A Crowd-AI Collaborative Duo Relational Graph Learning Framework towards Social Impact Aware Photo Classification

In artificial intelligence (AI), negative social impact (NSI) represents the effect on society as a result of mistakes conducted by AI agents. While photo classification problem has been widely studied in community, NSI made misclassification is largely ignored due to lack quantitative measurements and effective approaches reduce it. this paper, we focus an NSI-aware where goal develop novel crowd-AI collaborative learning framework that leverages online crowd workers quantitatively estimate effectively misclassified photos. Our motivated limitations current either 1) cannot accurately because they simply model semantic difference between true categories or 2) require costly human annotations pairwise class categories. To address such limitations, SocialCrowd, crowdsourcing-based explicitly reduces designing duo relational graph with estimated workers. The evaluation results two large-scale image datasets show SocialCrowd not only but also improves accuracy both datasets.