A Literature Review of New Direction in Implicit and Explicit Stereotypes Researches

Abstract

Implicit stereotypes are first noticed in the 1980s, nearly half century later then explicit ones. Since then, an array of indirect methods such as Implicit Association Test (IAT) and Attention Misattribution Procedure (AMP) were invented to further analyze this subtle subject. Recently, with the development of Internet and AI, researchers have generally combined these emerging techniques with conventional study on explicit and implicit stereotypes. The current review mainly introduces three new direction of stereotype study and analyzes a basket of representative and comprehensive researches, as examples, to propose the main strengths and weaknesses of these techniques. First, stereotype tests via Internet have strengths of large samples, improved accuracy, and raising public awareness, but they suffered from non-random recruitment and unstable judgment. Second, applying word embedding to stereotype study has strengths of improved efficiency and comprehensiveness, as well as tracing back history, but it suffered from limitation as book texts and limited samples. Third, utilizing Large Language Models (LLMs) to stereotype study is yet a new direction with little published researches, but the current review cautions that the divergence between human thoughts and LLMs’ outputs may matter. In conclusion, the current review postulated that these three directions have a promising future and can deeply inspire further study.