Teaching Machines to Talk to Humans is Delicate Business
April 25, 2017
In the era of conversational commerce, user experience experts are becoming completely absorbed by talking about how to talk.
We are learning the art of conversation at a whole new level, learning not only how to talk to clients, but also how to make the conversation meaningful. We want the conversation experience to be empowering, within the correct social and global context.
We don’t want to speak in tongues – we want the conversation to roll off the tongue, to be genuine.
We are researching, creating customer journeys, prototyping, and experimenting. We use the design thinking process to try to get to the core of a customer problem or need, to find a solution to customer communication that is not only elegant and frictionless, but also pleasant, natural, and worthy of trust.
Teaching machines how to learn to communicate with us is no longer just science fiction.
“Alexa, open the pod bay doors.”
(Incidentally, if you do this, Alexa will answer “I’m sorry, Dave, I’m afraid I can’t do this, I’m not Hal and we’re not in space!”
By combining voice recognition technology and artificial intelligence, we are seeing that we can get there, that meaningful conversation between machines and humans are beginning to happen, but along the way, we are discovering new challenges that we never thought we would encounter.
These are the challenges we continue to work on:
While well advanced, the technology still needs improvement
Language is complex. While we are way more advanced than ever, there are still limitations on what we can do to make machines understand our languages.
Large companies such as Google, Facebook and Amazon are investing a lot in research and in the developer community to improve AI. Collaborative development will get us there.
Examining the Context of Usage
The line has become skewed between public and private spaces. Do customers want to approach a bank machine using their voice? What about privacy? Is the robot, listening for a command, capable of hearing things it should not hear? It’s not yet natural, in our public spaces, to speak to a computer, and not everyone is comfortable with it. It cannot be forced; we have been typing for a long time to communicate with others. Are people comfortable with AI and voice interfaces always listening? Are they recording? Is there a Cloud connected?
Is voice interface appropriate? Dig Deep.
Just because it’s new, it doesn’t mean we need it. While the technology is useful in many context, sometimes, maybe a graphical user interface would work better, or maybe both voice and graphics. With new options available to designers, design thinking is more important than ever. We really have to dig deep and figure out where the true conversation pains are, to create solutions to real problems, instead of creating new problems.
Can robots encompass the brand personality and build trust?
Trust of App = Trust of brand
As we embark on the journey with new technologies, we are bound to have errors—we’re not going to catch everything. Siri is a good example of a conversational interface that suffered from early adoption, and lack of maturity in the technology. Often, Siri is slightly off the mark. In addition to misunderstandings, the Siri technology just couldn’t follow a complex conversation. Even though you just asked her about a city, she wouldn’t connect the dots when you asked her to find a coffee shop in the next sentence, without you renaming the city. Accuracy and speed are needed for a trust in machines to begin.
Multilingual challenges abound
In Montreal, where I live, a bilingual city in Canada, this is especially challenging. Different accents, street names in another language, expressions that are unique to the environment, cannot be understood. Language is sophisticated, and it will take time for machines to learn how we speak.
Tread carefully as you prepare to have the conversation about conversation.
Teaching machines to talk to humans is delicate business.
