Programming learning at an early age effectively fosters logical thinking and self-centeredness, but an appropriate evaluation method for young learners has yet to be established. Herein we propose a new learning evaluation method that incorporates problem constructs. Specifically, we investigate the relationships between changes in block programming complexity, personality traits, and learning effects. Complexity is evaluated using rubric evaluations and log data, while personality traits are based on Big-5. Then the learning effects of 34 kindergarten children participating in workshops are analyzed in terms of complexity and personality traits. After learning, first-time programmers tend to show a large increase in complexity. The correlation with the rubric score is ρ = 0.43, and the correlation with log data is r = 0.92. Furthermore, analysis using the Big-5 gives ρ = 0.609 for the rate of increase in extraversion and complexity, indicating a strong relationship between learning effects and personality traits. In the future, this data will be used to build an AI tools for automatic evaluation, feedback, and learning curriculum recommendations.