The rapid growth of AI-powered voice assistants (VAs) like Siri and Alexa presents a challenge for companies seeking to enhance product competitiveness. This study addresses the need for a better understanding of voice-based AI interaction, focusing on the impact of language style in different service contexts. While existing research explores various factors influencing VA adoption, the role of language style—a direct carrier of VA-human interaction—remains under-investigated. The authors highlight the potential for different impacts from concrete (precise, data-driven) versus abstract (general, impressionistic) language, drawing on consumer research showing preferences for concrete language in sales contexts but abstract language in word-of-mouth communication. The study also considers the distinction between utilitarian-dominant (accuracy, efficiency) and hedonic-dominant (anthropomorphism, enjoyment) service contexts, suggesting that user preferences for language style may vary across these contexts. The central research question is whether matching language style (concrete or abstract) to service context (utilitarian or hedonic) improves user evaluation and continuous usage intention of VAs. The authors hypothesize a congruency effect, proposing that concrete language will be preferred in utilitarian contexts, while abstract language will be preferred in hedonic contexts. Processing fluency is posited as a mediating variable.

The literature review examines existing research on VA utilization, highlighting the importance of utilitarian benefits, symbolic benefits, and social benefits in driving user adoption, while acknowledging perceived privacy risks as a disincentive. The review then focuses on language style in AI-human interaction, noting the significance of warmth and competence as fundamental dimensions of interpersonal judgment. It discusses how warmth is often associated with abstract language and high levels of empathy, while competence is connected to concrete language and precise information. The review explores the construal-level theory, showing how matching the construal level (abstract vs. concrete) of a message to the psychological distance of the receiver can enhance message effectiveness. Previous studies examining abstract versus concrete language in consumer behavior are reviewed, showing that abstract language is more effective in positive word-of-mouth but that concrete language can increase consumer satisfaction and purchase intention in certain situations. Finally, the review examines the combined influence of language style and service context (hedonic vs. utilitarian) in shaping user preferences for AI interactions, suggesting the possibility of a congruency effect.

The study employed a mixed-methods approach involving a pilot study and three main studies. The pilot study (n=240) used a survey to validate the classification of four service contexts (music recommendation, movie recommendation, online shopping, financial investment) as either hedonic-dominant or utilitarian-dominant. Study 1 (n=380) used an online experiment to test the hypothesized congruency effect of language style (abstract vs. concrete) and service context (hedonic-dominant vs. utilitarian-dominant) on continuous usage intention, focusing on music recommendations and online shopping as representative scenarios. Participants were exposed to simulated VA conversations with either abstract or concrete language and then rated their continuous usage intention. Study 2 (n=348) replicated Study 1 with different scenarios (movie recommendation and financial investment) and additionally measured processing fluency as a mediating variable. Study 3 (n=161) employed a field experiment using actual VA videos with robotic characteristics to minimize anthropomorphism effects. Participants interacted with the VAs in the chosen scenarios and rated their continuous usage intention and processing fluency. The data was analyzed using ANOVAs and moderated mediation analyses with PROCESS macro (Model 8, Hayes, 2018) to test the hypothesized relationships and the mediating role of processing fluency.

The pilot study confirmed the classification of service contexts as hedonic-dominant or utilitarian-dominant. Study 1 revealed a significant interaction effect between language style and service context on continuous usage intention: concrete language led to higher continuous usage intention in utilitarian contexts (online shopping), while abstract language led to higher intention in hedonic contexts (music recommendations). Study 2 replicated this congruency effect with new scenarios (movie recommendations and financial investment) and demonstrated the mediating role of processing fluency. Higher processing fluency was observed when language style matched the service context. Study 3, using actual VA videos, confirmed the congruency effect and the mediating role of processing fluency. The findings consistently showed that matching language style to the service context resulted in higher processing fluency and increased continuous usage intention. Alternative explanations, such as perceived response accuracy and usefulness, were ruled out.

The findings support the hypothesized congruency effect between language style and service context in influencing continuous usage intention for AI voice assistants. The results consistently demonstrated that users preferred concrete language in utilitarian contexts and abstract language in hedonic contexts. The mediating role of processing fluency suggests that the ease of processing information contributes significantly to the positive evaluation of VAs. These findings extend existing research on AI-human interaction by demonstrating the importance of considering both language style and service context in designing effective VA interactions. The study’s contribution lies in offering actionable insights for developers and managers to optimize VA language systems, tailoring their responses to the specific context of use, thereby enhancing user experience and continuous usage.

This research provides substantial evidence for a congruency effect between language style and service context in influencing user experience with AI voice assistants. Users prefer concrete language for utilitarian tasks and abstract language for hedonic tasks, with processing fluency mediating this effect. Future research could explore other language styles, cultural variations, and the effects of physical VAs versus virtual assistants to further refine the understanding of AI-human interaction.

The study primarily used simulated interactions rather than real-world interactions with physical VAs, which might have limited the sense of immersion and realism for participants. The sample predominantly consisted of Chinese participants, limiting the generalizability to other cultures and languages. Further research with diverse populations and settings is needed to strengthen the external validity of these findings.