2. Agenda
PROBLEM CONTEXT SOLUTION -
OPERATIONALIZE AI
ETHICS AND DATA
DON’TS FOR AI ETHICS
RISK MITIGATION
APPROACH: HOW TO
OPERATIONALIZE DATA
AND AI ETHICS
3. Problem
Context
While AI scales solutions at the same time it scales risks as
well as an example ethics. AI ethics and data are business
necessities. With this emerging trend companies need to
publish an action plan for facing the ethical uncertainties.
4. Solution – Operationalize AI ethics and data
Identify existing infrastructure that a data and AI ethics program can leverage.Identify
Create a data and AI ethical risk framework that is tailored to a given industry.Create
Change how ethics is perceived.Change
Optimize guidance and tools for product managers.Optimize
Build organizational awareness.Build
Inspire employees to play a role in identifying AI ethical risks.Inspire
Monitor impacts and engage stakeholders.Monitor
5. Challenges faced
if AI ethics is not
operationalized?
Wasted resources
Inefficiencies in product
development and deployment
Inability to use data for
training AI models
6. Don’ts
for AI ethics
risk mitigation
• Academic approach: results in absence of clear
directives to the developers on the ground and the
senior leaders who need to identify and choose
among a set of risk mitigation strategies.
• On-the-ground approach by engineers, PMs and
data scientists: Lack of expertise and org support to
answer systematically and efficiently any questions.
• High-level AI ethics principles: Rolling out
principles does not help because there are no clear
answers to what is meant by fairness and which
metrics is the right one for decision making.
7. Approach: how
to operationalize
AI ethics
• Frame the problem before developing AI algorithms by engaging with relevant stakeholders early in
the development process and articulate what the product does and does not do.
• Establish KPIs, ethical standards and a governance structure to measure the continued
effectiveness of the tactics carried out for the AI ethics strategy.
• Form an ethics council with members of data, security, legal, compliance and external subject
matter experts (includes ethicists). Create processes to vet for biased algorithms, privacy violations,
and unexplainable outputs. For every AI initiative go through this council for review prior to
development and deployment of AI models.
• Develop quality assurance programs during product development with:
• Self-serve tools for PMs to validate bias and understand explain-ability of algorithms to make
informed trade-offs.
• Define how people’s data is collected, used, and shared
• Break down big ethical concepts like privacy, bias, and explain-ability into infrastructure,
process and practice.
• Increase organizational awareness with necessary trainings and collaterals.
• Reward people for their efforts in promoting a data ethics program is essential.
• Monitor the impacts of the data and AI products that are on use.