4. To learn about each other,
the world around us,
and exchange ideas.
An informal exchange of ideas by spoken words.
Two people, talking back and forth.
Conversation
Why?
Through context
and subtext.
How?
5. Words, sentences, languages, gesture, expression.
Can be imprecise, inaccurate, ambiguous.
Near infinite aperture.
The human condition, in all its breath and depth.
And sometimes, shockingly efficient.
With other People
7. In a conversation, we all think about both what is being said
as well as who is saying it.
These two concepts are inseparable.
Regardless if you’re having a conversation with a person or a thing.
Thinking about who you’re talking to
8. Who Said…
“No, I am your father.” “I have a dream.”
“To be, or not to be…” “Mmoommmyyy. Ddaaaddddyyy.”
9. It is in our nature to attribute human
qualities to things.
Like puppies, the weather, and wifi.
Your voice app has a character, a
backstory, motivations, wants, and
needs.
Because we, as humans speaking
with your voice app, ascribe those
qualities to it.
Anthropomorphism
& Personification
10. “Human fidelity” is a convenient start.
More a measure of complexity than
quality.
Today, we’re within an order of
magnitude in width / # of different
topics.
But we are nowhere near human
fidelity in depth of conversation today.
How should we
measure quality?
Success?
12. This isn’t mobile. This isn’t web.
• No Graphical User Interface (GUI).
• With a GUI, user’s choices are limited by the options presented.
• Even with visuals/touch (Echo Show), it’s still voice-first.
• With voice, the user can say anything, and often does.
13. The basic building blocks
• static
• dynamic
• fallback
Content
• machine learned
• rules based
• blended
Intent
• global
• local
• scoped
Context
14. …the user says something you
expect?
Your basic functionality, your golden
path, and the core utility in your
experience.
You’re likely to get this right.
It’s the business reason you are
building your voice app in the first
place.
What happens when...
15. …the user says something you
DON’T expect?
You communicate the tone, mood,
and style of your experience.
Which is a direct expression of the
core personality of your brand.
When you least expect it.
What happens when...
16. True character is
only revealed
under pressure
How your voice app respond when the
user surprises or challenges it is
crucial.
True character is revealed when one
makes decisions under pressure. And
your voice app is now under pressure.
How will it respond? What will it say?
This is when the true nature of your
experience is shared with the user.
Spend as much time working on this
as you do working on your core
functionality.
18. This is hard. Very, very hard.
It’s how we all talk to each other.
To achieve this, we will all need…
• high quality intent recognition
• scoped intent matching
• content, intent, and content
• integrated in the same toolset
Wide and deep
conversations
19.
20. Some content should be scripted,
some should be dynamic.
Need runtime access to previous
intents and conversation flow.
Full use of state at all points in the
conversation (memory).
Natural language
responses
21.
22. Technical and creative teams,
working together
This is an interactive art.
Think gaming, but with voice and language.
Need technical and creative contributions to build the best work.
Need tools and a platform to encourage, solicit, and combine these together.
Technologists likely best at integrations into existing runtime systems (webAPIs).
Creatives likely best at dialog and conversation flow to craft the tone, mood, and style of
your voice app.