Anthropomorphic AI assistants promise the best interaction experience. But only because it’s technically possible to create digital humans and make them work online 24/7 doesn’t mean we should.
UneeQ’s Sophie is a flesh-and-bones digital human. Source
The promise of human-like interaction with computers
We need to admit that the idea behind digital humans is fascinating. It has already been introduced on UX Collective with anticipation of how they will revolutionise human-machine interaction and improve UX.
Technology has reached the point where it has become possible to arrange an interaction with a machine that resembles the conversation with a real person on a level that was not achievable before. It is like a dream to create a living man artificially come true (think of all the Frankensteins and Pinokios). But digital humans are not a miracle. We have all the puzzles to make it work: dialogues are powered by ChatGPT while 3D face animations are done through NVIDIA Audio2Face technology.
As their producers and promoters advertise, digital humans will enrich interactions with a “human touch”: build emotional connections with users, leading to higher conversion, loyalty, and engagement. And if you are lost on the website because of its poor usability, you can reach out to the digital human for guidance.
That sounds like a solution to all the problems that UX designers are trying to solve, which, in essence, all originate from one fundamental problem: how do we make the interaction between the human and machine natural and easy? To put it in other words, how can technology be made more human? If you take this problem literally, it looks like the answer is simple — make the computer speak and look like a human with little AI, and it is done. The temptation to follow this path is high, but let’s not rush, as it can do more harm than good to the user experience. And here are some reasons why we shouldn’t.
Human interaction is a work
Interaction with a digital human evokes the feeling of social presence. Since Eliza chatbot, researchers proved that we perceive intelligent computers as social entities. That makes applying the existing knowledge on social interactions to human-computer interactions justified.
Erving Goffman, an American sociologist who studied social interactions, coined a term that describes the particular kind of effort people put into daily contact with others. This term is “face-work.” When we interact, we play a better version of ourselves; we manage our faces to make a desirable impression on others. Only when we are in a private space, with no judging eyes around, we can relax from this social performance.
Now imagine you are at home lying on the couch in your track pants and an oversized T-shirt, shopping online on your phone. On one of the pages, the digital human appears. This virtual agent is as perfect as it is supposed to be — the embodiment of the brand values and customer-centricity. “How can I help you today?”—they ask with encoded care. Fortunately, there is a familiar X icon that you can tap, and the digital human is gone. What a relief — you stay in your private space without judging digital gaze.
Humans are rather shy. That’s why XIX-century retail boosted when price tags were attached to products and displayed in shop windows. Before that, it required some courage to enter the store and start a conversation with the salesman. E-commerce takes this idea even further: you can buy from your couch with no social interactions at all.
If you claim to be a human, you’d better be one
But even if some users are self-confident enough to have no issues starting an interaction with a digital stranger, this conversation could quickly become frustrating.
That is due to an overestimation of anthropomorphic AI’s abilities. If it looks human and speaks like a human, we will treat it as if it were a real person. But no matter how smart generative AI seems, real conversation is a much more complex process than just an exchange of information.
When we speak to people, we sometimes interrupt them, but we also wait for them to finish their line when they pause to make up their minds. We read the body language and try to adjust our way of communication to the cultural and social profile of the interlocutor.
If digital humans are going to deliver on the promise, they should do all of it as well as we do. Otherwise, we humans will be at least disappointed.
But the experience could be much worse. Digital humans come from the uncanny valley. Apart from the motion inaccuracies that can trigger the feeling of eeriness, there is something deeper that makes us dislike them. They are questioning our identity by mimicking us. We may feel threatened, and we have reasons: they are good-looking, non-ageing, working 24/7, and very low-cost. Are they going to take our jobs? You don’t want your customers to feel uncomfortable, do you?
Why not design AI assistants more like C-3PO?
There is a solution to the shyness, disappointment, fear, and disgust we may experience when interacting with hyper-realistic digital humans. Let’s not design AI assistants this way!
Do you remember this funny golden android called C-3PO from Star Wars? Everybody loves this kind of robot — a bit imperfect, sometimes silly and therefore forgivable (no unrealistic expectations). If it tries to be like a human, then it’s rather cute than scary. It would not judge us, as we are in a position of power here.
C-3PO is an example of a human-friendly AI assistant. Source
Think of all the design possibilities for such a digital character and how it can enhance the brand! Making a non-human character inclusive and neutral should be much easier than a digital human of a specific age, gender, and race. And don’t forget to make it imperfect so your customers can sympathise with it when it makes mistakes or hallucinates a bit.
After all, you can make your AI assistant a text-chat or a voice-bot. It will provide the same value as an animated character but in a much more discrete way.
Apart from the looks, the more important question a designer should ask when integrating an AI bot into a product or service is what interaction problem it solves or what additional value it offers. Wowing your users with a piece of cutting-edge technology is not enough in the long run.
When a digital human is a good choice
However, there are cases when AI-powered digital humans can elevate our experience. Digital humans will be used as NPCs (non-player characters) in video games. Social norms can be suspended in the game, and our identity becomes the subject of experimentation. We take it easier than real-life situations, even if real life happens mainly online.
The uncanny valley effect is also not something that should bother the players, as video games are usually surreal and much more scary things can happen there. We expect it to be weird. We also know that the game is a fake world, so we will not be frustrated by slight communication misalignments.
Digital humans can be a good choice when designing for a context similar to video games or, more generally, for entertainment purposes.
Do you remember the trend of skeuomorphism? All the wooden and leather textures and shiny embossed buttons? A graphic user interface is a visual metaphor of the real world, and some designers decided to copy the real world with the highest accuracy or even make virtual representations better than the original. This trend is over, as users get the metaphor pretty quickly, add all those heavy visual additions become obsolete.
Now, a trend of anthropomorphism is on the horizon. It’s already a dominant way of depicting AI as a cyborg, very commonly as an attractive female, which is not only a transmission of gender biases but also an amplification of the misconceptions about generative AI. The fear of AI that surpasses humans is quite common, and by making it look like a human we are just adding to it.
We should not repeat past mistakes, as the stakes are higher this time. A technological breakthrough is opening a gate for a new user interface paradigm. That is an exciting time to be a UX designer. And its our job to make new technology understandable and accessible through the right metaphors. So, let’s select them wisely.
This article is 100% human-written
Digital humans: do we really need this? was originally published in UX Collective on Medium, where people are continuing the conversation by highlighting and responding to this story.
