ҹ1000

How we fell in love with our voice-activated home assistants

Personal assistants like Amazon’s Alexa already do our bidding. Why do we want them to be our friends too?
Alexa
A good listener, but not a friend
Howard Lipin/San Diego Union-Tribune/TNS/Alamy Live News

Alexa, play some classical music.

Alexa, set an alarm for 7am.

Alexa, will you marry me?

THEY sit unobtrusively in your home and, with a simple command, check the weather, play a song or set a kitchen timer, rendering many of life’s routine tasks hands-free. Voice-activated devices like the Google Home and Amazon Echo – which responds to the name “Alexa” – are designed to make your life easier.

They seem to have struck a chord – . But they aren’t just a convenience. The way we interact with these devices is beginning to betray a more complex relationship between humans and machines.

Daren Gill, director of product management for the Alexa personal assistant used by Amazon’s Echo, says he has been surprised by how often people try to engage the assistant in purely social interaction. “Every day, hundreds of thousands of people say ‘good morning’ to Alexa,” he says. Half a million people have professed their love. More than 250,000 have proposed. You could write these off as jokes, but one of the most popular interactions is “thank you” – which means people are bothering to be polite to a piece of technology.

“The really disruptive part, I think, is the difference between treating something as a tool and treating something as an agent,” says , a social roboticist at Carnegie Mellon University in Pennsylvania. So why are we getting so attached to a jacked-up smart speaker?

Devices have been doing our bidding for years, thanks to voice recognition technology in phone-based assistants like Apple’s Siri and Microsoft’s Cortana. However, subtle differences in the interface design of the new home assistants make us regard them more as characters, says Knight – like a friendly concierge, perhaps, or even a personal secretary (see “The invisible woman“).

For example, unlike Siri or Cortana, you don’t need to press a button or fumble for your phone to invoke Alexa once you’ve set up your Echo. The device is constantly “listening” for its name (Alexa comes as standard, but the user can change the word that wakes it up to “Amazon” or “Echo”). Call, and it will respond to your request, just as you might say a friend’s name to get their attention.

500,000

Number of people who have told Alexa ‘I love you’”

Using speech in this way means the interface “almost disappears”, says Gill. Communicating by speech comes naturally to most of us, which makes talking to Alexa feel a lot more seamless than typing commands into a computer. Thanks to natural language processing and machine learning, it understands your requests without you having to speak robotically or learn specific phrases. “We want people to interact the way they would with friends,” says Gill. The team was inspired by the computer in the TV series Star Trek.

Another design advantage comes simply from where we keep the devices, which are designed for use at home. “They’re in a very intimate space, and that’s going to change the way we think about them,” says Knight. Add the Echo’s unassuming physical design – essentially a cylinder with an LED at the top – and the device quickly becomes a familiar presence. “It almost becomes like a family pet,” says , director of the University College London Interaction Centre.

Trust me

But it’s not a pet. Neither the Echo nor the Google Home look like a robot. They don’t move. They don’t have a face. So why are we so ready to bestow a social role on something that looks so undeniably mechanical?

Social robotics research has shown that we can’t help treating technology like people. It doesn’t have to be fancy – the response is elicited even by relatively primitive kit. In a , Clifford Nass and his team at Stanford University asked people to assess the performance of a desktop computer. They gave more positive feedback if they entered their response into the computer itself rather than into an identical one – just as you may be less critical to someone’s face than behind their back.

And despite their apparent simplicity, the home assistants offer other cues that make us treat them socially. “Sometimes we can be tuned to think about things as having faces even if they don’t have eyes or a mouth, just by things like orientation,” says Knight. The Echo’s LED ring lights up in the direction of your voice when you say its name, for instance, which sends a cue about where the device is focusing its attention. This non-verbal behaviour is very powerful, says Knight. “We really care when people hear us.”

In fact, she says, the home assistant’s minimalist design can be an asset in convincing us to form relationships with them as we are so good at projecting ideas of character onto a blank canvas. Interestingly, humanoid robots have a much harder time living up to our expectations.

Over 100,000 a day

Number of people who say ‘good morning’ to Alexa”

All this raises the question of whether there could be any negative consequences of our willingness to treat these devices like the friends we share our intimate thoughts with. It’s easy to forget when building a rapport with Alexa that all your whispered sweet nothings are being collected and processed by Amazon.

Devices that constantly listen out for commands pose particular concerns if they accidentally pick up audio someone didn’t intend to record, or if the data is used for other purposes, says at the University of California, Berkeley. He encourages people to examine privacy policies and thinks manufacturers should be more transparent about how they use people’s data. Amazon says users can delete any records of their communications with Alexa, both on the app and the server.

250,000

Number of people who have proposed to Alexa”

That said, at the moment, Alexa doesn’t remember much about you. The only way to personalise it – so that it remembers what city to give the weather forecast for, for example – is by feeding that information into an app.

This may explain why most personal questions are likely to elicit its standard deflection: “Sorry, I did not understand.” (Or, if you are lucky, it makes a joke: when I suggested marriage, Alexa replied that she thought we should just be friends.)

But all this could change as home assistants become more personable. The next step for Alexa is a capacity for longer and more dynamic dialogues, says Gill. Amazon also has a “long road map of personalisation”, he says, to allow your device to get to know you better, perhaps learning what books or music you like.

As these machines become more personable, the way we treat them could inform the design of more advanced technology.

Jibo, a personal robot designed for the home and set for release in 2017, builds on the simple movements and subtle cues of today’s home assistants. It might cock its “head” to indicate it’s listening, for example, or blink its “eye” to acknowledge a request.

Technologies like Alexa are just the beginning, says Knight – “robot protozoa”, waiting to crawl out of the water onto land.

The invisible woman

Ask people about Alexa and they won’t talk about it – they’ll talk about her. After all, “she” has a female name and speaks with a feminine voice. She’s not the only one: Microsoft’s Cortana is voiced by a woman, as is Google Assistant, the voice service that powers the Google Home.

Why? Amazon’s Daren Gill says Alexa’s developers tried many different voices, male and female, and selected the final voice based on what users preferred to hear.

But technology product designer questions if applying female gender cues to virtual assistants risks perpetuating gender stereotypes (see “Lazy coding is teaching software to be sexist“). The roles these devices take on – personal assistant, secretary, carer – are often categorised as “women’s work”. Characterising devices that tackle these tasks as female could further cement this gender trope. Perhaps this is why people might find female voices for these applications more pleasing in the first place.

Yvonne Rogers at University College London suggests users could be given the option to use a male or female voice as is the case with Apple’s Siri and many GPS systems. But Gill says Amazon stuck with one voice for Alexa to focus its efforts on optimising the technology.

This article appeared in print under the headline “Home invasion”

Topics: Artificial intelligence / Psychology