Artificial intelligence (AI) has become an increasingly integrated aspect of daily life. Particularly, businesses have been incorporating AI into many of their features from customer support to product personalization. While there has been a body of research exploring the interpersonal impacts of AI and human participants, there is limited research on the effects of human-like AI on its influence in the purchasing process. Taking the theoretical framework of the Computer Are Social Actors (CASA) paradigm, I utilized a 2 x 3 experiment to measure if the perceived sex of an artificial intelligence impacts consumers' purchase intention, trust, and sense of agency. Participants interacted with either a male-gendered or female-gendered AI chatbot, or a static website, which then recommended a water bottle based on the participants' preferences. The study indicated significance with both male and female participants preferring the control website over both AI sexes. The study also indicated significance in women participants feeling more overall agency than men participants during the experiment. The results indicate a potential need for a new level of human realism before CASA can be framed within some AI applications. / MACOM / Artificial intelligence (AI) has become an increasingly integrated aspect of daily life. Particularly, businesses have been incorporating AI into many of their features from customer support to product personalization. While there has been a body of research exploring the interpersonal impacts of AI and human participants, there is limited research on the effects of human-like AI on its influence in the purchasing process. Taking the concept that people naturally treat human-like technology as real humans, I have evaluated whether consumers' purchase intention, trust, and sense of agency are impacted while interacting with either a male-gendered or female-gendered AI chatbot , or a static website, which then recommended a water bottle based on the participants' preferences. The study showed that both male and female participants preferred the control website over both AI sexes, and women participants overall felt more agency than men participants during the experiment. The results indicate that there may need to be a new level of human realism before users will treat some AI applications as fellow humans.

Previous research has found that users' interactions with others in online environments are often guided by the same rules and stereotypes we apply in our everyday lives. However, fewer studies have used virtual worlds as an experimental setting for the systematic examination of how avatar appearance and offline identity affect the outcome of users' actual interactions. This online field experiment measured the effect of avatar attractiveness, avatar gender, purported user gender, and favor size on the rate at which users received help across 2,300 separate user interactions. In addition, the main study's avatar gender, purported user gender, and favor size manipulations were replicated with a human avatar condition with 761 participants to examine whether trends for these factors' effects were similar with human avatars. In the main study, attractive avatars generally received more help than less attractive avatars. However, purported female users were helped less frequently than purported male users when represented by avatars that were either male or less attractive. Trends in the human avatar condition were similar to those observed in the main study. Implications for avatar-mediated communication and the persistence of sex roles in virtual environments are discussed. / Master of Arts

The Computers are Social Actors (CASA) paradigm was proposed more than two decades ago to understand humans’ interaction with computer technologies. Today, as emerging media technologies including social robots and smartphones become more personal and persuasive, questions of how users respond to them socially, what individual factors leverage the relationship, and what constitutes the social influence of these technologies need to be addressed. As an expansion of the CASA paradigm, the Media are Social Actors (MASA) paradigm was applied in the current dissertation to understand users’ social perception, social attitudes, and social behavior in their interactions with humanoid social robots and smartphones. Two lab experiments with between-subjects factorial design were conducted. A total of 110 participants were asked to interact with a humanoid social robot and a smartphone respectively in a socio-emotional context and a task-oriented context. Four pairs of social cues were compared to understand their influence on users’ anthropomorphism of the technologies. Multivariate analyses and textual analyses were conducted. Results suggested that users developed more trust in the social robot with a human voice than with a synthetic voice. Users also developed more intimacy and more interest in the social robot when the robot was paired with humanlike gestures. However, individual differences such as users’ attitudes toward robots, robot use experiences, and suspension of disbelief affected users’ psychological responses to the social robot. Although users’ responses to the smartphone did not vary based on the language styles and the modalities, factors such as individuals’ intensive smartphone use, mobile use habits, and their source orientation and re-orientation moderated the social influence of the smartphone. The dissertation has theoretical value in expanding the CASA paradigm to social robots and smartphones. It also tests the validity of the propositions of the MASA paradigm. The results can lead to more comprehensive, nuanced, and exciting discoveries of the social implications, ethical implications, and practical guides of using these emerging media technologies in the future. / Media & Communication