This dissertation is guided by an overarching commitment to advancing cutting-edge, AI-powered, personalized learning environments that scaffold writing development in scalable, explainable, ethical, and pedagogically informed ways. Through this agenda, and as part of a broader design-based research (DBR) program, two interconnected studies were conducted to establish the foundation and guide the subsequent design and development trajectory. The first study presents a systematic synthesis of 18 empirical studies published between 2014 and 2023, synthesizing research on the use of Automated Writing Evaluation (AWE) systems in adult English as a Foreign Language (EFL) argumentative writing context. The review offers a comprehensive review of the field, mapping the intellectual landscape of research and illuminating how AWE tools support or constrain the development of argumentative writing. It further articulates evidence-based design principles and propositions to inform the development of pedagogically grounded, theoretically sound, and contextually responsive AWE systems that meet the rhetorical demands of argumentative writing. Building on insights from the first study, the second study developed and validated two cutting-edge, AI-powered automated essay scoring approaches for evaluating argumentative writing. Seven state-of-the-art transformer-based architectures were trained using the PERSUADE 2.0 dataset, and four Generative AI (GenAI) scoring models were constructed using few-shot prompt engineering. The findings revealed that transformer-based models aligned closely with human ratings, with several architectures (e.g., DeBERTa base model: QWK ≈ 0.84; MAE ≈ 0.44) achieving almost perfect agreement (Landis & Koch, 1977), both within the PERSUADE 2.0 context and when extended to EFL settings involving different learner populations (DeBERTa base: QWK ≈ 0.75; MAE ≈ 0.49). GenAI-based models showed comparatively weaker performance but emerging potential (e.g., LLaMa 3: QWK ≈ 0.69; MAE ≈ 0.61), reflecting substantial agreement but indicating the need for intensive refinement and optimization. The validity discourse, however, must extend beyond statistical considerations to foreground the ethical dimensions, ensuring that scoring processes uphold fairness, transparency, explainability, interpretability, and construct relevance. Together, the two studies establish a coherent foundation for developing next-generation, personalized, AI-powered writing support systems grounded in validated scoring strategies and pedagogically informed design principles at scale, laying the groundwork for the next DBR phases.