ChatGPT is a conversational agent built on a large language model. Trained on a significant portion of human output, ChatGPT can mimic people to a degree. As such, we need to consider what social identities ChatGPT simulates (or can be designed to simulate). In this study, we explored the case of identity simulation through Japanese first-person pronouns, which are tightly connected to social identities in intersectional ways, i.e., intersectional pronouns. We conducted a controlled online experiment where people from two regions in Japan (Kanto and Kinki) witnessed interactions with ChatGPT using ten sets of first-person pronouns. We discovered that pronouns alone can evoke perceptions of social identities in ChatGPT at the intersections of gender, age, region, and formality, with caveats. This work highlights the importance of pronoun use for social identity simulation, provides a language-based methodology for culturally-sensitive persona development, and advances the potential of intersectional identities in intelligent agents.
Dark patterns and deceptive designs (DPs) are user interface elements that trick people into taking actions that benefit the purveyor. Such designs are widely deployed, with special varieties found in certain nations like Japan that can be traced to global power hierarchies and the local socio-linguistic context of use. In this breaking work, we report on the first user study involving Japanese people (n=30) experiencing a mock shopping website injected with simulated DPs. We found that Alphabet Soup and Misleading Reference Pricing were the most deceptive and least noticeable. Social Proofs, Sneaking in Items, and Untranslation were the least deceptive but Untranslation prevented most from cancelling their account. Mood significantly worsened after experiencing the website. We contribute the first empirical findings on a Japanese consumer base alongside a scalable approach to evaluating user attitudes, perceptions, and behaviours towards DPs in an interactive context. We urge for more human participant research and ideally collaborations with industry to assess real designs in the wild.
ChatGPT, a chatbot powered by a large language model (LLM), is used by millions globally. Whether it is perceived as an anthropomorphic agent is currently being mapped out. Notably, the degree and nature of how ChatGPT is gendered remains unclear, despite agent gender cues being known to impact user attitudes and behaviors. ChatGPT, with its natural communication style and minimal gender cues, may offer radical insights on agent gendering. In this online mixed methods study, we invited first-time and experienced users to undertake tasks with ChatGPT. We surveyed their perspectives at two time-points following the interaction. Results indicated that ChatGPT’s communicative abilities correlated with gendered traits and its perceived capabilities reflected gender stereotypes. Our work suggests that ChatGPT is not gender-neutral, with implicit gender stereotyping occurring in the absence of gender cues and regardless of exposure. We advocate for conscientious design interventions to overcome agent-based perpetuation of human stereotypes.