Since 2012, social product development (SPD) has emerged as a potential paradigm for new product development (NPD). However, the emerging SPD practices are difficult to remain and promote after their early success. The lack of sustainable human-centric collaborative settings will hinder the release of crowd-based collective intelligence in NPD. Currently, the core mechanisms of human-centric collaboration (HCC) in SPD are not clear. To narrow the gap between theory and practice, a two-way empirical study is conducted through an in-deep industrial practice investigation and theory-oriented research based on cooperative inquiry. An extreme case of a famous automotive company is sampled as the object of the cooperative inquiry. The proposed theoretical models are verified by two real SPD pilot projects. Inspired by the discipline of Collaborative Engineering, this study explores the HCC issues and it core mechanisms in the context of SPD. Two novel foundation theoretical models are developed to address the focused scientific problem and engineering challenge. A bottom-up model is developed to structurally elicit the HCC requirements and develop reasonable information modules for specific HCC work practices. A top-down model is developed to improve the manageability and reusability of the learnable experience from historical HCC work practices for better business-technology-system alignment. This study put forward novel theoretical models for designing and deploying collaborative work practices for HCC in SPD. A classification-based Descriptive Model of Collaboration (DMoC) is proposed, which is domain-agnostic and customized for SPD requirements. Our model inherits and outperforms in modelling capability than five other proven modelling approaches.