Skeuomorphism In Conversational Design
How metaphors impact conversational experiences
Bret Brautigam wrote a thought provoking article on how adding a human voice to a computer could be a form of skeuomorphism. This idea made me wonder… Is any form of conversational design Skeuomorph to some extent?
Before we start, let’s address the fat turkey in the room: what the hell is Skeuomorphism?
Simply put, Skeuomorphism is a fancy design term based on the simple idea of creating elements that resemble their real-world counterparts (switches etc). The first example that comes to mind is the old Apple Calendar UI:
You might wonder, how could anyone design such a hideous interface? Ok first of all, respect your elders! Keep in mind that there was a time when screens and touch interfaces weren’t second nature. At the time, using real world analogies could actually prove helpful to convey how an interface worked.
But eventually, all good things come to an end. As technology and people became more and more inseparable, flat design emerged.
All of a sudden, Skeuomorphism wasn’t cool anymore. It became an embellishment. A boogeyman. A fat turkey we hid in our backyard, so no one would see it.
Will the same thing happen to Skeuomorphism in conversational interfaces?
Once technology merges with our bodies, sure, I can see it happen. But until then, we’re probably stuck babbling to our devices. So let’s look at how Skeuomorphism sneaks into conversational experiences and how it impacts the way we design them.
1. Language
Human Computer Interaction has never been about graphical user interfaces. At its core, it’s always been about communication.
One of the main reasons why conversational interfaces are so fascinating is because conversation is a form of communication everyone uses and understands.
For instance, when users have a hard time navigating a graphical user interface, they often get frustrated and blame themselves (after all, there is a reason every single user test starts with we’re testing the interface, not you).
On the other hand, when users encounter problems in a conversational interface, they immediately blame the system instead. Users are confident in their abilities to articulate themselves and therefore expect the system to adapt to them, not the other way around.
Conversational interaction fundamentally changes users’ expectations but is often still looked at as a technical challenge, rather than a social one as well.
When we talk to someone, we constantly adapt our voice, body language, and vocabulary depending on who we’re talking to. Context shapes language. And it’s how skeuomorphic elements sneak into digital experiences too.
Skeuomorphism in conversations
Metaphors can be a powerful way to convey how an interface works. But we know from traditional UI design that they always come with a lot of baggage:
Digital metaphors carry assumptions users have about its real world counterpart
For instance, when you think about trash cans, you think about filling and emptying them. But what about conversations?
Conversation, social norms, and habits are inextricably linked. As a result, those norms and habits sneak into digital conversations as skeuomorphic elements.
For instance, why are we often tempted to say thank you, even though we know we’re talking to a system, not a human. Or why do we actually use natural language to begin with?
Natural language is a choice. Most conversational bots can easily parse both “Alarm 8am” and “Hey! Wake me up tomorrow morning at 8” but the use of natural language often feels less cumbersome.
We’re so used to speaking naturally that robotic and abbreviated language ironically often result in more cognitive load than just speaking the way we would with a friend.
The conversation is in many ways not a metaphor anymore. It’s become a lie that’s telling the truth.
In order to design better conversations, we need to better understand how these invisible forces affect our conversations and how we can use them in a beneficial way.
2. Time
Time is one of the most elusive and under-appreciated elements of design. Instead of thinking of it as just another tool in our design belts, we often think of it as a burden.
Two years ago, I wrote about how time perception shapes user experience, and why slower experiences have its own unique benefits. Fast forward to today and I’ve come to believe, that conversational interfaces amplify these findings even more.
Let’s see why.
Last year, Intercom published 7 principles of good bot design. The first principle resonated with me the most:
Don’t pretend to be human.
In the article, the author however suggests that “typing-indicators” and artificial waits should be avoided at all cost.
But why?
One of the things I learned designing the UX Bear and making my website conversational is that timing and delays can actually improve the experience.
Delays are in many ways a skeuomorphic element in conversational interaction. Most bots could answer much faster than any human ever could. But does that necessarily mean it’s better?
I don’t think so.
Time can be used as a form of mental whitespace — a means to cut users some slack and give them the time they need to follow along and process what’s happening.
Let’s remind ourselves that a conversation always involves two parties. It’s not a one-way street. Just because you can post three chat bubbles at once, doesn’t necessarily mean you should. It’s in many ways uncooperative, spammy, and even worse, impolite.
Traditional information and visual design taught us how chunking can greatly increase people’s short term memory and help remember things better:
The same applies to pauses. Pauses help separating one idea from another. In a medium without body language and non-verbal cues, every small detail can help communicate an idea more effectively.
When it comes to conversational design, time and pace are just as crucial as the use of comas and periods in written language. We can’t get rid of it. But we can take advantage of it. And you know what?
I think we should.
3. Emotions & personality
Personality and emotions can be thought of as a form of Skeuomorphism.
