Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
SlideShare a Scribd company logo
Designing Conversational AI at Microsoft:
a Design Toolkit
Josephine Scholtes | December 2021
| UX Services
Josephine Scholtes | UX Consultant
Conversational AI
- bots, virtual agents, digital assistants, and all other text-based dialog interfaces
UX STRAT Online 2021 Presentation by Josephine Scholtes, Microsoft
UX STRAT Online 2021 Presentation by Josephine Scholtes, Microsoft
UX STRAT Online 2021 Presentation by Josephine Scholtes, Microsoft
MICROSOFT BOT FRAMEWORK
UX STRAT Online 2021 Presentation by Josephine Scholtes, Microsoft
Conversational UX (CUX)
- a modality of interaction that’s based on natural language. When interacting with each
other, human beings use conversation to communicate ideas, concepts, data, and emotional
information. CUX allows us to interact with our devices, apps, and digital services the way we
communicate with each other, using phrasing and syntax via voice and text or chat that
come naturally.
Reference: Microsoft CUX Guide
Chat-only Hybrid Voice-only
Context dependent
- Conversational UX is very context dependent, for example:
Voice-only Hybrid Chat-only
Single-user Multi-user
Private Public
Quiet Noisy
Pleasenote that this is an example.These factors are no strict determinants.
Character
Design
Conversation
Design
Character Design
Dave
How do I look?
Just like a mobile app
your assistant icon is
really important.
What’s my name?
Think about your name and
how it reflects your brand
and the type of assistant
you are designing.
How should I sound?
Tone of voice is the primary
way you will inject your
brand into an assistant.
When should I ask for help?
No matter how smart your
assistant is, there will be
cases where you still need to
let a human take control of
the conversation.
Conversation Design
- Dialogs are for bots, like screens are for apps. They separate concerns and organize flows in
a similar way.
Home page
Sub page Search feature
Traditional application
Root dialog
Sub dialog Search dialog
Bot
Understanding intents
- You can think of user intent as the reason why a person is interacting with your conversational experience.
Identifying the true intent ensures your users are matched with the most accurate content to help them
complete their goal.
User query:
“The jeans I bought last week are too big.
Can I get them one size smaller?”
Intent: exchange an item for a different size
Reference: Microsoft CUX Guide
Disambiguating intents
- Disambiguation is the process of narrowing down a user intent by asking clarifying questions, or in some cases
presenting options to choose from, to get a better understanding of the true intent and direct to the correct content.
Ask more questions to narrow
down the intent
Provide options to reduce the
need to ask multiple questions
Ask targeted questions that inform the
user of what information is necessary
to fully understand their intent.
Reference: Microsoft CUX Guide
Design Toolkit
- Canvases and cards for different CUX topics
Why this toolkit?
- Enable people who are new to Conversational UX to design bots, virtual agents and digital assistants
Tone of voice
The primary way to inject your brand in the bot
Saying hello
First impressions matter
Rich user controls
Buttons, carousels, forms etcetera
Avatar
The visual representation of your brand
The “stubborn & clueless” bot
Help, cancel and limit retry options
Ask for help
Even the most intelligent bots need human intervention
Content
- Four canvases and 21 cards
UX STRAT Online 2021 Presentation by Josephine Scholtes, Microsoft
Attendees of
UX STRAT Online
Can quickly pointattendees
to preferred talks
Maintenancebased on
eventcadence
Personal data should
not be recorded
Text-only bot for UX
STRAT Online
website
Practical eventinfo –
refer to pages
Maximumof three
errors
English (US)
Can give customized
information & send calender
invites
UX STRAT 2021 01
UX STRAT Online 2021 Presentation by Josephine Scholtes, Microsoft
Not personal enough 3D design, human
character
UX STRAT 2021 01
Reference: Microsoft CUX Guide
1. insightful
2. community
3. open
4. collaboration
5. international
1. closed
2. sarcastic
3. protective
4. traditional
5. nagging
Buzz Lightyear
1. Mural/Miro
2. Apple
UX STRAT 2021 01
Goal-oriented
UX STRAT Online 2021 Presentation by Josephine Scholtes, Microsoft
Hi [name],welcometo
UX STRAT Online!
UX STRAT 2021 01
My nameis Lux, how
can I help you?
Do the talks have
live closed
captions?
Will therebe live
closed captions?
Yes, please
UX STRAT 2021 01
Sorry, I do not
understand your
question?
Could you repeat?
Hmm I think I’mnot
smartenough to
answer this question
Would you like me to
forward you to my
human colleague?
Content
- Four canvases and 21 cards
Key takeaways
• Conversational AI & Conversational UX
• Voice-only, Hybrid & Text-only
• Character design vs. Conversation design
• Access CUX Design Toolkit @ Figma Community:
https://www.figma.com/community/file/1044999262220337719/CUX-Design-Toolkit
Thank you
Let’s stay in touch: Josephine Scholtes @ LinkedIn
| UX Services

