Introduction

The use of digital Ai generated avatars in business settings is nascent, so it is too early to understand how they will be accepted by people at all, let alone across different demographics. 

After all, chatbots were widely derided before the advent of Large Language Models (LLMs). LLMs increased the quality of chatbots so dramatically that usage not only grew to huge volumes, but the usage itself more closely resembled a normal human-to-human interaction. One researcher on Google’s LLM, LaMDA, even famously believed it to be sentient. 

And a similar process could easily happen to digital people which, due to increased technical capabilities, feel more natural, human, and present.

That said, although the future is unknown, there are studies in this area that give us some clues as to possible adoption behavior.

The Evidence

Research from MIT showed that people feel better represented by avatars that look like them.

“Participants rated their level of satisfaction with those avatars as virtual self-representations and provided the level of perceived resemblance to themselves. The results show that our avatars are perceived to be similar to the self, rated at 7.5/10. Those avatars with faces derived from the participants’ face mixed with an ethnically similar face were also rated with high scores. These results differ significantly from how arbitrary avatars are perceived”

With this in mind, there is a solid argument for using a system like D-ID, that can produce a greater variety of avatars with relative ease, as opposed to using CGI-based systems which tend to be more limited.

In a paper by the EU EMPATHIC project,  whose goal is to develop an empathic virtual coach capable of enhancing seniors’ well-being, they found that “Seniors (independently from their gender) prefer to interact with female humanoid agents…”

They also found that ”seniors with technological experience felt less motivated and judged the proposed agents less captivating, exciting, and appealing”. This suggests that experience with technology is as important as age in determining levels of acceptance. 

Finally, a paper from the Cognitive Science Lab, at the International Institute of Information Technology, Hyderabad, India found that acceptance was situationally based. Which is to say that it depends on the role the avatar was playing. For example:

“The results indicate that the sense of eeriness for highly human-like avatars… was low, with higher acceptance in subordinate job roles.” 

Most of the use cases that D-ID is focused on are subordinates – assistants, service roles, etc… 

Interestingly the paper goes on to say:

“In recent years, [avatars] in highly cognitive roles like psychiatric therapy counselors… have made a mark. The premise is, talking to a virtual agent gives a sense of anonymity and reduces inhibitions [12, 21]. People preferred talking to a virtual realistic-looking agent in these contexts more than an agent that looked cartoonish”

Conclusion

There is evidence that the differences in reaction and satisfaction with digital human service providers between generations tend to follow patterns associated with users’ familiarity and comfort with technology. Generally speaking, younger generations who have grown up with technology tend to be more accepting and comfortable with digital human service providers, while older generations may be more skeptical or prefer human interaction. 

However, these are just general trends and as we have seen, there are large exceptions. Older individuals who are tech-savvy and open to digital human service providers, while some younger individuals may have a preference for human interaction. It’s also important to note that reaction and satisfaction also depend on factors such as the specific application and the individual’s personal experience with the technology.

And of course, the technology itself is changing and this could radically affect the acceptance of digital people across the board. 

After all, it was only a decade ago that watching a video on a mobile device was a bit odd, something that has now become commonplace.

We have not yet seen an avatar that provides service with perfect domain expertise and full emotional consistency. They could well be better perceived than human agents lacking knowledge and with the inevitable emotional inconsistencies that come with being human. 

In other words, digital people who look and sound human, who know as much as the best service agent, and who don’t have the ups and downs mood that humans do, might well be better accepted.

Matthew Kershaw is D-ID’s VP of Commercial Strategy.