Training Industry Magazine - March/April 2019 - 53

WHAT'S NEXT IN TECH
STELLA LEE, PH.D.

THE ETHICS AND
ACCOUNTABILITY OF AI
FOR LEARNING
Artificial intelligence and machine
learning continue to take on a greater
role in the learning and development
field: providing personalized learning
paths, predicting our performance, and
making recommendations on content and
activities. This year, AI systems will support
our workplace learning trajectory more
than ever, but only if we can be deliberate
in establishing ethical frameworks and
make its use more transparent. Below
are some of the challenges we face as
we develop AI applications for learning
and some potential ways to address
ethical concerns.
Current Challenges
* Black box algorithms: Increasingly,
researchers are saying that algorithms
used by organizations can be opaque
and discriminatory. Black box algorithms
refer to the general inability to see inside
of a system and see how it arrives at a
decision. For example, when a learning
platform makes recommendations to
a learner, how would you know what
learner attributes the platform uses
as criteria for such recommendations
and if they are conducive for learning.
Even more troublesome, there are few
options, unless you are a computer
programmer, for updating and tweaking
these assumptions and bias that are
inevitable in these systems.
* Issue of trust: In measuring learning
and understanding how and what
to personalize to individual learners,
companies and platforms are collecting
a huge amount of personal data. At the
same time, ownership and governance
of data is often ill-defined, or not
defined at all, for many organizations.
Lack of definition promotes a feeling of
mistrust among employees, while data
is being collected without clear and
conscious learner consent. In today's

"big brother" surveillance culture,
people don't inherently trust machine
learning algorithms, especially when
the decision-making process behind
the scene is lacking transparency
and accountability.
* Lack of guiding principles and
governance: While tech giants such
as Google, Microsoft , and SAP have
all introduced ethical principles for AI,
there is a lot of room for more concrete
definitions and better oversight. One
criticism over AI principles set by tech
firms is that they are more concerned
with protecting themselves from legal
actions and less on coming up with a
realistic, actionable set of guidelines.
In the L&D world, the issue is further
complicated by changing standards in
e-learning and learning management
system technology whereby learner
profiling and data tracking are often not
made explicit and without user consent.

THIS YEAR, AI SYSTEMS
WILL SUPPORT OUR
WORKPLACE LEARNING
TRAJECTORY MORE
THAN EVER.
Potential Resolutions
* Explainable AI: One notion to counter
black box algorithms is advocating
the use of explainable AI, XAI.
Essentially, XAI programs enable users
to understand the technology, provide
input on the decision-making process
and improve algorithmic accountability.
In the case of dynamically generated
learning paths, learning platforms
should be explicit about the decisions
it makes to recommend certain
learning paths or options over others.
In addition, learners, course designers,

and instructors are given the ability to
review and update as they see fit.
* Privacy by design: Privacy needs to
be considered by the whole design
process, not just as a bolt-on approach.
Privacy by design is a system
engineering approach proposed by
Dr. Ann Cavoukia. Essentially, privacy
by design advocates that organizations
need to consider privacy at the initial
design stages and throughout the
development process of new products
or services that involve handling
personal data.
* Regulatory frameworks for AI:
Regulatory frameworks are needed to
safeguard transparent and auditable
education data. Prejudices of the past
must not be unwittingly built into AIenabled learning systems. Systems
and algorithms need to be carefully
designed, explained, and audited. While
government plays a key role in setting
regulatory frameworks, organizations
should collaborate and encourage
discussions on the ethics of AI by
setting up advisory bodies, developing
core ethical principles and responsibly
handling data.
For AI to deliver its promise on learning
enhancement, it will require people
to trust the technology, organizations
to safeguard the use of personal and
educational data, and for government to
collaborate with industries to research
and discuss ethics in AI.
Dr. Stella Lee has over 20 years of experience in
consulting, planning, designing, implementing,
and measuring learning initiatives. Today,
her focus is on large-scale learning projects
including LMS evaluation and implementation,
learning analytics, and artificial intelligent
applications in learning. Email Stella.

T RAIN IN G I N DU STR Y M AGAZ INE - TRAINING IN ANOTHER DIMENSION 20 1 9 I WWW. T RAI NINGINDU S T RY . C OM/ MAGAZ I NE

| 53

https://www.ipc.on.ca/wp-content/uploads/Resources/7foundationalprinciples.pdf https://www.ipc.on.ca/wp-content/uploads/Resources/7foundationalprinciples.pdf https://www.ai.google/principles/ https://www.microsoft.com/en-us/ai/our-approach-to-ai https://www.sap.com/canada/products/leonardo/machine-learning/ai-ethics.html https://www.thenextweb.com/artificial-intelligence/2018/06/08/googles-ai-principles-7-shades-of-gray-area/ https://www.thenextweb.com/artificial-intelligence/2018/06/08/googles-ai-principles-7-shades-of-gray-area/ https://www.wired.com/story/ai-experts-want-to-end-black-box-algorithms-in-government/ https://www.darpa.mil/program/explainable-artificial-intelligence https://www.trainingindustry.com/magazine

Table of Contents for the Digital Edition of Training Industry Magazine - March/April 2019