More Related Content

UX STRAT Online 2021 Presentation by Josephine Scholtes, Microsoft

  • 1. Designing Conversational AI at Microsoft: a Design Toolkit Josephine Scholtes | December 2021 | UX Services
  • 2. Josephine Scholtes | UX Consultant
  • 3. Conversational AI - bots, virtual agents, digital assistants, and all other text-based dialog interfaces
  • 9. Conversational UX (CUX) - a modality of interaction that’s based on natural language. When interacting with each other, human beings use conversation to communicate ideas, concepts, data, and emotional information. CUX allows us to interact with our devices, apps, and digital services the way we communicate with each other, using phrasing and syntax via voice and text or chat that come naturally. Reference: Microsoft CUX Guide
  • 11. Context dependent - Conversational UX is very context dependent, for example: Voice-only Hybrid Chat-only Single-user Multi-user Private Public Quiet Noisy Pleasenote that this is an example.These factors are no strict determinants.
  • 13. Character Design Dave How do I look? Just like a mobile app your assistant icon is really important. What’s my name? Think about your name and how it reflects your brand and the type of assistant you are designing. How should I sound? Tone of voice is the primary way you will inject your brand into an assistant. When should I ask for help? No matter how smart your assistant is, there will be cases where you still need to let a human take control of the conversation.
  • 14. Conversation Design - Dialogs are for bots, like screens are for apps. They separate concerns and organize flows in a similar way. Home page Sub page Search feature Traditional application Root dialog Sub dialog Search dialog Bot
  • 15. Understanding intents - You can think of user intent as the reason why a person is interacting with your conversational experience. Identifying the true intent ensures your users are matched with the most accurate content to help them complete their goal. User query: “The jeans I bought last week are too big. Can I get them one size smaller?” Intent: exchange an item for a different size Reference: Microsoft CUX Guide
  • 16. Disambiguating intents - Disambiguation is the process of narrowing down a user intent by asking clarifying questions, or in some cases presenting options to choose from, to get a better understanding of the true intent and direct to the correct content. Ask more questions to narrow down the intent Provide options to reduce the need to ask multiple questions Ask targeted questions that inform the user of what information is necessary to fully understand their intent. Reference: Microsoft CUX Guide
  • 17. Design Toolkit - Canvases and cards for different CUX topics
  • 18. Why this toolkit? - Enable people who are new to Conversational UX to design bots, virtual agents and digital assistants Tone of voice The primary way to inject your brand in the bot Saying hello First impressions matter Rich user controls Buttons, carousels, forms etcetera Avatar The visual representation of your brand The “stubborn & clueless” bot Help, cancel and limit retry options Ask for help Even the most intelligent bots need human intervention
  • 19. Content - Four canvases and 21 cards
  • 21. Attendees of UX STRAT Online Can quickly pointattendees to preferred talks Maintenancebased on eventcadence Personal data should not be recorded Text-only bot for UX STRAT Online website Practical eventinfo – refer to pages Maximumof three errors English (US) Can give customized information & send calender invites UX STRAT 2021 01
  • 23. Not personal enough 3D design, human character UX STRAT 2021 01
  • 25. 1. insightful 2. community 3. open 4. collaboration 5. international 1. closed 2. sarcastic 3. protective 4. traditional 5. nagging Buzz Lightyear 1. Mural/Miro 2. Apple UX STRAT 2021 01 Goal-oriented
  • 27. Hi [name],welcometo UX STRAT Online! UX STRAT 2021 01 My nameis Lux, how can I help you?
  • 28. Do the talks have live closed captions? Will therebe live closed captions? Yes, please UX STRAT 2021 01 Sorry, I do not understand your question? Could you repeat? Hmm I think I’mnot smartenough to answer this question Would you like me to forward you to my human colleague?
  • 29. Content - Four canvases and 21 cards
  • 30. Key takeaways • Conversational AI & Conversational UX • Voice-only, Hybrid & Text-only • Character design vs. Conversation design • Access CUX Design Toolkit @ Figma Community: https://www.figma.com/community/file/1044999262220337719/CUX-Design-Toolkit
  • 31. Thank you Let’s stay in touch: Josephine Scholtes @ LinkedIn | UX Services

