Scott McCall
Empathy for the Machine - Design Considerations for Voice
Scott McCall
September 20, 2017

Empathy for the Machine – Design Considerations for Voice

Open on a busy downtown street–it’s evening, cold and drizzling. Our traveler walks into a lush, well-lit hotel lobby after an arduous trip. The modern chic decor is stunning. No doubt the service here will be excellent, she imagines, as her thoughts drift to imminent room service and a plush comforter. She walks up to the desk to check in and sees no one. Meanwhile, our zero UI auto-concierge waits with bated breath under the front desk:

[urgent whispering] “I’m down here–just say the word! I know you! Your previous rooms, your meal preferences, and even that it’s your birthday tomorrow–I have a surprise for you! Just notice me…say the words to invoke my check-in application and I’ll be at your service!”

The weary hotel guest raps her nails on the marble, looks around, rings the bell…her gaze eventually falls to a small branded placard with a bulleted list of written instructions for screenless check-in. She rings the bell again…

Well, that sounds like a modern service disaster of Richard Scarry proportions–the kind where all the trucks on the road collide and there are mustard and maple syrup and hot dogs everywhere. Yes, I am aware that it is unlikely that a prestigious hotel chain would consider replacing a crucially important component in the service chain with anything like a basic Alexa Skill. Regardless of technology, there are sticky design problems that need to be effectively addressed if we want to reach the goal of zero UI bliss— technology that responds to people in ways that feel as if it were alive.

The Design Problems Remain The Same

Obviously, a voice interface may not be the best road to excellent service for this particular hotel service application. Even with a custom voice application, some likely hurdles for this scenario might be:

  • How do we make the user aware that voice check-in is possible, and encouraged?
  • How do we help our traveler get started? What does she need to say?
  • How do we build trust and deliver accurate, top-notch service?
  • How do we address all of the potential questions or problems our traveler (or any traveler, actually) might have? Can we do this?
  • How do we sense frustration and turn the interaction around?

From a human-centered design perspective, we would start from the perspective of the traveler. We would empathize, explore and prioritize different possibilities in the traveler’s journey, then iterate and solve the problems. However, screenless solutions excel and come to life when we can “humanize” the interaction, meeting users on their terms.

To address these additional complexities, I’m proposing an additional layer of consideration to help put an interaction in perspective: empathize with the machine. Treat the machine like the other side of a conversational interaction. Looking at the problem from both the user’s and the machine’s perspectives can help you better understand the opportunities and gaps.

So, just how might we go about empathizing with the machine? Assuming you have identified a problem worth solving for users and your business, and you believe a conversational interface might be a good avenue to explore, consider the following:

Play to Voice Interaction Strengths

It’s not magic, it’s technology. Well executed, conversational interfaces can be a fantastic way to increase a feeling of intimacy, save time and physical effort, or serve situations where being tied to a screen is inconvenient or less than ideal. But some scenarios are more ideally suited for voice than others.

Voice interfaces can be great for:

  • Hands- or eyes-free environments where safety or screen visibility is a concern
  • Complicated things that people can articulate (for example, “Show me psychological thrillers from the 1990’s, available for free, with a rating of 3 or more stars”)
  • Languages that are hard to type

On the other hand, voice interfaces are generally not good for:

  • Input that is hard to describe
  • Situations where negotiation or lots of variables are required
  • Huge amounts of input or output

adapted from Laura Klein’s Designing from Voice Interfaces

Another consideration is that voice interfaces without screens often lack sufficient visual cues to indicate to users that 1) voice interaction is possible, and 2) here’s how you do it. Often, paper “tooltips” are used but then again, this is getting away from zero UI. Voice interaction standards have yet to solidify across devices, so there’s no clear and obvious way to start up an interaction, yet.

With this in mind, is the scenario you’re considering suited for a voice application? View it from the machine’s perspective, playing to its strengths in order to set it up for success.

Research, Explore, and Test Potential Interactions. Repeat.

Human interaction is nonlinear by nature and the type of input to which a machine needs to respond can vary greatly. In order to make the machine side of a zero UI interaction feel lifelike and human, take the necessary time to research and understand the numerous ways users can provide input. Think through the scenarios, experiment, test with users, and iterate. Do this early and often.

Also, be mindful of the scope of an interaction–how much are we asking of the machine and what does it need to do? As the potential inputs and types of tasks grow exponentially, does the overall experience begin to suffer? If it’s too broad, focus it. Consider the elements that make for a great core interaction and build from there.

Understand Context and Intent

When people talk to one another, so much information is conveyed by nonverbal cues like tone, body language and facial expressions. Using natural language is key to making the interaction feel like a 1:1 conversation, but a surface level human-esque veneer can set up unrealistic expectations if the machine cannot demonstrate an understanding of situational context, build trust via accuracy and consistency, recognize return users, or retain and reference a history of interactions.

When designing for a specific voice application, work to find out what the machine needs to know in order to respond effectively and be truly useful. What data can it access? What critical cues, based on content, inflection or tone could help bring the interaction to the desired level of “human-like” that makes the overall interaction a success?

Of course, machines are still machines. Offer elegant ways out of a jam, providing an ability to sense a problem and elegantly recruit an actual human to intervene.

Keep Empathizing

It’s critical to develop empathy for users when practicing human-centered design, just as it has always been. But when crafting human-like zero UI interactions, strive to gain perspective from both sides, human and machine, in order to create more lifelike experiences.

Interested in chatting more about designing for voice? Let’s chat!


Filed under:

If you like this, you'll like these:

More Blog Posts


With the strength of Globant behind us, we’re powering up our focus on transformative digital solutions.

On July 30, 2018, find all new content at

See More