Abstract
This review article synthesizes ten pivotal publications from 2025 and 2026 to evaluate the evolution of human-centric Artificial Intelligence agents within software engineering. It identifies a transition from reactive auto-complete tools to proactive Multi-Agent Systems and explores the “orchestration” paradigm shift where developers manage autonomous agents. Central to this evaluation is the tension between productivity gains highlighted by the Human-in-the-Loop LLM-based Agent framework’s 82% plan approval rate and the cognitive disruption of the developer’s flow-state. By analyzing the Decentralized Planning Decentralized Execution mechanism, the study establishes that while proactive agents such as “CodeAct” (achieving a 79.3% success rate on the HumanEvalFix benchmark) offer significant efficiency, their real-world integration is hampered by “cognitive lag.” The review further addresses technical deskilling in junior developers, contrasting the high-level intent of “vibe coding” with the need for underlying logical awareness. It argues for AI-based verification and Social Transparency to maintain trust in semi-automated environments. Strategic recommendations include architectural patterns such as context awareness and progressive complexity disclosure to mitigate workflow friction. Furthermore, the evaluation critiques the gap between single-file research prototypes and enterprise-scale multi-file interdependencies, advocating for sandboxed execution environments such as OpenHands. The paper concludes that agentic systems expand rather than replace human capabilities, provided that multi-modal verification frameworks and social transparency protocols are prioritized to ensure accountability. This synthesis provides a roadmap for achieving Hybrid Intelligence through standards of reliability and human-agent synergy.
Cite
MLA
Baig, Hamd Bin Fayyaz, and Fayyaz Ahmad. “Human Centric AI Agents for Software Development: A Review of Emerging Paradigms and Empirical Efficacy.” Journal of Secondary and Undergraduate Research, vol. 4, no. 1, 2026
APA
Baig, H. B. F., & Ahmad, F. (2026). Human Centric AI Agents for Software Development: A Review of Emerging Paradigms and Empirical Efficacy. Journal of Secondary and Undergraduate Research, 4(1)
Chicago
Baig, Hamd Bin Fayyaz. “Human Centric AI Agents for Software Development: A Review of Emerging Paradigms and Empirical Efficacy.” Journal of Secondary and Undergraduate Research 4, no. 1 (2026).