Intents every voice app should respond to

vui-design

#1

Following the recent episode of #testing-tuesday, in which @jan and I tested “The Muffin Man” by Chatty Creations, we got interested in whether voice apps should coventionally respond to some set of common requests beside help, stop and cancel requests.

Obvious candidates for such common requests might be:

  • Change the volume! (A topic in ourt testing of The Muffin Man)
  • What can I buy? (A required request for ISP Skills)
  • Repeat! (A good reason to implement a repeat handler is that it’s especially frustrating if this utterances gets mapped to another intent)
  • Launch a different Skill! (Something I get quite a bit in my Mau Mau game Skill, to my slight irritation :sweat_smile:)

This topic touches upon two great talks I that come to my mind:

  1. Philip Hunter’s (of PulseLabs) Voice Summit 2018 talk on encouraging and rewarding users to be spontaneous and explorative with your voice app
  2. Lauren Kunze (of Pandorabots) on common interactions between users and chatbots across various domains (My recommendation is to skip the intro and start at about 4:30)

With these talks in mind, some more candidates for standard handlers are:

  • Romantic/sexual approaches
  • Abuse
  • Thanks
  • Chit-chat like “How are you?”, “Are you a person/machine?” or “How old are you?”

From there on there’s probably a long tail of utterances where the effort of implementing such a response would be in poor relation to the user value it brings.

So: What do you think, should there be a convention about which requests all Skills should handle? Which do you consider most relevant? Or did you already notice particular unexpected utterances that show up a lot in your own intent histories, and that you might have already implemented handlers for? Looking forward to the discussion!


#2

Yes! I really liked this talk and it made me think a lot about how we as VUI designers and developers can provide better user experiences by allowing users to interact more freely.

The main difference between text-based chatbots and voice-based experiences is currently, that users on platforms like Facebook Messenger try out way more (“who’s going to be the next president?”, “do you want to marry me?”), while apps for Alexa and Google Assistant seem to be more structured. Maybe it’s also because people have less time to think about what they’re saying next due to the 8s time window.

Voice apps can’t currently answer a lot of questions for technical reasons (mostly because a too broad language model can cause many mismatches, so adding “easter eggs” could affect the main experience in a bad way), but I think it is important to add more and more things over time.

Maybe seeing that there is a set of intents (like HelpIntent) that works across apps will shift people’s behavior over time and allow for more free, playful interactions.