OK Google, what makes an accessible voice interface?

We looked at Google Assistant to see what we can learn about creating accessible voice interfaces.

More than 1 in 5 of households in the UK have a smart speaker (as of June 2019), and this number is expected to rise. Therefore, we are likely to see more and more companies and products developing interactions for the voice interface.

As with all interfaces, companies have a legal and ethical responsibility to ensure that the voice interfaces they build are accessible. By accessible, we mean not only should everyone be able to use it, but everyone should be able to use it equally, effectively and joyously.

At System Concepts, we help our clients build and create accessible products by conducting user research at the early stages of product development as well as assessing the accessibility of products once they have been launched.

Google, along with Amazon and Apple, is one of the leaders of smart technology and voice interfaces. Their voice interface is called Google Assistant and is designed to help users get answers, plan their day, enjoy entertainment, manage tasks and control their home. It can be used across many devices, including Google Home (a smart speaker), which a number of our team own. We were curious to investigate to what extent Google had designed their voice interactions with disabled users in mind. We therefore started to play around with the Google Home and made some observations about what we thought worked well and where we thought there was room for improvement.

What works well:

1.      Google gives you feedback

When we asked Google to play Landslide, Google said, “Landslide by Fleetwood Mac, here it is on Spotify” before playing the track. By hearing your instructions relayed back to you, you can verify whether Google understood your request correctly and whether it is going to complete the action you wanted it to. If it did not understand you, you can correct Google.

This is particularly useful for users with certain accessibility needs, such as anxiety, who may be more troubled if they think Google misunderstood them. Receiving feedback on the action Google will take would give this user group greater peace of mind.

2.      Google prioritises content

When we asked Google what the most popular drinks were in the UK, Google told us that it had found the top 11 drinks in the UK on the YouGov website and proceeded to announce the first four drinks on the list. It then offered an easy way for users to hear the remaining seven drinks on the list if they wanted. By prioritising content and only announcing the first four, it helps focus users on key pieces of information and prevents them from becoming overwhelmed or quickly distracted from the information they were looking for.
This may be particularly beneficial to users with certain cognitive impairments who struggle to concentrate on large quantities of content and / or users who have attention deficits.

3.      You can interrupt Google

If Google is telling you some information you can interrupt it and tell it to stop talking or to repeat something it just said. This gives users control over the voice interaction and allows them to clarify information or instructions as they need.
This would be particularly helpful for users with a hearing impairment who may find it more difficult to understand everything Google said the first time around.

4.      Google allows for pauses

If you pause mid-sentence or take some time to respond to a question Google asked, Google allows some time to pass (generally between five and ten seconds) before it responds and asks you to clarify what you said or tells you that it didn’t understand.
Building in these pauses helps users with speech or cognitive impairments who may need more time to gather their thoughts and say what they want to say.

Recommendations for improvement

1.      Break up instructions

If you ask Google how to sew on a button, Google finds information from a website and reads all the instructions in one go. Although the content given is helpful, the way it is delivered is not. It can be overwhelming to hear all the instructions at once and makes it difficult to follow the steps in real time.

It is also not possible to ask Google to repeat only one of the steps; making it quite cumbersome to remind yourself of a single step, as you will need to listen to all of the instructions again.
This increases the cognitive load on users as they are required to remember large quantities of information at once. This can be frustrating for everyone, but specifically for those with access needs such as a learning difficulty or memory impairment who may not be able to remember all the steps in one go or want Google to repeat one specific step.

2.      Do not use colloquialisms

When we told Google we were going to go to bed, Google said goodnight and told us to have a good time with ‘DJ Pillow and MC Duvet’. On other occasions Google said, ‘Happy hibernating’ and ‘Have fun in sleep mode’. These colloquialisms, I imagine, are supposed to bring some humour and light-heartedness between Google and the user. However, to people with autism spectrum disorder (ASD), colloquialisms can be extremely difficult to understand and to interpret correctly. Many people with ASD have a very literal understanding of language, so incorporating colloquialisms into voice interfaces could bring confusion and distress into the voice interaction, instead of Google’s intended humour and light-heartedness.

3.      Prevent dead ends in conversation

When we asked Google about things to do in London, Google responded, ‘Attractions in London include Big Ben, Coca Cola London Eye, Tower of London and others.’ Google did not offer a way to continue the voice interaction by providing prompts to learn more about one of the three attractions or offer a way to learn about any other attractions. Even when we responded with, ‘OK Google, what others?’ Google responded with, ‘Sorry, I’m not sure how to help’.
This dead-end in the conversation makes it difficult for users to find out more detail or information about what’s going on in London. If users want to be told about more attractions in London, they are required to think of new instructions to give Google to find the information they want, rather than Google offering them ways to learn more. The experience becomes disjointed and frustrating. This can be particularly difficult for users with various cognitive impairments, such as a learning difficulty or short-term memory impairment, who may struggle to think of different ways to get the information they’re looking for or understand what to do next.

4.      Allow users to adjust the speech rate

There are many blind users who listen to their screen readers at an incredibly fast speech rate. This skill to understand fast speech can help them to consume information a lot quicker and save time. Like screen readers, Google should have a function to adjust the speech rate to suit personal preferences and thereby create a more positive user experience.
Furthermore, being able to slow down the speech rate may help users with learning disabilities understand what Google is announcing. Slow speech is particularly helpful for interactions with people who have a learning disability if there are no physical cues to help aid communication, such as facial expressions, which do not exist in voice interfaces.

These are just some considerations to make when designing voice interfaces. With voice interfaces becoming more prevalent, it is increasingly important to think about how to design these interactions with everyone in mind.