Title:Towards an Atlas of Cultural Commonsense for Machine Reasoning

Abstract:Existing commonsense reasoning datasets for AI and NLP tasks fail to address an important aspect of human life: cultural differences. We introduce an approach that extends prior work on crowdsourcing commonsense knowledge by incorporating differences in knowledge that are attributable to cultural or national groups. We demonstrate the technique by collecting commonsense knowledge that surrounds six fairly universal rituals -- birth, coming-of-age, marriage, funerals, new year, and birthdays -- across two national groups: the United States and India. Our study expands the different types of relationships identified by existing work in the field of commonsense reasoning for commonplace events, and uses these new types to gather information that distinguish the identity of the groups providing the knowledge. It also moves us a step closer towards building a machine that doesn't assume a rigid framework of universal (and likely Western-biased) commonsense knowledge, but rather has the ability to reason in a contextually and culturally sensitive way. Our hope is that cultural knowledge of this sort will lead to more human-like performance in NLP tasks such as question answering (QA) and text understanding and generation.