There is a growing interest in AI, but the field is understood only by a few, often by technicians. What’s our involvement as designers? What do we know about these new disruptive technologies? Often we know little. This talk will unveil benefits and pitfalls of AI through concrete cases offering designers a framework for understanding. It will introduce a toolkit of ethical principles, that will make you reflect upon the potential of AI to change what is our conventional understanding of a product or a service. Design is about developing a higher awareness of the role we play as creators and builders.
6. Is it right? Is it fair? Are there obligations?
Are there ethics guidelines, checklists I could follow?
…
7. Leyla Acaroglu
Designer and sociologist
Ethical conventions are the product of social dialogue
and debate resulting in a normalized collective
agreement of approaches that fit the most humans for
the most benefit.
It’s not only about what is legal…
8. Any individual can spark an ethical movement, but it
takes a group to sustain it […]. Discuss ethics, and
engineer the time and space for these
conversations to happen
Future Ethics – Cennydd Bowles
9. Co-Creation
Any discussion will resonate within your
team as well as outside of your office.
Advocacy – Co-create together with
stakeholders and users. Toolkits and the
associated workshop demystify
technology and help non-tech experts to
create smart service concepts.
Testing the ground, qualitative
interviews, co-design exercises,
Prototypes build to provoke.
UR & Provocatypes Debates
IoT service toolkit
iotservicekit.com
The Intelligence Augmentation
iadesignkit.com
12. Virtue
is a trait or quality that is deemed to be
morally good and thus is valued as a
foundation of principle and good moral being.
Christopher Peterson and Martin Seligman
Healthy and stable personality given by 24 virtues
15. John McCarthy
AI is “the science and engineering of making
intelligent machines.”
1955
AI is Machine Learning
Machine Learning is AI
AI, then is… 🤔
16. Intelligence
Ability to accomplish complex goals such as
understanding, self-awareness, problem solving,
learning, etc…
Life 3.0: Being Human in the Age of Artificial Intelligence
Max Tegmark
17. Artificial
General
Intelligence
System able to accomplish virtually any of the
complex goals listed previously.
Human-like Intelligence
Artificial
Super
Intelligence
System able to accomplish virtually any complex
goals beyond our imagination/understanding.
Greater than Human Intelligence
Artificial
Narrow
Intelligence
Less than Human Intelligence
The holy grail of AI research
Intelligence
Augmentation
(IA)
The application of digital technology to amplify and im
capabilities including thinking, analysis and planning.
Work in close cooperation with humans
18. Artificial
General
Intelligence
System able to accomplish virtually any of the
complex goals listed previously.
Human-like Intelligence
Artificial
Super
Intelligence
System able to accomplish virtually any complex
goals beyond our imagination/understanding.
Greater than Human Intelligence
The holy grail of AI research
Intelligence
Augmentation
(IA)
The application of digital technology to amplify and improve human
capabilities including thinking, analysis and planning.
Work in close cooperation with humans
22. Deep Learning
on Raw Audio
Neural network model which achieves an
understanding of the song, including
characteristics like estimated time signature,
key, mode, tempo, and loudness.
Spotify’s songs (old and NEW)
Natural
Language
Processing
Model which analyse text in
News and Blogs
Collaborative
Filtering
Model which analyse both your and other’s
behaviours.
Similar users:
• Implicit Feedback (Spotify)
• Rating system (Netflix)
Recommendation models at Spotify:
24. Adaptation
Encourage users to give feedback that is
mutually beneficial to them and to your
service. Show them the benefits of providing
feedback.
Enablement
Automation is cool but on a long run it can
flatten what before was an enriched
experience. Provide the user the possibility
of adjusting the predictions.
Friendliness
The mechanisms sitting behind an interface are
getting more complex and hidden. The challenge
is to pursue a design that is glanceable and
at the same time provides a clear picture of
how the system works to the user.
25. Boosting business productivity
Efficiency & Velocity
Perform tasks more effectively
Fraud detection
Supply management
Customer segmentation
Predictive maintenance
Complete tasks more rapidly
Legal document and claims processing
Manufacturing process optimisation/defects
Medical diagnosis
Managing finances
2/3
26. Stock shelves more efficiently based
on past sales data and other factors
such as weather forecast
VEKIA
Your Supply Chain enhanced by our Artificial
Intelligence. In the retail activity, the
challenge is always to have the right stock at
the right place at the right time and in the
right quantities.
-8%
inventory
+2%
sales have risen
*numbers provided by Vekia
27. Warehouses workers could fill orders faster
The proposed wristbands would use ultrasonic tracking to
identify the precise location of a worker’s hands as they
retrieve items. One of the patents outlines a haptic
feedback system that would vibrate against the wearer’s skin
to point their hand in the right direction.
