by JT Mudge
Today, voice is not just an emerging technology but a huge part of our lives. Back in 2017, when our family first got an Alexa, my then five year old would mumble to it and be understood about 80% of the time. Now she is eight and it is not much different. In fact, it may be worse. Why? One reason is she is using it for more sophisticated tasks.
Alexa processes more of her requests, but a larger percentage do not work. Instead of just asking for jokes, she now has Alexa in her room and she uses it to control her lights, set alarms, help her with spelling, and play obnoxious music. Repeatedly. Seriously, there is only so many times an adult can hear “Cat Flushing a Toilet” on repeat. Over and over. And over.
The other thing my daughter likes to do is ask Alexa questions. Lots of questions. She does this all day long and is by far the biggest user of Alexa in our house. And it makes sense too. Reading and writing are skills that take much longer to develop while speech comes more naturally. We continue to develop the subtleties of conversation as we grow.
In a lot of ways Alexa and my daughter are growing up together and learning the finer points of communication. But both of them have a long way to go. Voice assistants need to not only be smart, but they need to communicate with us better. They need to make us feel confident that they hear and understand us. My daughter has long ago surpassed Alexa’s conversational skills but there are three areas of communication where they are both still growing: trust, personalization, and personality.
While communication is at the heart of what voice assistants are about, trust is crucial if we are going to adopt voice as a vital part of our lives. When someone says one thing and does another, that breaks our trust with them, and the meaning of their words change. We cannot take them at their face value.
My daughter has a Philips Hue light called “candle” she uses as a night light. Like all of our lights, sometimes Alexa does not get it quite right:
“Alexa, dim candle”
The candle dims
Alexa: “Candle is not responding”
“Alexa, turn off candle”
Alexa: “There is no light with the name candle”
“Alexa, turn off candle”
The light goes off.
These are only a couple of the many examples of trust with voice assistants. I am not sure why Alexa has a hard time with truth, but the point is we cannot trust her to do things correctly. In order to gain our trust, voice assistants need to learn to communicate with more finesse.
A better response would be “Hmm, I am having issues talking to Candle. Did the light dim as expected?” It may still be frustrating, but the level of trust increases.
We do not communicate in the same way to everyone. Even as an infant my daughter would cry differently in her mother’s arms as opposed to those of a stranger. Communication tends to work better when we adjust to our audience.
Voice assistants need to go well beyond personal preferences and recommended suggestions. They need to anticipate us, know our habits, and understand our intentions.
This is not an easy problem to solve. First, voice assistants need to distinguish who their audience is for any given request. Then they also need to learn about that audience and continue to adjust their responses.
Who, When, and Where
Take music for example, one of the most common uses for voice assistants. If I ask Alexa to play music, first she will need to know who it is, me or my daughter (I do not want to hear “Cat Flushing a Toilet” by mistake). Next, Alexa needs to learn about my music preferences. When I wake up I like more mellow music. When I am working late I like hair metal. When I am working in the basement I like oldies. By knowing the when and where, Alexa should be able to process a more personalized response.
Then there is one of the biggest issues with personalization, privacy. As humans we make privacy mistakes all of the time: spoilers, gossip, letting secrets slip. Machines usually only know what to share based on very explicit access rules. This is much more difficult when it comes to shared devices like Alexa.
Christmas shopping with my daughter used to be very difficult. She got so excited that she could not keep the secret. As soon as we would walk in the door she would shout “We got you a book!”. Personalized ads on the home computer shout the same thing when they recommend the same gifts I am buying. The difference is now my daughter has learned when to say something and when to be quiet.
It comes back to trust. If we cannot trust our voice assistants, then we cannot expect a decent personalized experience.
Amazon’s Machine Learning team is working on improving personalization. Like my daughter, I expect that Alexa will get much better at this, but there is still some growing up to do.
How well we communicate is as much to do with personalization (knowing your audience) as personality (who each of us is). It is so important that we have many different models to classify our personalities. Are you a Libra? Is your Myers–Briggs type ESTJ? Are you a Type A?
What consumer voice assistants currently lack is a unique personality. Alexa will respond the same for all consumers. Everyone experiences the same basic Alexa.
As much as we love Alexa, her personality doesn’t quite blend with our household. Our family likes corny jokes, sarcastic answers, and short responses to direct questions. We say “please” and “thank you”. We shout through the house (even though we shouldn’t). We get along well together because our personalities work well together.
Before we can have a deeper relationship with our voice assistants, they need to have a personality that works well with ours. When she turns on the light and I say “thanks”, I would like her to say “you bet!” or “de nada”. If Alexa wants to be larger part of our lives, she will need to feel more like a part of our family.
Some will not want a deeper relationship with their voice assistants, and that is the point. Alexa’s personality needs to be personalized. Just as parents influence their children, Alexa should learn to grow and respond to adapt to its environment.
Are We There Yet
The world we experience every day is becoming “smarter” as part of the Internet of Everything. It’s not just Alexa or voice assistants but homes, public spaces, transportation, workplaces - they all are going through a rapid digital transformation and we are still learning what that means.
As someone with the privilege of helping companies solve these problems I’ve learned it’s not just about building a technical product, but also how we feel about it. Our world is getting smarter but it does not have to lose its personality. Every day, by working with some of the most talented people I know, I get to help shape our future and pass that future on to my daughter (who will probably still be listening to cats flushing toilets.)