Authors
Keywords
Abstract
An automated label identification and recommendation system using deep convolutional neural networks, understanding user perceptions regarding the reliability, accuracy, and interpretability of the computer system is essential for developing robust and explainable AI solutions. This study contributes to this need by exploring several performance dimensions, including accuracy, contextual relevance, clarity, dataset quality, sensitivity, specificity, recall rate, false positive rate, and explain ability. Using a cross-sectional survey design, data were collected from respondents via Google Forms in November-December 2025. One-way ANOVA and two-way ANOVA (p < 0.05) were used with IBM SPSS to explore the relationships between socio-demographic variables and perceived performance metrics. The findings indicate consistent perceptions across experience levels and model types, while the model output format significantly impacted label accuracy and dataset quality. The results emphasize the importance of output representation in improving automated labeling performance and reliability, suggesting directions for future system optimization and validation frameworks.
