Psychology of Technology
Sensory Marketing (Voice)
Title: A Threat or a Promise?: Essays on Consumer Perceptions of Machines and Recommendation Agents in the Digital Marketplace
Co-chairs: Samuel Bond and Michael Lowe
Committee: Aradhna Krishna (U of Michigan), Christian Hildebrand (U of St. Gallen), and Yegyu Han (IE Business School)
“Personality Perceptions of Consumer Smart Agents”
with Samuel Bond
(Essay 1, invited for revision at Journal of Consumer Research)
This project develops a psychometric instrument that captures perceptions of smart agent personality and demonstrate its utility for addressing important, managerially-relevant questions regarding consumer interactions with machines.
Across a series of studies, we develop a hierarchical model of smart agent personality that contains two high-level factors (“friendly” and “reliable”) and seven underlying facets. We demonstrate the reliability and validity of the measurement instrument using various methods – testing discriminant validity and differentiation between known-groups, and conducting experimental examinations of nomological validity.
“Vocal Similarity, Trust, and Persuasion in Consumer-Recommendation Agent Interactions”
with Michael Lowe and Aradhna Krishna
(Essay 2, invited for revision at Journal of Marketing Research)
This project focuses on how vocal characteristics in consumer and recommendation agent interaction drives persuasion and trust. We utilize unsupervised machine learning method to capture "objective" vocal similarity in timbre (with MFCCs) between an individual user and an agent, and examine its influence on consumer choice as well as the role of trust as a mediator of the underlying process across three experimental studies.
Then, using data from 2,791 Kickstarter campaigns, we show that an average spokesperson’s voice (i.e., the average MFCC scores from a large sample of sampled voices) results in higher persuasion (fundraising) than more atypical voices, again driven by vocal similarity. Our methods and findings provide a deeper understanding of consumer and recommendation agent interactions, including new tools for vocal analytics.
“Can YOU Be Funny?:
How Humor Impacts Our Pleasure of Consumption”
with Michael Lowe
(Essay 3, data collection in progress)
Marketers and designers of artificial intelligence products imbue humanlike characteristics to their products to promote natural and efficient interactions with its consumers. However, little is known regarding the extent to which when and how humanness characteristics of a machine is perceived to be appropriate. Drawing from previous research on humor and expectation violation, I propose that adding humor to consumer-machine interactions may play a role, and even be beneficial, in situations when machines sound or behave to be robotic or somewhat “unnatural”.
Specifically, the usage of humor in AI may moderate the relationship of AI humanness (versus machinelike) on enjoyment of AI-created content and persuasion, and I posit that an appropriate humor in AI may help to mitigate from any negative responses of expectation-disconfirmation.
Other Ongoing Projects
“Voice of Leadership: The Role of Vocal Characteristics as Indicators of CEO Promotion”
with Michael Lowe and Aradhna Krishna
(manuscript in progress, target: Psychological Science)
This research explores the role that vocal characteristics (richness of the voice as depicted through voice timbre) play on perceived leadership capabilities. We present a systematic analysis of how voice may affect career path, utilizing a set of collected voice data of Fortune 500 CEOs, including clips of interview audio to extract vocal features. We quantify an objective measure of vocal timbre (or voice richness) using MFCCs to capture detailed data regarding patterns of overtones in CEO voices.
An examination of these overtones as quantified by the MFCCs reveals that, all else equal, individuals with deeper voices and richer overtones in their mid-vocal frequencies are more quickly promoted to leadership roles within their company. In addition to the model, we provide further evidence in the lab that rich vocal timbres are related to perceptions of leadership potential, and may play a role in actual promotion decisions.
“Consumer Power and Chatbot Humanness”
with Yuhosua Ryoo and Yongjun Sung
(manuscript in progress, target: Journal of Consumer Psychology)
While firms are widely adopting “humanlike” AI agents into their designs, its behavioral consequences on consumer-brand relationships is largely underexplored. In this research, we are interested in how AI’s humanness can alleviate its potential negative consequences in the context of service failure.
Across four studies, we find that consumers feel increased perceptions of power when interacting with a humanlike (versus machinelike) chatbot. In turn, this increased power perceptions lead to higher likelihood to forgive the chatbot (and the brand) and self-brand connection can help mitigate the negative effects on consumers.