Semarang – A research team from the Department of Informatics, Faculty of Science and Mathematics, Diponegoro University, has successfully developed an artificial intelligence (AI) method capable of detecting user interface (UI) design flaws quickly and efficiently. This innovation is expected to provide a solution that helps application developers deliver digital services that are user-friendly, inclusive, and of high quality.
User interfaces are the main gateway for human interaction with technology, whether in banking applications, e-commerce platforms, transportation services, or public services. Unfortunately, even minor design flaws—such as buttons that are difficult to tap, overlapping text, confusing icons, or poor color contrast—often cause frustration. Manual UI testing, which has been the common practice so far, tends to be time-consuming, costly, and prone to human error.
To address these challenges, the study led by Dr. Aris Puji Widodo, S.Si., M.T., along with his colleagues Dr. Eng. Adi Wibowo, S.Si., M.Kom. and Dr. techn. Kabul Kurniawan, S.Kom., M.Cs., leveraged advances in deep learning technology to automate UI design evaluation. The research, published in the International Journal of Advanced Computer Science and Applications in 2023, introduces a new approach through the application of few-shot learning, or learning from only a handful of examples.
Through this method, the AI model used—EfficientNet-B1—was able to recognize ten categories of common design flaws such as misaligned elements, poor text hierarchy, and low color contrast. With only five examples per category, the model achieved an average accuracy of 76.05 percent on a standard dataset, and 42.6 percent on a specially developed dataset containing various visual flaws. These results are considered highly promising, especially for startups or small development teams with limited access to large-scale data.
According to Dr. Aris, the benefits of this discovery can be directly felt in everyday digital life. Applications will become easier to use and reduce user stress, digital inclusion will be enhanced as designs become friendlier for elderly users and people with disabilities, and digital innovations can be delivered more quickly thanks to more efficient testing processes. For instance, whenever a UI design change is made, the AI system can instantly evaluate and provide recommendations for improvements, ensuring application quality while significantly shortening development time.
Nevertheless, the researchers acknowledge that this study still faces challenges, particularly regarding data diversity and the interpretability of AI results. The team hopes to further expand the dataset, involve end-users in model training, and apply similar methods to other UI components such as navigation and animations. Dr. Aris emphasized that the purpose of this development is not to replace designers or testers, but to position AI as a strategic partner that accelerates and strengthens the software innovation process.
With this innovation, FSM UNDIP once again demonstrates its tangible contribution to advancing artificial intelligence technologies that benefit society at large. This research not only strengthens Diponegoro University’s position as a leading research hub in informatics but also opens opportunities for cross-disciplinary collaboration to produce digital solutions that are increasingly inclusive, adaptive, and impactful in the future.
The news about this research has also been published on Kompas.id and can be accessed through the following link: Detecting Digital Design Flaws with Artificial Intelligence