Editor's Notes

  1. In 2016 Tay was launched on Twitter by Microsoft as an AI bot, within 24 hrs taken offline Microsoft’s intent: Tay, the creation of Microsoft's Technology and Research and Bing teams, was an experiment aimed at learning through conversations. She was targeted at American 18 to 24-year olds--primary social media users, according to Microsoft--and "designed to engage and entertain people where they connect with each other online through casual and playful conversation.“ Tay was designed to learn from interactions it had with real people in Twitter. Seizing an opportunity, some users decided to feed it racist, offensive information. And in less than 24 hours after her arrival on Twitter, Tay gained more than 50,000 followers, and produced nearly 100,000 tweets. The problem? She started mimicking her followers. According to Microsoft, Tay is "as much a social and cultural experiment, as it is technical." But instead of shouldering the blame for Tay's unraveling, Microsoft targeted the users: "we became aware of a coordinated effort by some users to abuse Tay's commenting skills to have Tay respond in inappropriate ways.“ "Any AI system learning from bad examples could end up socially inappropriate," Yampolskiy said, "like a human raised by wolves.“ The failure of Tay, she believes, is inevitable, and will help produce insight that can improve the AI system.
  2. Incorporated in our products (surface laptops, hololens, M365) Similar to Siri, Alexa, Google Assistant
  3. Back-end > offer as a B2B product Make use of Cognitive Services (includes LUIS – Language Understanding Information System)
  4. BMW, together with Microsoft, is looking to make future conversations with the Bavarian automaker’s Intelligent Personal Assistant even more personalized, not to mention more natural-sounding and “multi-modal”. BMW customers could for example make an appointment at their preferred BMW dealership, simply by talking with the personal assistant – a conversation that might start with a reminder that service is due, and finish with the system arranging the appointment. Users could also manage their personal e-mails and calendar appointments the same way while on the move.
  5. If the user isn’t matched with the correct intent, they may receive wrong or misleading information, causing frustration and the potential loss of trust in your brand.
  6. If the user isn’t matched with the correct intent, they may receive wrong or misleading information, causing frustration and the potential loss of trust in your brand.
  7. If the user isn’t matched with the correct intent, they may receive wrong or misleading information, causing frustration and the potential loss of trust in your brand.
  8. If the user isn’t matched with the correct intent, they may receive wrong or misleading information, causing frustration and the potential loss of trust in your brand.
  9. If the user isn’t matched with the correct intent, they may receive wrong or misleading information, causing frustration and the potential loss of trust in your brand.