Congruity of Language Styles in AI Voice Assistants Across Service Contexts

By Dr Silpaja Chandrasekar, PhDReviewed by Susha Cheriyedath, M.Sc.Jan 19 2024

In a paper published in the journal Humanities and Social Sciences Communications, researchers investigated the unprecedented growth of voice assistants (VAs) equipped with artificial intelligence (AI) and their implications for product competitiveness. The study focused on optimizing the user experience by aligning VAs with utilitarian or hedonic service contexts and utilizing concrete or abstract language accordingly.

The aim was to establish a congruity effect that fosters more favorable evaluations. The findings from three studies highlighted a user preference for VAs employing abstract language in hedonic contexts, while VAs using concrete language were deemed more competitive in practical contexts. Researchers identified the mediating role of the perception of processing fluency, providing valuable insights for managers and technology providers to enhance users' continuous usage intention.

Related Work

The maturation of natural language technologies, particularly in intelligent voice systems like Siri, TmallGenie, and Alexa, has led to widespread commercial use. VA is crucial for service innovation, with significant growth predicted in markets like the United States, Japan, and China. However, research gaps persist in understanding user reactions to VA language.

Congruity Effect in VA Interaction

In three studies, the researchers conducted experiments to investigate the impact of language style and service context on users' evaluation of VAs. The pilot study involved 240 participants assigned to different VA service contexts, such as music recommendation, movie recommendation, online shopping, and financial investment. The results confirmed significant differences in participants' perceptions of utilitarian and hedonic attributes in these contexts. Users considered music and movie recommendations more hedonic while considering online shopping a financially helpful investment.

This classification guided the selection of scenarios for subsequent studies. Study one included 380 participants and employed a 2x2 between-subjects design, manipulating the perception of VA language style (abstract vs. concrete) and service context (hedonic-dominant vs. utilitarian-dominant). The results revealed a significant interaction effect, indicating that the congruity between language style and service context influenced continuous usage intention. Specifically, participants showed higher intention to continue using VAs when the language style matched the context, with concrete language preferred in practical contexts and abstract language preferred in hedonic contexts.

Expanding from study one, study two delves into movie recommendation and financial investment contexts. The study also introduced processing fluency as a potential mediator. The results confirmed the congruity effect, and they found that processing fluency mediated the relationship between language style, service context, and continuous usage intention. In hedonic contexts, users favored the abstract language style, while in practical contexts, they preferred the concrete style.

Study three aimed to validate the findings using an on-site investigation with 161 participants interacting with real VAs. The results replicated the congruity effect, and processing fluency again played a mediating role. The study addressed potential alternative explanations, such as perceived response accuracy and usefulness, confirming that these factors did not mediate the observed effects.

The studies provide robust evidence for the congruity effect in human-VA interaction. Users exhibit a higher intention to continue using VAs when the language style aligns with the utilitarian or hedonic nature of the service context. Processing fluency emerged as a critical mediator, emphasizing the role of cognitive ease in shaping users' perceptions and intentions. The findings contribute valuable insights for designing more effective and user-friendly VAs in different service contexts.

Optimizing AI VA Interactions

Researchers explore the evolving dynamics of AI VAs, initially designed for hedonic purposes but increasingly employed for utilitarian functions. A continuum, rather than a binary structure, characterizes their applicability, revealing a nuanced interplay between language style preferences and service contexts. The study introduces a novel approach, proposing a dynamically changing language strategy to enhance user evaluations, addressing a gap in AI–human interaction research.

The research extends AI–human interaction studies by delving into the impact of hedonic/utilitarian trade-offs on language style preferences and subsequent VA evaluations. Unlike previous studies on specific language styles, this approach introduces distinct construal levels (concrete and abstract), revealing a congruity effect that aligns language styles with service contexts. This theoretical advancement enhances the understanding of AI–human interactions and provides a new language strategy to optimize VA evaluations.

By deconstructing the internal mechanism of the congruity effect, the study emphasizes the role of processing fluency in AI–human interactions. Integrating construal-level theory sheds light on the factors influencing users' perceptions, contributing theoretical insights to the broader field of AI–human interaction research. The findings underscore the importance of aligning language styles with service contexts to improve users' processing fluency and, consequently, their evaluations of AI VAs.

Urge developers and product managers in the rapidly evolving AI landscape to consider VA service contexts during language programming. The research emphasizes the need to recognize users' preferences for language styles based on the dominant attributes of service contexts. This awareness enables dynamic switching of VA language styles, with the suggestion to assess and collect user interaction data through big data analytics and user surveys. Furthermore, the study advocates for the practical importance of developing diverse language packages tailored to different VA usage scenarios, emphasizing the significance of fluency and user comprehension in language design.

Conclusion

In conclusion, this research highlights the evolving role of AI VAs, initially designed for hedonic purposes but increasingly utilized for practical functions. The study introduces a dynamic language strategy, addressing a gap in AI–human interaction research and emphasizing the alignment of language styles with service contexts. The congruity effect, influenced by hedonic/utilitarian trade-offs and distinct construal levels, contributes theoretical insights. Developers and product managers are encouraged to consider VA service contexts during language programming, fostering dynamic language style switching.