=Paper=
{{Paper
|id=Vol-3901/paper_5
|storemode=property
|title=Embedding Responsible AI in Technical Education Curriculum: A Case Study in Asynchronous Online Advanced Data Analytics Course
|pdfUrl=https://ceur-ws.org/Vol-3901/paper_5.pdf
|volume=Vol-3901
|authors=Rhoda Abadia
|dblpUrl=https://dblp.org/rec/conf/tethics/Abadia24
}}
==Embedding Responsible AI in Technical Education Curriculum: A Case Study in Asynchronous Online Advanced Data Analytics Course==
https://ceur-ws.org/Vol-3901/paper_5.pdf
E m b e d d i n g Responsible Al in Technical Education
Curriculum: A Case Study in an A s y n c h r o n o u s O n l i n e
A d v a n c e d Data Analytics Course
Rhodora Abadia 1
1
University of South Australia, Adelaide, South Australia, Australia
Abstract
The rapid growth and widespread adoption of Artificial Intelligence (Al) highlighted the urgent need for
higher education institutions to reform how ethics in computing, data science, and related fields are taught.
While computing and professional ethics are typically included in the curriculum, they are often presented
from a social science, legal, or philosophical perspective. This paper presents a case study on integrating
responsible Al principles into an asynchronous online advanced data analytics course from a technical
perspective. The study found that embedding ethics throughout the curriculum using experiential learning
and applied ethics was effective in fostering students' comprehension and application of responsible Al
concepts. However, challenges included finding relevant case scenarios, lack of expertise among teachers,
and developing suitable activities and assessments. Best practices identified include utilizing real-world case
studies, implementing hands-on ethical coding exercises, and adopting interdisciplinary approaches. Lessons
learned emphasize the importance of timing, practical application, and flexible curriculum design. This
approach enabled students to assess and plan for the human consequences of Al applications, and design
and implement risk mitigation strategies. The study represents a step forward in preparing students for
ethical challenges in Al, while highlighting areas for future work, including teacher training and curriculum-
wide integration of responsible Al principles.
Keywords
Responsible Al in Education, Ethics in Computing Education, Principles of Responsible Al
1. I n t r o d u c t i o n
Teaching ethics in computing-related degrees are often required by accrediting bodies or board with
oversight of the profession. International standards such as ABET’s accreditation standard [1] and
ACM’s Code of Ethics [2], and Australia’s ACS Code of Ethics [3] lay the foundation of ethics
education here in Australia. The increasing permeation of Artificial Intelligence (Al) systems lead to
the concerns in the use of data, lack of trust and challenges with the use of Al. Governments and
major players in the field have all started specifically looking at their practices and evaluate how they
can scale up their use of Al systems and at the same time minimize Al risks. Every Al, data or related
professionals has a responsibility to understand the social, political, and ethical consequences of their
work. The higher education sector who produces these specialists who are more likely to use Al
algorithms are compelled to rethink how they can teach the responsible use of Al. A survey made by
[25] looked at machine learning courses and found that students were not taught ethics and if they
were, students enroll in a stand-alone ethics elective course. Most approaches to ethics education are
case-based teaching where students are presented with cases, and they respond by discussing their
approach and decisions using existing ethical frameworks [6] [7] [30]. Although this approach seems
to be successful for some professions, education in Artificial Intelligence-related courses struggle to
provide sound training in ethics that are essential for them to be successful in their
career. Frauenberger, Rauhala, & Fitzpatrick [16] argued that ethical concerns are still managed in
the mindset of past paradigms that largely remain static and have determined outcomes. It is only in
recent years that ethical topics are integrated and infused in computing related curricula [15] [18],
7th Conference on Technology Ethics (TETHICS2024), November 6-7, 2024, Tampere, Finland
Q rhoda.abadia@unisa.cdu.au
O 0000-0002-8265-0503
© 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
This paper presents the experiences in integrating the responsible use of Al into an upper-level
undergraduate data analytics course. Specifically, the following are the objectives of this study:
1. Describe how the course was designed to embed the principles of responsible Al in an
asynchronous online course.
2. Discuss challenges faced in the course design.
3. Present best practices and lessons learned that can be used to replicate the success of the case.
The technical practical (applied) ethics approach was employed to address ethical issues in Al
education. An investigation of key ethical concerns and their application within the technical
practical activities proved conducive to ethical reasoning, as these activities were specifically
designed to align with the subject matter of the ethical inquiry. The curriculum emphasized topics
highlighted by the ACM Conference on Fairness, Accountability and Transparency (ACM FAccT).
Moreover, the Principles of Responsible Al, as established by industry leaders such as Microsoft [26],
Google [4], IBM [27], and Meta [13] were adapted in for the course. The UNESCO recommendation
on Ethics of Artificial Intelligence [45] and European Commission ethics guideline for trustworthy
Al [46], who were also used as a guide in this course, emphasize similar core principles such as
fairness, transparency, privacy, safety, and human oversight. They also stress the importance of
accountability and the need for Al to benefit society.
2. T h e o r e t i c a l B a c k g r o u n d
This section presents the relevant literatures, theories and frameworks that were used in the
development of the course.
