Over the past decade, a crisis of confidence in published scientific findings has catalyzed widespread response from the research community, particularly in the West. These responses have included policy discussions and changes to existing practice as well as computational infrastructure to support and evaluate research. Our work studies Indian researchers' awareness, perceptions, and challenges around research integrity. We explore opportunities for Artificial Intelligence (AI)-powered tools to evaluate reproducibility and replicability, centering cultural perspectives. We discuss requirements for such tools, including signals within papers and metadata to be included, and system hybridity (fully-AI vs. collaborative human-AI). We draw upon 19 semi-structured interviews and 72 follow-up surveys with researchers at universities throughout India. Our findings highlight the need for computational tools to contextualize confidence in published research. In particular, researchers prefer approaches that enable human-AI collaboration. Additionally, our findings emphasize the shortcomings of current incentive structures for publication, funding, and promotion.

翻译：过去十年间，已发表科学发现的可信度危机引发了学术界的广泛回应，尤其在西方学界。这些回应涵盖政策讨论、现有实践变革，以及支持与评估研究的计算基础设施建设。本研究聚焦印度科研人员对科研诚信的认知、态度及面临的挑战，从文化视角探讨人工智能驱动工具在评估可重复性与可复制性方面的潜力。我们论述了此类工具应具备的特性，包括论文内部信号与需纳入的元数据，以及系统混合性（全AI模式与人类-AI协作模式）。基于对印度各地大学研究人员的19次半结构化访谈及72项跟踪调查，研究结果表明：计算工具有助于语境化评估已发表研究的可信度。尤其值得关注的是，研究人员更倾向采用人类-AI协同方案。此外，研究结果凸显了当前发表体系、经费资助与职称晋升激励机制的缺陷。