Working at Amazon
Streamline “time consuming” tasks
through wristbands, like responding to
orders and packaging them for speedy
delivery.
28. Unfair treatment of
employees
“After a year working on the floor, I felt like I had
become a version of the robots I was working
with. You had to process the items in seconds
and then move on. If you didn’t meet targets, you
were fired.”
Orwellian snooping
on workers
New York Times, February 2018 – Phone interview with Max
Crawford, a former Amazon warehouse worker in Britain.
29. Circumspection
Careful consideration of circumstances and
consequences. E.g. Do we want to apply IA to
screening job candidates? What are the risks?
The decision of applying IA is not only based
on business case suitability.
Trustfulness
Acting in a way that inspires confidences and
trust. New types of agreement between parties
like employer and employees. For example, most
employment contracts in US give employers
rights to monitor employees and collect data
while Europe has stronger privacy laws.
30. Accountability
The condition to report and justify
algorithmic decision-making, as well as to
mitigate any potential harms on an individual
and societal scale.
Accuracy
How to identify source of
errors and uncertainty in
an algorithm.
Explainability
Any decisions (produced by an
algorithm) should be accessibly
explained to whom is affected by
those decisions.
Fairness
The results of discriminatory
evaluation should be released publicly
together with the criteria used.
Responsibility
There needs to be someone
responsible to deal with
harming effects on
individuals/society.
Auditability
Algorithms should be
developed to enable third
parties to probe and
review the behaviour of an
algorithm.
31. Fostering safety
AI-Safe, Automated Intelligent System for
Assuring Safe Working Environments’
funded by InnovateUK
3/3
Cortexica
a London based, global provider
of B2B Vision AI solutions for
Digital Transformation.
InnovateUK
the UK's innovation agency
reporting to the Department for
Business, Energy and Industrial
Strategy (BEIS).
32. 1.85m
CCTV cameras in UK, mainly
owned by private company (2011)
1/32
CCTV cameras/citizens
(UK, 2011)
https: // www.cortexica.com/solutions-video/
Personal
autonomy
300
times/day caught on
CCTV in London (2017)
Social
responsibility
Action Recognition for CCTV security
systems for detecting crime and anti-social
behaviours.
33.
34. Creativity
Use constraints as a source of inspiration.
Find alternative solutions that care about
doing good.
Respect
Plan, design and build services with privacy
as well as other human rights at hearth.
Foresight
Consideration of the consequences of one's
action; thinking ahead. For example, be
prepared for potential regulatory changes.
35. Creativity
Use constraints as sources of inspiration.
Find alternative solutions that care about
doing good.
Foresight
Consideration of the consequences of one's
action; thinking ahead. For example, be
prepared for potential regulatory changes.
Circumspection
Careful consideration of circumstances and
consequences. E.g. Do we want to apply IA to
the legal system? What are the risks? The
decision of applying IA is not only based on
business case suitability.
Respect
Plan, design and build solutions with privacy as
well as other human rights at hearth.
Trustfulness
Acting in a way that inspires confidences and
trust. New types of agreement between parties
like employer and employees. For example, most
employment contracts in US give employers rights
to monitor employees and collect data while
Europe has stronger privacy laws.
Accountability
The condition to report and justify algorithmic
decision-making, as well as to mitigate any
potential harms on an individual and societal
scale.
• Responsibility
• Auditability
• Accuracy
• Explainability
• Fairness
Adaptation
Encourage users to give feedback that is mutually
beneficial to them and to your service. Show
them the benefits of providing feedback.
Enablement
Automation is cool but on a long run it can
flatten what before was an enriched experience.
Provide the user the possibility of adjusting
the predictions.
Friendliness
The mechanisms sitting behind an interface are
getting more complex and hidden. The challenge
is to pursue a design that is glanceable and at
the same time provides a clear picture of how
the system works to the user.
… and there are many more virtues to explore
36. RecoverMitigate
Anticipate
Law of unexpected consequences
There will always be outcomes we
overlook but unintended doesn’t
mean unforeseeable.
1
2 3
Externalities
Someone Else’s Problem.
37. Future Ethics – Cennydd Bowles
“Most (tech employees) genuinely want to improve the
human condition, or at least tackle interesting problems,
and have good intentions. The industry’s problems are
mostly down to negligence, not malice.”
38. Politics of the Very Worst – Paul Virilio
When you invent the plane you also invent the plane crash;
when you invent electricity, you invent electrocution… Every
technology carries its own negativity, which is invented at
the same time as technical progress.