Robot Anthropomorphism and Customer Adoption

Product/Service InnovationBlut, M.; Wang, C.; Wuenderlich, N.; Brock, C. · 2021Journal of the Academy of Marketing Science
Topicsservice robots·anthropomorphism·chatbots·technology acceptance·meta-analysis·uncanny valley·intention to use

A bank keeps employees in the service encounters that depend on personal connection. Now it is about to approve the rollout budget for a new service robot. Before it does, it must decide how humanlike the robot should be.

Customers say they are more willing to use a service robot the more humanlike it seems. That link runs partly through customers seeing the robot as lifelike, smart, safe, useful, and easy to use.

Data chart

Seeing a robot as humanlike goes with intention to use, not with eeriness or rapport

Average correlation across studies between how humanlike customers judged a service robot and each measure shown, corrected for measurement error (0 = no link, 1 = perfect link, below 0 = an opposite link).

.35Stated intention to use the robot.00Negative feelings toward the robot, suchas eeriness (no reliable link)-.07Personal connection with the robot, orrapport (no reliable link)
  • Neither negative feelings nor rapport rises reliably with humanlikeness.

Customers' intention to use a service robot rises with how humanlike it seems, while their negative feelings toward it and their personal connection with it show no reliable link.

Average correlations pooled across service-robot studies and corrected for measurement error; intention to use pools 30 study results from 2,160 customers. A separate pooled analysis finds a small negative link with rapport.

Data chart

A humanlike design matters less for cleaning, storage, education, and entertainment services

Average correlation between how humanlike customers judged a service robot and how willing they said they were to use it, for each service type against all other services, corrected for measurement error (0 = no link, 1 = perfect link).

This service typeAll other servicesServices on customers' belongings(cleaning, freight transport,storage, laundry).17.36Education, entertainment, and newsservices.28.42
  • The link is roughly half as strong for customers' belongings.
  • The link is also weaker for education, entertainment, and news.

The link between how humanlike a robot seems and intention to use is about half as strong for services on customers' belongings as for other services, and also weaker for education, entertainment, and news services.

Average correlations pooled across service-robot studies and corrected for measurement error. The belongings estimate rests on 4 study results from 132 customers. Both gaps also appear when the robot, service, and study differences are tested together, the education, entertainment, and news gap only at a looser statistical threshold.

Action guide

  1. Ask for customer evidence before approving the rollout.Require the proposal to report intention to use and whether customers see the robot as lifelike, smart, safe, likable, useful, and easy to use.
  2. Test actual use before committing the rollout budget.Have your team report actual customer use in the planned service.
  3. Prioritize humanlike design outside cleaning, storage, education, and entertainment services.Its link with intention to use is weaker there, about half as strong for cleaning, freight transport, and storage as for other services.
  4. Give your robot emotional expression and humanlike features such as a face.These cues are most strongly linked to customers seeing a robot as humanlike. Size, cuteness, and an animal-like body make no reliable difference to the link with intention to use.
  5. Compare robot gender cues before setting a design standard.The link is stronger for female-gendered robots, so confirm the pattern in your own service first.
  6. Keep employees alongside the robot where personal connection matters.A more humanlike robot is not reliably linked to more rapport, so do not count on it to build that connection.

Evidence

  • Across studies, the link between how humanlike a robot seems and intention to use is strong, positive, and linear, averaging .35 on a scale where zero means no link. Humanlikeness has no reliable link with negative feelings such as eeriness, so there is little support for an uncanny-valley penalty.
  • Seeing a robot as humanlike goes with seeing it as lifelike, smart, safe, useful, and easy to use. Each set of perceptions partly explains the link with intention to use, but a direct positive link remains.
  • Humanlikeness has no reliable link with rapport, a customer's personal connection with the robot. A separate pooled analysis finds a small negative link.
  • The link is about half as strong for services on customers' belongings, such as cleaning and freight transport, as for other services (.17 versus .36). It is also weaker for education, entertainment, and news services (.28 versus .42). Both gaps come from comparing studies, not from tests within one company.
  • Across studies, the link is stronger for female-gendered robots than for male robots or robots with no clear gender cues (.52 versus .31).

Key takeaway

The more humanlike a service robot seems, the more willing customers say they are to use it.

Limitations

The studies cover stated intention to use service robots, chatbots, and other service AI, not actual use, sales, willingness to pay, or repeated interactions. The link is stronger for critical services, such as ticket selling, and for banking, but only when service types are compared one at a time.

Source

Blut, M., Wang, C., Wuenderlich, N., & Brock, C. (2021). Understanding anthropomorphism in service provision: a meta-analysis of physical robots, chatbots, and other AI. Journal of the Academy of Marketing Science, 49, 632–658. https://doi.org/10.1007/s11747-020-00762-y. Exact values: correlations corrected for measurement error are .35 for intention to use, .00 for negative feelings, and −.07 for rapport; .52 for female-gendered robots versus .31 for male robots or robots with no clear gender cues; .17 for services on customers' belongings versus .36 for all other services; and .28 for education, entertainment, and news services versus .42 for all other services.

https://doi.org/10.1007/s11747-020-00762-y