#NFIM18 - Anna Felländer - Senior Advisor, The Boston Consulting Group
1. In order to embrace the positive gains from AI, the ethical implications need to be identified, measured and governed
By understanding and communicating the implications of AI in a broader and ethical context, trust, integrity, diversity, gender
equality and environmental sustainability could be strengthened
The short-term goal is to create an operational framework for Sustainable AI.
The medium-term goal is to create a standard for certifying sustainable handling of AI data.
The long-term goal is to strive for positive impacts on society.
AI Sustainability - responsible governing of AI
Anna Felländer, September 17
2. AI at a cross road
Massive negative intended and unintended consequences if Artificial
Intelligence is not governed in a sustainable and ethical way…
…AI Sustainability intends to focus on mitigation of risks as well as the
realization of vast gains to society by acting proactively, in a
multidisciplinary manner together with large organisations, start-ups,
governments and academia
4. Should agencies be contacted if a person drives
drunk?
How far can we go in our collection of personal
information for credit risk scoring?
Is this compatible with our Swedish and
organizational core values? But what about the
values in China?
Is it ok for a gaming app to capitalise on persons
with a gambling addiction?
Is it ok when media recommend articles to a
person with racial opinions supporting his or her
orientation?
In the AI era, ethical questions need to be on the top of the agenda
5. AI Sustainability - "corporate responsibility 2.0"
Data is the new gold
Equally important to capture the
value from data and AI, is the
need for active and holistic
ethical principles regarding the
use data and AI.
The difference of AI is scale
AI is scaling in a new and autonomous way.
Explicit & implicit assumptions are reproduced
by self learning algorithms. Given goals such
as sales and productivity, the overview over
ethical considerations, and how they are
scaled in a broader context, are easily lost.
Do good through tech
Ethical concerns around AI
are increasing from
academia, politicians, civil
society & the private sector.
New competitive edge
1. Differentiate by proactively addressing
customers’ increasing demand for transparency in
AI decisions
2. Avoid future pitfalls by applying frameworks and
standards that strengthen trust
3. Unique opportunity to extend and strengthen
sustainability in general, and gender equality in
particular.
6. Companies are increasingly recognizing the need for an AI
ethical framework
• We are already seeing discrimination and
unintended harm to people
• Most organisations, and their partners are
exposed due to wide and deep client database.
• Lagging regulatory framework on AI and ethics,
as well as data driven business more broadly.
• Increasing trend in the business and human
right context - it is no longer accepted to say
“we didn’t know…”
• Customers will demand transparency and
communication around ethical considerations.
• Few organisations have confidence in the
fairness and ability to audit their AI systems
• Technological development is galloping,
seductive profits in different parts of
organizations
• Unclear organisational responsibility over AI
ethics - sustainability, marketing, compliance,
digital and IT?
• Organizational structures - data scientists often
work in silos and lack skills on how AI can be
scaled in a broader ethical context.
• GDPR compliance has been an urgent driver
Why is an AI ethical framework urgent? Why has it not been done yet?
7. AI specific ethical unintended pitfalls
The bias of
the creator
Lack of theor.
understanding
• Violate customer
integrity by using
too broad or deep
open data to
maximize short
term profits
– E.g. separate
open data
sources can be
combined and
create
intelligence that
is privacy
intrusive
• AI is not neutral.
AI decisions lie in
the hands of a
technologist
without full
overview or skills
of the ethical
broader
implications on
customers &
suppliers
Data &
machine bias
• The data
available is not
an accurate
reflection of
reality or the
preferred reality
and may lead to
incorrect and
unethical
recommendatio
ns to customers
Immature AI
• Insufficient
training of
algoritms on data
sets could lead to
incorrect and
unethical
recommendations
to suppliers &
customers
• Decisions based
on AI without
being understood
or evaluated by
responsible org.
• Potential negative
impact on
employees or
suppliers from
inaccurate
business
decisions
Misuse of
AI & Data
8. Regulatory blindspot and lack of governing skills -
Standards for Sustainable AI
Artificial
Intelligence,
Robotics and
‘Autonomous’
Systems
Statement on
European Group on
Ethics in Science and
New Technologies
Research and
Innovation
• Transparency
• Predictability
• Accountability
9. What does an AI Sustainability Strategy entail?
1. Being a proactive thought leader in the application of new technologies
2. The pace of change has never been faster and the promise of efficiency and profits is
seductive. In today's purpose driven economy, sustainability strategies must aim to
combine profits with sound and sustainable considerations in AI based decision.
3. Avoiding AI specific unintended pitfalls
4. Translating existing ethical principles into the AI framework
5. Strengthening and extending traditional sustainability and environmental impact
assessments
10. Processes for AI Sustainability and Ethical Maturity Index
Description of how
client data is used
based on risks and
fears
Communicating
existing and future
handling of client data
as well as AI ethical
principles
Communication
Identification of current
and future data and AI-
based processes with
an ethical dimension.
Description of how
these are scaled with AI
Analysis of how current
ethical framework
should be applied in the
AI setting
Ethical analysis
of data and AI
processes
Developed tailored
approach and policies
for handling data and
AI
Ethical guidelines for
“clean” input data,
algorithms and follow-
up of results
Design of a governing
modell as well as
warning and control
systems
Framework for
AI Sustainability
Identifying the largest
ethical risks with AI from
value based, technical
and communicative
angels
Assessment of
probability and
consequences of risks
Plan for mitigation of
identified risk
Risk identification
och risk
management