Human-Centered AI: Designing Collaborative Intelligence for Decision-Making

Abstract

The field of human-centered AI (HCAI) emphasizes designing AI systems that enhance, rather than replace, human decision-making by centering human values, usability, and trust. This paper examines collaborative intelligence—where humans and AI jointly perform cognitive tasks—with an emphasis on decision support across domains. The methodology includes a comprehensive literature review from human-computer interaction (HCI) and AI, threat and bias modeling, design taxonomy analysis, and evaluation of human-in-the-loop (HITL) systems, including interactive tools like Coevo. Key findings indicate that successful Human-AI collaboration hinges on explainability, accountability, and ethical design principles. Systems like Coevo demonstrate co-creative interfaces that align human and AI reasoning. Human-in-the-loop models improve performance and decision trust, though cognitive biases (like algorithm aversion) may reduce effectiveness unless mitigated. Trust grows when AI is interpretable, not opaque. However, scaling HITL systems faces challenges in cost, workload, and complex feedback loops. We propose a design workflow: start with task analysis, co-design interfaces, integrate explainable AI components, conduct user testing, deploy human-in-the-loop mechanisms for mitigation, and iterate based on feedback. Advantages include improved decision quality, user acceptance, and error reduction; disadvantages involve increased complexity and slower responses. The conclusion underscores that collaborative intelligence represents a path toward trustworthy, effective AI-enabled decision-making. Future research should explore bias-aware collaboration frameworks, dynamic human-AI role adaptation, and scalable HITL mechanisms across diverse domains.