Critical commentary based on my professional experience in designing apps with artificial intelligence and on desktop research. Presentation slides for Botscampe 2016.
Danger of lowest
common
denominator by
statistical
inference
Pitfalls of common pattern
While constructing classes
edges cases are excluded to
avoid overfit.
Anthrax, metal fume fever,
leukaemia said to have
common symptoms to flu
Body size adjustment for
gestural interface
Closer look on edge data for innovation
Questioning, learning to give a
proper interpretation and outcomes
on the edges.
Design tip 1
Effort to control on
the inferred logic
Controlling one’s error
Users by mistake have
communicated wrong data. The
System computes as a final
answer without users being
able to take it back.
Tiny
Making the learning mechanism transparent
and explicit
Who are you?
How do you learn?
➔ Reinforcement learning (rewards,
punishment etc.)
➔ Transfer learning
➔ Deductive logic learning
➔ Adaptive learning (user becomes
teacher)
➔ Geometric intelligence (shapes,
movement)
➔ etc.
Design tip 2
Persuasive vs.
Friendly tone of
voice
Letting users’ control the system
behavior for its own detriments for
behavioral change ?
When users are using the
system to change their
behaviour should the system
be friendly? Can “friendly”
be persuasive enough to
change users’ behavior?
Behavior change triggers in persuasive design
ExpertAmateur
Good Methods:
● Focus on failure
● Coaching
● Persistence
● Deliberate practice:
● Basic physical practice
Goal:
Professional
Tennis player
Bad Method:
● Baby step without
setting strict goal
Adjusting bots communication strategy
according to the personality of user?
With Sensing + Thinking People
● Be specific, confident, well-reasoned
● Demonstrate immediate advantages
● Provide examples; use visual aids
With Sensing + Feeling People
● Be supportive, expressive and
confident
● Provide examples; demonstrate
immediate advantages, profit
● Appeal to emotions and feelings by
making it “personal”
With Intuitive + Thinking People
● Be specific, well-reasoned; use
diagrams
● Use concepts, theories to appeal to
intellectual capabilities
● Give them a challenge
● Show how the subject of communication
fits in “big picture”
With Intuitive + Feeling People
● Be expressive, well-reasoned via
visual aids
● Appeal to their intuition
● Give them a challenge
● Show how the subject of communication
fits in “big picture”
Design tip 3
Expression style
and personality
Today’s weather?
Team’s cognitive bias
Team is unaware of how their
cognitive bias is shaping
the personality of the ML
system.
Letting users’ sense to
evaluate the weather
Giving a
straightforward data
Gap between ML
system and
avatar’s
representation
How do you represent appropriately the system’s
cognitive capability?
Right expectation on the system
Avatar representation may
not translate the capability
of the system, therefore
setting false expectation by
the users
3~6 year old
1~2 year old
Unknown = expectation is hard
to set
Awareness on how
it shapes our
culture
New social skills, new education
system for our society?
Unwittingly media
technology, including
communication tools such as
chatbots, shapes the way we
communicate and think. The
Machine is a social being.
http://sunsetlovesarii23.de
viantart.com/art/hatsune-mi
ku-live-in-tokyo-japan-3390
70511
Miku hologram
concert
Shaping cultural extremes or
bubble based on simplistic
“average” algorithm of
click.
Evaluating impacts on society in a
multi-disciplinary design team
pre-2016
Design concerns over usability of
UI artefacts in a given context.
Inclusion of designers and user
researchers to integrate human
aspects.
2016 Future
Design concerns over long haul
effects on humanity.
Inclusion of philosopher,
historian, social scientists to
make an informed decision on
cultural and ethical values.
Design tip 4
New skills in ML system design project team
Social scientist
Writer
Philosopher
How about integrating different skillset
in AI engineering team?
Designer/Artist
Historian
Behavior psychologist
About me
My past design experiences
related to machine learning
system is in the domains of
natural language process
(translation tool) and
computer visions applications
(gestures, image recognition)
as an interaction designer
collaborating with engineers.
Akemi Tazaki id_farm
https://medium.com/idfarm
References
A survey of socially interactive robots
http://www.cs.cmu.edu/~illah/PAPERS/socialrobot
icssurvey.pdf
Empirical Evidence for a diminished sense of
agency in speech interfaces
http://www.davidcoyle.org/docs/paper1302.pdf
The problem with ‘Friendly’ Artificial
Intelligence
http://www.thenewatlantis.com/publications/the-
problem-with-friendly-artificial-intelligence
The chatbot will see you now: AI may play
doctor in the future of healthcare
http://www.digitaltrends.com/cool-tech/artifici
al-intelligence-chatbots-are-revolutionizing-he
althcare/
For sympathetic Ear, More Chinese Turn to
Smartphone Program
http://www.nytimes.com/2015/08/04/science/for-s
ympathetic-ear-more-chinese-turn-to-smartphone-
program.html?_r=0
How a chatbot could help people take their
medication
http://venturebeat.com/2016/09/03/how-a-chatbot
-could-help-people-take-their-medication/
Behavior Wizard: A method for Matching Target
Behaviors with Solutions
https://pdfs.semanticscholar.org/4f7b/2df5d0d5a
6fbc0cb3fe7d0baf352ec69c100.pdf
Interacting with an Inferred World: The
Challenge of Machine Learning for Humane
Computer Interaction
https://www.repository.cam.ac.uk/bitstream/hand
le/1810/248690/Blackwell%202015%20%20Critical%2
0Alternatives%202015%20-%20The%205th%20Decennia
l%20Aarhus%20Conference.pdf;sequence=1
March of the machines
http://www.economist.com/news/leaders/21701119-
what-history-tells-us-about-future-artificial-i
ntelligenceand-how-society-should
Is Anything worth Maximizing?
http://nxhx.org/maximizing/
How Technology Hijacks People’s Minds - from a
Magician and Google’s Design Ethicist
http://www.tristanharris.com/essays/
Behavior measurement and change
http://matrix.berkeley.edu/research/behavior-me
asurement-and-change
Alone Together: Why we Expect more from
Technology
Sherry Turkle
Behavior change matrix
http://www.nirandfar.com/
Behavior Change strategy cards
https://www.artefactgroup.com/resources/behavio
r-change-strategy-cards/
Transparent Active Learning for Robots
http://www.cc.gatech.edu/social-machines/papers
/chao10_hri_transparent.pdf
All photos are credited inline.