorasi strategi perancangan antarmuka pengguna (user interface ) yang imajinasi bagi masyarakat awam dengan tingkat literasi digital terbatas melalui pemanfaatan kecerdasan buatan. Fokus utama penelitian ini adalah menganalisis bagaimana kualitas proses dalam prompt engineering mempengaruhi efektivitas Large Language Model (LLM) dalam menghasilkan solusi desain. Peneliti mengusulkan penggunaan framework CO-STAR sebagai struktur instruksi formal untuk meningkatkan kejelasan dan relevansi output AI. Melalui eksperimen komparatif terhadap enam model LLM, hasil desain dievaluasi oleh 129 responden menggunakan skala Likert dan analisis statistik Mean Rank . Temuan menunjukkan bahwa kerangka kerja CO-STAR secara signifikan membantu AI dalam menghubungkan elemen visual yang kompleks menjadi tata letak yang mudah dipahami oleh pengguna pemula. Studi ini menyimpulkan bahwa rekayasa prompt yang presisi mampu mengatasi limitasi sistem tokenisasi AI, sekaligus memberikan kontribusi praktis dalam menciptakan produk digital yang mendukung inklusivitas digital bagi seluruh lapisan masyarakat.
Downloads
References
[1] A. Srivastava et al., “Beyond the Imitation Game: Quantifying and Visualizing the Capabilities of Language Models,” Journal of Artificial Intelligence Research (JAIR), vol. 75, pp. 123-145, 2023.
[2] G. S. Madaan et al., “Language Models of Code are Few-Shot Commonsense Learners,” arXiv preprint arXiv:2210.07128, Oct. 2022.
[3] I. Medhi, A. Prasad, and K. Toyama, “Optimal Usage of Non-Textual Cues in Interfaces for Illiterate Users,” in Proc. SIGCHI Conference on Human Factors in Computing Systems (CHI '07), 2007, pp. 727-735.
[4] I. Medhi-Thies et al., “Voice-facilitated mobile services for low-literate users,” in Proc. 17th International Conference on Human-Computer Interaction with Mobile Devices and Services (MobileHCI '15), 2015, pp. 547-554.
[5] J. White, Q. Fu, S. Hays, and M. Sandborn, “A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPT,” arXiv preprint arXiv:2302.11382, Feb. 2023.
[6] L. Reynolds and K. McDonell, “Prompt Programming for Large Language Models: Beyond the Few-Shot Paradigm,” in Proc. 2021 CHI Conference on Human Factors in Computing Systems, 2021, Art. no. 314, pp. 1-14.
[7] N. Dwivedi et al., “Opinion Paper: ‘So what if AI can write?’ – Imperatives of the next generation of AI research,” International Journal of Information Management, vol. 71, p. 102646, Aug. 2023.
[8] N. J. Schick and H. Schütze, “Toolformer: Language Models Can Teach Themselves to Use Tools,” arXiv preprint arXiv:2302.04761, Feb. 2023.
[9] S. Kurniawan and P. Zaphiris, “Research-Derived Web Design Guidelines for Older Users,” ACM SIGCAPH Computers and the Physically Handicapped, no. 73-74, pp. 129-135, 2002.
[10] T. Brown et al., “Language Models are Few-Shot Learners,” in Advances in Neural Information Processing Systems (NeurIPS), vol. 33, 2020, pp. 1877-1901.
[11] Y. Bang et al., “A Comprehensive Capacity Evaluation of ChatGPT: Robustness, Hallucination, and Interactivity,” arXiv preprint arXiv:2302.04023, Feb. 2023.
[12] Z. Yang et al., “LayoutGPT: Compositional Visual Planning and Generation with Large Language Models,” arXiv preprint arXiv:2305.15393, May 2023.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Journal of Information Technology and Computer Science
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
