Jakarta, INTI - The researchers from the Research Center for Data Science and Information (PRSDI) of the National Research and Innovation Agency (BRIN) conducted a study on validity in artificial intelligence (AI) systems.
Ira Maryati, a researcher at PRSDI BRIN, explained that validity is a fundamental aspect determining the reliability of an AI system in producing outputs that align with user objectives and the context in which it is implemented.
"The success of an AI system is determined not only by the accuracy of the model, but also by the suitability of the data, the context of use, and the model's ability to maintain performance when implemented under different conditions. Therefore, understanding validity is increasingly important to ensure that the technology can be used appropriately and responsibly," said Ira on Wednesday, June 24, 2026.
Ira explained that the current discussion being developed covers a number of important aspects, starting from the conceptual foundations of AI validity, the relevance and representativeness of training data, the validity of the model training and testing process, to the challenges of cross-domain validity that often arise when models are applied to environments different from those for which they were developed.
"Challenges that can affect the validity of AI systems include overfitting, underfitting, and context mismatch. These conditions can cause the model to produce good predictions on training data, but suboptimal results when confronted with real-world data," explained Ira.
Besides technical aspects, the study also discussed the importance of evaluation metrics and validation strategies to ensure alignment between system development objectives and the resulting results. This approach is necessary for AI implementation to provide optimal benefits while minimizing the risk of bias and decision-making errors.
AI Validity in Various Sectors
Ira also reviewed several examples of AI validity failures that have occurred in the real world, including in the healthcare sector, organizational behavior, and automotive systems. These cases demonstrate that inadequate data quality and validation processes can introduce bias and reduce the reliability of AI systems.
Ira also presented the progress of research on perceptions and practices of the use of Generative AI in academic and research activities in Indonesia. The research was conducted through a national self-assessment-based survey involving various respondent groups.
The research team has now completed the development and distribution of the questionnaire, as well as the descriptive analysis. The next stage includes conducting Focus Group Discussions (FGD), analyzing the relationships between variables using the Structural Equation Modeling – Partial Least Squares (SEM-PLS) approach, and compiling a scientific article entitled "Perceptions and Practices of Generative AI Use in Academic and Research Activities: A Self-Assessment-Based National Survey in Indonesia."
Ira hopes the research results will provide a more comprehensive picture of the level of acceptance, utilization patterns, and challenges of Generative AI use in national academic and research environments.
Conclusion
Researchers at PRSDI BRIN studied the importance of validity in artificial intelligence (AI) systems to ensure the technology produces reliable, context-appropriate outputs and can be used responsibly. Ira Maryati explained that AI success is determined not only by model accuracy but also by the quality and representativeness of data, the suitability of the context of use, and the model's ability to maintain performance under different conditions. The research team is also developing research on perceptions and practices of Generative AI use in academic and research environments in Indonesia through a national survey.
Read more: Microsoft Found that Indonesian Workers are in the Frontier Professionals Category in AI Utilization