Study finds ‘uncanny valley’ may influence customer preference in robotic shopping assistants
Holiday shoppers who venture into retail stores this season may find themselves asking robots, not human sales associates, for guidance.
But how should these robots look? And how should they respond? Barry J. Babin, the Phil B. Hardin Professor of Marketing at the University of Mississippi, gives insight into customer preference when it comes to robotic shopping assistants in a new study published in the Journal of Consumer Behavior.
“The research investigates how robotic shopping assistants, or RSAs, should be designed,” said Babin, who also serves as chair of the Department of Marketing, Analytics and Professional Sales. “Two basic questions are addressed: How human should a robot appear? Also, should the robot just try to solve a problem for a customer, or should the robot go beyond task completion and provide entertainment?”
Robots can be made with humanlike appearances and mannerisms, or they can be as simple as a flat-screen panel – even looking like an ATM, Babin said. They can provide a practical benefit, such as directing a customer to a product. They can also provide a personal or emotional benefit, even telling jokes.
Babin and his co-authors surveyed more than 200 customers who have interacted with SoftBank‘s Pepper robot, a humanoid robot launched in Japan almost a decade ago. The researchers conducted follow-up experiments examining customer responses to RSAs that could be controlled in the lab to reflect humanoid and nonhumanoid characteristics paired with practical or more personal attributes.
The results suggest that customers prefer humanoid RSAs with practical attributes over a humanoid RSA that provides entertainment.
“The consistent result across all studies is that humanlike RSAs’ attempt to entertain did not result in favorable outcomes relative to other conditions,” Babin said. “Thus, particularly when a robot is humanlike, it should just be designed to provide utilitarian (practical) benefits.”
Anthropomorphism, or nonhuman objects taking on human characteristics, is a factor in the findings, Babin said.
“Our results suggest that the ‘uncanny valley’ theory is at play,” he said. “Basically, when nonhuman things look or act too human, we react with a sense of creepiness.
“In highly anthropomorphic conditions, the sense of uncanniness can create a negative reaction. Thus, when the robot looks human and tries to joke around, it does not come across well.”
Nobuyuki Fukawa is a co-author of the study and associate professor of marketing at the Missouri University of Science and Technology. The research employs a novel approach that analyzes both unstructured and structured data simultaneously, Fukawa said.
“A side benefit of our method is that the analysis software that we used allowed us to combine closed-ended, structured data, such as ratings, with unstructured data, like the written comments that coincide with a numbered rating,” he said.
“Investigators may struggle to combine those things. Our method could be applied to a wide range of big data research.”
Both Babin and Fukawa agree that the utilization of RSAs will increase and they hope the study will provide insight into how best to design them.
Even Ole Miss‘ implementation of Starship food delivery robots is an example of widespread adoption, Fukawa said. In this case, the RSA is nonhumanoid.
“There is a lot of successful marketing of RSAs, and I have no doubt they will be used more and more,” he said. “There are all kinds of examples, such as cleaning robots that vacuum a store.”
Deploying RSAs has been beneficial for many companies worldwide, Babin said.
“McDonald’s in France – and in some other countries – has replaced counter service with a large touch-screen panel where customers place orders and pay,” Babin said. “They step forward and their food drops into a chute. Chatbots are another illustration.
“The cost advantage will be difficult to ignore in the decision to deploy RSAs.”
By Erin Garrett