The experts agree: we’re living in a new machine age.
Since the wide adoption of the internet in the 1990s, the Internet has been focused around people. Beginning with subscription access to the internet in the days of AOL and Netscape, the Internet has given us new ways of communicating with one another.
Online advertising was based around connecting with visitors to specific web pages before it was superseded by Google’s paid search and more sophisticated versions of communicating with consumers.
The age of the social network was based around connecting with your friends and peers.
The move to mobile has enabled us to stay constantly connected to one another.
So what’s next? We’re leaving the internet of people behind. The next stage is an Internet based around machines. Often termed the Internet of things, this includes household appliances such as your lightbulbs and your fridge, and extends to infrastructure and transport, such as autonomous cars.
But the Internet of people needs to interact with the Internet of things. And this is where we fit in. We’re building software that lets people interact with their machine. In our case, this is specifically for the in-car environment.
What’s the best way to do this? This is the challenge. And although there are numerous technical hurdles to overcome, the main challenges are about what we want in our interactions with machines.
When building voices for machines, it isn’t the case that the more human the voice and the more natural the dialog, the better we like it. This is known as the ‘Uncanny Valley’ phenomenon. This describes the effect that increasing human likeness has on artificial figures. When the artificial figure is perceived as more humanlike it is also perceived as more agreeable--but only until it comes to look so humanlike that we start to find its nonhuman imperfections unsettling. When this happens, acceptance declines dramatically and the figure is regarded as uncanny, eerie and creepy.1
This is something we have to pay particular attention to. When building a digital assistant for the vertical of mobility (driving), there are bound to be a few mistakes and imperfections, so it’s important that we continually test user reactions to ensure that we’re not falling into the uncanny valley--particularly important as we need people to trust their voice assistant as much as possible in the in-car environment.
Besides exhaustive user research, there are a few other solutions that we’re taking into account.
Firstly, we can encourage users to view their voice assistant as something they might have a relationship with. We can do this by adding elements of personality: by giving it a name, for example. For our voice assistant, we went with Chris. To enhance the effectiveness of this, we can also encourage the user to address the device with a ‘hot word’, which wakes it up--’Hey Chris’, for example. By cultivating a relationship, users are more willing to forgive slight deviations from the perfect appearance of humanity.
Secondly, we can use our device’s features to inform the user of its activity. If a device shows a blank screen but does what a user says without any sign of what it’s doing, humans find this strange--in some scenarios, such as when pedestrians want to cross the road and require autonomous cars to stop for them, this can be dangerous. To overcome this, our device’s screen displays a microphone to show that it’s working, just as some autonomous cars use lights to let pedestrians know that they’ve been seen.
Finally, we can work on making the device’s dialog manager as seamless as we can. This means working with as much data as possible, which we get from user tests and use of the device.
Minimizing the uncanny valley effect is not easy, but by approaching the problem from multiple angles we’re sure we can do it and build successful software.
- Technology is now shifting from the internet of people to the internet of things--a new machine age.
- In order for this change to be effective, people need to trust their interfaces
- One problem with this is the ‘uncanny valley’ issue, where people stop trusting their devices as they become more humanlike but retain flaws.
- There are several ways to overcome this, and we are deploying them to make sure our device, Chris, doesn’t fall into the uncanny valley.
1 Mori, 1970; Mara and Appel, 2015
Image: By Half-pass (Own work) [CC BY-SA 4.0] https://creativecommons.org/licenses/by-sa/4.0, via Wikimedia Commons