Extending CASA for DIY Bot Gender: Effects of Customization on Gender Stereotyping of Social Chatbots

• In Event: 41021 - Student Forum Refereed Roundtables
  In Refereed Roundtable (90 minute): Roundtable 13: Technology and the Evolution of Artificial Intelligence

Abstract

Recent advances in generative AI technologies have significantly boosted the popularity of social chatbots, deepening their societal implications while challenging conventional patterns of HMC (human-machine communication). Notably, the opportunity to customize chatbots through DIY ("do it yourself") bot traits offers new possibilities for personalized interactions in HMC. While existing studies, under the CASA (Computers as Social Actors) paradigm, show that gender cues can lead people to ascribe gender-biased traits to chatbots, it is less well-understood how gender customization in social chatbots may moderate this gender affect and influence user experiences. We explore this gap in research with a 2x2 between-subjects design with two chatbot gender conditions (pre-assigned and customized) and two bot genders (female and male). Our findings show that chatbot gender customization is not simply a functional design choice but reshapes stereotypical patterns in HMC, suggesting its potential to promote equitable human-bot relationships by mitigating social biases.

Author

• Mengmeng Wu, university of chicago