Conversational UI

Lots of what passes for an AI revolution is actually a change in interaction model, enabled by machine learning that has gotten “good enough” to suggest a single answer to a query.

This is an important shift. As compute becomes smaller and cheaper, it is infecting everyday objects that we don’t think of as “computers”. In the future, everything will be a little smart. Conversational UI will be an important tool for interacting with these smart things, because many will be too small to have screens.


  • Works anywhere – scales from screen-less devices to mobile to Desktop.
  • Works with or without a screen.
  • Conversations are an anthropomorphic way to interact with computers.
  • Conversational UI can seamlessly blend interactions between people and artificial intelligence agents.
  • Messages, conversations and notifications have heavy conceptual overlap.
  • Search, conversations, notifications, messages and cards — these things have a strong overlap in terms of jobs and features.
  • This is why WeChat is able to expand into a platform.

Risks and Challenges

  • The rule of thumb for machine learning is that 80% success rates are easy, but the next 10%-15% gain is hard.
  • Failure hurts. How often are you quietly exasperated by Siri? When users feel punished by failure, they stop exploring and stick to the beaten path.
  • Where AI falls short, careful UI can patch up the differences. This is where design patterns come in. WeChat does this very well, with structured responses.


  • When voice assistants get good enough, every business will have a programmable API, because every business has a phone.
  • Another big question for me is what happens when we have many of these bots, conversing with each other?
  • Where we’re headed: instead of a phone, you have a watch with cellular modem and a pair of AirPods (headphones). A tablet gives you a screen when you need it.
  • We’ll need app frameworks to help author these bots, you and the frameworks will make assumptions about how conversations work in small groups.

Design Patterns

  • The Cooperative Principal: we trust that others we converse with are being cooperative, and will contribute usefully to the conversation. Avoid breaking this deal at all costs.
  • Be As Smart as A Puppy. If the computer is dumb, make it charming.
  • Keep conversations structured. If the you can’t have 100% accuracy, narrow the problem space with careful UI design.

The Jack Principles of conversational UI are a game design guide for the early text-based quiz game “You don’t know Jack”. There is some truly excellent design thinking here:

Maintain pacing:

  1. Give the user only one task to accomplish at a time.
  2. Limit the number of choices the user has at any one time.
  3. Give the user only meaningful choices.
  4. Make sure the user knows what to do at every moment.
  5. Focus the user’s attention on the task at hand.
  6. Use the most efficient manner of user input.
  7. Make the user aware that the program is waiting.
  8. Pause, quit or move on without the user’s response if it doesn’t come soon enough.

Respond with human intelligence and emotion to:

  1. The user’s actions
  2. The user’s inactions
  3. The user’s past actions
  4. A series of the user’s actions
  5. The actual time and space that the user is in
  6. The comparison of different users’ situations and actions

Maintain the illusion of awareness:

  1. Use dialogue that conveys a sense of intimacy
  2. Make sure characters act appropriately while the user is interacting
  3. Make sure dialogue never seems to repeat
  4. Be aware of the number of simultaneous users
  5. Be aware of the gender of the users
  6. Make sure the performance of dialogue is seamless
  7. Avoid the presence of characters when user input cannot be evaluated

Brand Bots and Augmented Conversations

Something fascinating is going on with WeChat:

WeChat has popularized the concept of “official accounts” for brands and public figures. They’re kind of like the IRC and AIM bots of yore — think SmarterChild but for banks, phone companies, blogs, hospitals, malls, and government agencies. Many institutions that otherwise would have native apps or mobile sites have opted instead for official accounts.

You can send any kind of message (text, image, voice, etc), and they’ll reply, either in an automated fashion or by routing it to a human somewhere. The interface is exactly the same as for chatting with your friends, save for one difference: it has menus at the bottom with shortcuts to the main features of the account (though it can be toggled away to reveal the normal text field).

Because you’re having a conversation over text, you can swap out the AI with a person on the other side. Conversational systems could be hybrid Augmented Intelligences

This approach is a bit like being able to cheat the Turing Test… passing a message from a human when convenient.


  • Updated
    Oct 3, 2016