People are generally well aware that digital assistants don’t have feelings and yet they prefer responses that feel warm, witty, and human, rather than cold, static, and robotic.
Designers have long debated the importance of personality in user interfaces. Fonts, color, copywriting, illustration, and motion are all means to make experiences feel more personal.
The same applies to conversational interfaces. But the means we use to convey personality are less visual, but based on the use of language itself. Consider the following responses, which are not just technical and robotic, but downright frustrating.
Any person interested in reproduction will avoid language like this. Better:
“Sorry, I don’t know how to do that just yet.”
“Oh oh. Sorry something went wrong here.”
Let’s be clear, both error messages are not great since they don’t help users understand and recover from it. But notice how we automatically adapt language for a conversational context. Not even the geekiest kid in town, would write responses as seen in the system dialog in a conversational interface.
So why do we add personality to interfaces? Why do we not just let computers be what they truly are?
Because our interactions with them are inherently social.
Ever cursed at your phone, computer, or even your toaster? It’s you being social to devices you know don’t understand you. And yet there you are, babbling and cursing at random objects around you.
We attribute social traits to many things we probably shouldn’t — computers being one of them.
People like friendly computers
Let’s be honest, you like people who are friendly and have similar interests. It’s ok, we all do. But who would have imagined that it’s the same for computers?
Research shows that people prefer computers who are friendly and similar to them. When participants in a study were asked to complete a set of tasks, they preferred the friendly computer over the other. When computers praised the participants, even in an unwarranted and almost random way, they didn’t just feel better about themselves, they also felt better about the computer they were using [0].
Perhaps even more surprising, participants rated the friendly computer’s performance better than the other’s. Even though personality might be a skeuomorphic element, it impacts the way people think about both the system, and the interaction itself.
My friend, Peter Hodgson, who is a voice interface designer on the same team I’m on keeps saying:
Personality is not a choice. If you don’t create a persona for your bot, your users will.
Let’s make sure we’re in charge of that.
4. Voice
Voice is to some extent skeuomorphic too. A digital assistant’s voice could sound completely robotic. It could speak in an unnatural, short, and abbreviated way. But it doesn’t.
Companies invest a lot of time and effort in making voices sound less robotic and more human for a good reason.
It’s not just easier to listen to a human voice, it’s more pleasant too! Sure, command line and robotic voices have their own unique style but for longer speech interactions they quickly become irritating and annoying. It also wouldn’t help you feel less awkward talking to your phone in public, wouldn’t it?
Voice is part of the persona
I’ve recently listened to a podcast about the voice actress of Siri and it felt like listening to my phone being interviewed by someone else. It was incredibly bizarre.
Voice is a part of the persona and shapes its identity. Once we’ve associated a voice with something, it becomes part of its identity.
Stephen Hawking sticks to his robotic voice even though there are way more natural sounding voice synthesizers available today. He’s not keeping his voice because it’s delightful, but because it’s become part of who he is. On a recent appearance at a show where celebrities auditioned to give him a new voice, he mentioned:
I don’t think anyone would take me f — king seriously if I sounded like that,
Just like in real life, a voice comes with a lot of prejudices, expectations, and biases. And it affects the way users experience conversations.
Studies have shown that male voices are believed to be more credible for technical topics like math or navigation, while female voices are believed to be more truthful for emotional topics like love and relationships [1] .
How does this impact the way we design for voice? Does that mean I should stick to a male voice for my conversational traffic-advice bot?
No.
We have a social responsibility to change these biases. Deliberately taking advantage of stereotypes and biases only further reinforces them. Let’s not do that.
But it is important to be aware that voice does impact how people feel and interpret the words they hear and that it’s yet another way how the metaphor impacts our design—whether we want that or not.
Parting words
During my research on how design affects decision-making, I realized that neutral design is merely an illusion. Every single decision we make throughout the entire process, impacts the end user in one way or the other.
It’s the same in conversational design.
No matter how you design a conversation, people’s experience will be influenced not just by the conversational experience itself, but by all the expectations they have about it as well. It’s our job to make those expectations help shape the experience in a way that’s beneficial for the user, not the other way around.
Going back to the initial question of this article: are conversational interfaces skeuomorphic?
Perhaps they are. But the far more interesting question is… how can we use it to improve the experience?
Thanks for reading and Bert Brautigam great thought on the topic and Tibor Kranjc’s help with the illustrations!
If this article resonated with you, feel free to hit ♥︎. Thanks!
Further reading:
Studies by Clifford Nass:
[0] http://ldt.stanford.edu/~ejbailey/02_FALL/ED_147X/Readings/nass-JOSI.pdf
[1] http://onlinelibrary.wiley.com/doi/10.1111/j.1559-1816.1997.tb00275.x/abstract
