The rapid integration of Natural Language Processing (NLP) technologies into educational assessment platforms has fundamentally altered the landscape of automated essay scoring (AES). This comprehensive study investigates the pedagogical implications, accuracy, and student perceptions of AES systems deployed in high school English Language Arts curricula. By analyzing a dataset of over 5,000 student essays graded concurrently by human educators and a proprietary NLP algorithm, the research establishes a high degree of quantitative reliability, showing a 92% variance match between machine and human scores on macro-level criteria such as grammar, syntax, and structural cohesion. However, qualitative analysis reveals significant deficiencies in the algorithm's ability to evaluate creative nuance, rhetorical tone, and culturally specific idioms, often penalizing divergent thinking. Furthermore, student surveys indicate a 40% decrease in subjective writing motivation when feedback is solely generated by an AES system, with learners citing a lack of empathetic engagement and personalized encouragement. To address these multifaceted challenges, the paper proposes a "Human-in-the-Loop" (HITL) instructional framework. This model utilizes NLP to instantly process mechanical and structural errors, thereby freeing human educators to focus their grading time entirely on narrative development, logical argumentation, and conceptual creativity. The authors argue that while AES technology is a powerful tool for reducing the logistical burden of grading, it must be strategically tethered to human oversight to ensure that writing is taught as a communicative art form rather than a merely algorithmic exercise.