Giving human qualities — both emotionally and visual, physically — to virtual assistants could prompt people to reveal more personal information to brands than they otherwise would, according to a multi-phase study led by researchers in computer science and mathematics. It also could ease some privacy data concerns.

Researchers at the University of Waterloo found that people put more trust in virtual assistants when they have human-like qualities, mannerisms and visual features such as age, facial expressions and hairstyles.

The research, Genie In The Bottle: Anthropomorphized Perceptions of Conversational Agents, began in 2018 and ran through September 2019, when the paper was submitted. It was announced Thursday.  

“This anthropomorphism that we see in our study speaks to how people personalize conversational agents,” Edward Lank, a professor in Waterloo’s David R. Cheriton School of Computer Science, wrote in an email to Search Marketing Daily. “I think that – from the perspective of advertising – this work speaks to perception and attachment toward conversational agents.”

Led by Lank, researchers created an environment where 10 men and 10 women with diverse experience interacted with virtual assistants. Three conversational agents — Amazon Alexa, Google Assistant, and Apple Siri — were used. The 20 participants shared their perceptions of the agents’ personalities and were asked to describe them physically before creating an avatar for each. 

When asked whether this study considered consumer data privacy, along with California Consumer Privacy Act (CCPA) and General Data Protection Regulation (GDPR) laws, Lank wrote that the findings do not necessarily answer the question, but “understanding how and why users may trust technologies such as conversational agents allows us to better understand how to train (users), regulate (information), and design (products) in a way that balances empowering users against respect for privacy.”

He also pointed to the tension between a desire to become closer to technology and to use it more pervasively against the inherent risk this poses to privacy.

Lank wrote that the issue becomes increasingly complicated when considering things such as “conversational agents which are, technically, always listening to you, even if just passively until they hear their wake keywords. In many ways, negotiating these tensions motivates a lot of the discussions we have about the risks versus rewards of technology.”

For the study, the perceptions of an agent’s character were structured through behavioral characteristics such as approachability, sentiment, professionalism, intelligence, and individuality. Adding a visual element to these characteristics increased trust and the willingness to share information for those who came into contact with them.

The findings describe Siri’s sentiment as predominantly disingenuous and cunning, while Alexa is genuine and caring. The participants commonly described Alexa’s individuality as neutral and ordinary, while participants considered the individuality of Google — and Siri especially — to be more defined and pronounced.

When asked to physically describe the virtual agents, those participating in the study perceived Alexa to be of average height or slightly shorter and older than the others. Alexa wore casual or business-casual attire, mostly dark or neutral colors. Her hair was described as darker, wavy, and worn down.

Those who participated in the study perceived the Google Assistant as being of average height or taller, dressed in either casual attire with a focus on tech culture or in business-formal clothes in dark or neutral colors. The virtual assistant’s hair was perceived as being light in color, blond or brunette, and either long and straight or worn down or worn up in a bun or ponytail. Study participants also associated Google with higher professionalism. 

Survey participants commonly described Siri as being of average height and younger than the other agents, and rarely wearing glasses. They described Siri as wearing either casual but fashionable clothes such as a V-neck, tank top, and heels, or business-formal style dressed in either dark or particularly bright colors, especially red. The participants described Siri’s hair as short or as long, straight and worn down, either blond or black.

Anastasia Kuzminykh, a PhD candidate in Waterloo’s Faculty of Mathematics who helped to conduct the study, believes the findings provide a window into the way people think — and, unfortunately, shows bias. “How an agent is perceived impacts how it’s accepted and how people interact with it; how much people trust it, how much people talk to it, and the way people talk to it,” she stated.