Download as pdf or txt
Download as pdf or txt
You are on page 1of 41

Designing Chabots and

Human-AI Interaction

Soomin Kim
https://soominkim.xyz
https://brunch.co.kr/@flatdesign
• Conversational Agent Space
• Guidelines for Human-AI Interaction
• Practical Chatbot Strategies
• Conversational Agent Space
• Guidelines for Human-AI Interaction
• Practical Chatbot Strategies
ELIZA (1966)
Bots and Chatbots Populating the Online Space

Not Semi- Intelligent Task-focused


Conversational Conversational Assistants Chatbots
Web crawlers Twitterbots Siri (11), Cortana (14), Facebook M
Alexa (14), Google (16),
Bixby (17)

1990s 2000s 2011 2015


Hundreds and thousands of
chatbots were developed

2015 Facebook launch M


Chatbot 2016 Facebook, Microsoft, LINE, and
IBM launch chatbot platforms.
Rise Slack launch development
investment fund.

Gartner predictions
By 2020, 80% of new enterprise
applications will use chatbots.
By 2021, more than 50% of
enterprises will spend more on
bots than traditional mobile app
development. In the ‘post-app era,’
chatbot will become the face of AI...
Development continues with
challenges emerging

2017 70% if Messenger bots were


Chatbot unable to answer simple
questions (Media post)
Challenges 2018 Facebook shut down M, the
chatbot that helped inspire the
surge.
The Gartner Hype Cycle
2016

2019

2015
2017

2018
Common success factors

Specific tasks e.g. FAQ; survey.


Chatbots into Niche markets e.g. customer service;
counselor; healthcare.
the Plateau of Human-chatbot partnerships Design for
human-chatbot teams, with people handling
Productivity difficult queries.
Human-
Chatbot
Partnerships

Kim, Lee, & Gweon, 2019


Common challenges

Tasks are complex (even simple ones)


Continuing Conversation and natural language is
hard (humans are experts)
into the Plateau The demo trap
(scaling up is challenging)
of Productivity The Uncanny Cliff and Mission Creep
(complications of anthromorphism or
mission creep)

How to build chatbots with


effective social and
conversational strategies?
• Conversational Agent Space
• Guidelines for Human-AI Interaction
• Practical Chatbot Strategies
Establish a foundation for creating
usable, useful, and engaging
Goal user experiences with AI,
including conversational agents .
1. 2.
Make clear
what the
Make clear
how well the
Guidelines for Human-AI Interaction
Initially
system system can
can do. do what it
(Learn More: https://aka.ms/aiguidelines)
can do.

3. 4. 5. 6.
Time services Show Match Mitigate
During based on contextually relevant social
Interaction context. relevant social norms. biases.
information

7. 8. 9. 10. 11.
Support Support Support Scope Make clear
When efficient efficient efficient services why the
Wrong invocation. dismissal. correction. when system did
in doubt. what it did.

12. 13. 14. 15. 16. 17. 18.


Remember Learn from Update and Encourage Convey the Provide Notify users
recent user adapt granular consequences global about
Over Time cautiously.
interactions. behavior. feedback. of user controls. changes.
actions.
INITIALLY

Make clear
what the
system
can do.

PowerPoint’s QuickStarter Howdy bot introducing itself


INITIALLY

Make clear
what the
system
can do.

Make clear
how well
the system
can do what
it can do.
Declaring the purpose in on boarding Clarifying the Capability
DURING
INTERACTION

Time
services
based on
context.

Outlook’s
프로야구봇 알림설정
Time to Leave notification
DURING
INTERACTION

Show
contextually
relevant
information.

Match
relevant
social norms.
MS Word Acronyms Grammarly Goal Setting
DURING
INTERACTION

Mitigate
social
biases.
WHEN
WRONG

Support
efficient
invocation.

Support
efficient
dismissal.

Support
efficient
correction.
Flash Fill in Excel Response Modifiability
WHEN
WRONG

Scope
services
when
in doubt.

Make clear
why the
system did
what it did.
OVER TIME

Remember
recent
interactions.

Learn from
user
behavior.
Outlook Suggested People Personalized Siri App Suggestions
OVER TIME

Update and
adapt
cautiously.

Encourage
granular
feedback.

PowerPoint Designer 업데이트 Grammarly Feedback


OVER TIME

Provide
global
controls.

Word Options
OVER TIME

Provide
global
controls.

Woebot without global menu Chatbot with global/persistent menu


OVER TIME

Notify users
about
changes.

“What’s New” in Office Chatbot Notification


1. 2.
Make clear
what the
Make clear
how well the
Guidelines for Human-AI Interaction
Initially
system system can
can do. do what it
(Learn More: https://aka.ms/aiguidelines)
can do.

3. 4. 5. 6.
Time services Show Match Mitigate
During based on contextually relevant social
Interaction context. relevant social norms. biases.
information

7. 8. 9. 10. 11.
Support Support Support Scope Make clear
When efficient efficient efficient services why the
Wrong invocation. dismissal. correction. when system did
in doubt. what it did.

12. 13. 14. 15. 16. 17. 18.


Remember Learn from Update and Encourage Convey the Provide Notify users
recent user adapt granular consequences global about
Over Time cautiously.
interactions. behavior. feedback. of user controls. changes.
actions.
• Conversational Agent Space
• Guidelines for Human-AI Interaction
• Practical Chatbot Strategies
Chatbot’s Anatomy

AI Branding &
Personality

Context
& Memory Conversation
& Flow

Monetization
Rich Interaction
Building Chatbot’s Persona
• Name: Poncho
• Branding (Key Concept): Fun and humoristic
• Personality: Fun, humoristic, mischievous, comedic,
delightful, sassy
• Environment: Consumer, fun, social, morning, home
• Target Audience: aged 20~40, early adopters
• Task at hand: Get weather forecast and notifications
• Runtime variations: Errors should be handled with
humor
• Values: Get the weather out there and keep it light
Branding / Personality / Character
• How users perceive your chatbot service?
• Branding Components: Naming, logos, colors, character, voice and tone…
Naming & Logo
• Implying functionality
• Easily searchable
• Global issues (korean - english)

vs.
Personality and Tone of Voice

Ex-Pixar storyteller, Emma Coats, is teaching


artificial intelligence how to be affable.

“Pixar is all about finding an emotional


reality in a car or a fish,”

“So that’s something we’ve really used with the


Assistant. We don’t want it to ever be a human
being, right? That’s not what it is. But that
doesn’t mean that A.I. or software can’t have a
perspective on the world.”
Tone of Voice (by Nielsen Norman Group)
Conversation & Flow: Onboarding
• Onboarding: First interaction users see from the bot
• Declaring the purpose
• Teach the user how to use the bot
• Configuration
• Setting the tone and personality
Conversation & Flow: Message Tailoring
Conversation & Flow: Message Tailoring
Comparing Data from Chatbot and Web Surveys
Effects of Platform and Conversational Style on Survey Response Quality

Soomin Kim , Joonhwan Lee , Gahgene Gweon


1 1 2

1
Human-Computer Interaction + Design Lab
2 Cognitive Computing Lab

Seoul National University

You might also like