2.1. Principles of Responsible Al
The term responsible Al encompasses a variety of technical, legal, and ethical considerations that
apply not only to Al but also to data analytics and data science. The applications of Al have grown
exponentially throughout the years, and current laws, policies, and standards have not caught up
with the unique challenges and risks that Al poses and the changing ways society is adopting these
technologies. This is why there is a need for comprehensive guidelines. While there is increasing
pressure on organizations to scale up their use of Al, there is also a growing demand for the
responsible use of AL
Several big organizations and international bodies have identified principles that guide the
development and use of Al applications. UNESCO's Recommendation on the Ethics of Artificial
Intelligence [45] outlines ten key principles: Proportionality and Do No Harm, Safety and Security,
Fairness and Non-Discrimination, Sustainability, Right to Privacy and Data Protection, Human
Oversight and Determination, Transparency and Explainability, Responsibility and Accountability,
Awareness and Literacy, and Multi-stakeholder and Adaptive Governance. Similarly, the European
Commission's Ethics Guidelines for Trustworthy Artificial Intelligence [246 propose seven key
requirements: Human Agency and Oversight, Technical Robustness and Safety, Privacy and Data
Governance, Transparency, Diversity, Non-discrimination and Fairness, Societal and Environmental
Well-being, and Accountability. These frameworks demonstrate a growing consensus on the
fundamental principles of responsible Al development and deployment.
Major tech companies have also developed their own frameworks. Microsoft defines responsible
Al as an "advancement of Al driven by ethical principles that put people first; and making sure Al
systems are developed responsibly and in ways that warrant people's trust" [26], Accenture defines
responsible Al as "the practice of designing, developing, and deploying Al with good intention to
empower employees and businesses, and fairly impact customers and society" [28]. Google Al [4],
IBM [27], Microsoft [26], and Meta [13] have common principles that can be classified into five
categories: fairness and inclusion, transparency and explainability, reliability and safety, privacy and
security, and accountability.
Fairness refers to the lack of bias. Including and involving diverse people creates data diversity
which helps prevent bias. Transparency and explainability refer to the need for Al systems and their
related components to be understandable, explainable, and interpretable. Reliability and safety refer
to Al systems being built and tested for safety, performing what they were originally designed to do,
responding safely to new situations, and resisting unintended manipulation. Privacy and security
principles include the incorporation of privacy and security in the design of Al systems. The goal is
to ensure that all data, whether personal and/or sensitive, should be used ethically in all Al systems.
Lastly, the accountability principle refers to making people who design, develop, and deploy Al
systems accountable for the impact of actions and decisions of the technology. This includes making
users of the system accountable for how they use the systems.
A recent big player in the development of artificial general intelligence (AGI) applications is
OpenAI. OpenAI's principles, however, do not seem to address most of the concerns raised in the
responsible use of Al [29]. OpenAI's principles focus on broadly distributed benefits, long-term safety,
technical leadership, and cooperative orientation.
While there are some variations in emphasis and terminology, these various frameworks and
guidelines share common themes: algorithmic fairness and diversity, reliability and safety, privacy
and security, and transparency and accountability. Thes principles form the foundation for the ethical
development and deployment of Al systems across various domains and applications.
2.2. Teaching Ethics
Teaching ethics can be classified as theoretical and applied ethics to distinguish where is the focus of the
ethical investigations [16]. Theoretical ethics often concerns itself with the understanding of the nature,
language, and reasoning in ethics. Applied ethics is the practical application of the ethical theory to a
problem specific to the field in question.
There are several best practices in teaching ethics in law, health, engineering, Information
Technology, and business. Azim & Shamim [5] looked at educational theories that inform the
education strategies for teaching ethics in undergraduate medical education and found that reflection,
constructivist, and experiential learning theories are best suited in guiding strategies in teaching
ethics. A European survey that examined how computer ethics are taught in computer science, or
related degrees found that 66 % of the universities teach ethics as part of their computing degree [33],
The case method seems to be the common approach particularly in the field of computing, science,
and engineering [6] [17] [32]. There are a variety of cases used in the case method, some are used
independent of the others while some can be a combination of the others. Cases can be classified as
narrative vs dialogue, single perspective vs multi perspective, hypothetical vs actual, stories vs
problems, view as reader a participant vs an outside judge; success vs positive, single issue vs multi-
issue, single stave vs multi-stage, ordinary vs technical language, personal vs policy, living case vs
published cases [11]. All of these can be used in creating case studies for teaching ethics, it is highly
suggested that technical courses use technical language where students have the same technical
training or background as the instructor. Stratton [34] emphasized that moral judgement, as a skill,
must be practiced and simulations provide the students-controlled environment.
Stavrakakis etal [32] and Lewis & Stoyanovich [40] argued that traditional ethics education for
computing and/or data related curriculum are often taught separately as an ethics course and may
not include practical and timely training on how to weigh the consequences that can be applied in
their profession. Skirpan et al. [32] use of ethical thinking throughout the process of learning the
fundamentals of human-centered design while Lewis & Stoyanovich [40] used algorithmic
development to teach ethical data science. Both studies have shown increased interests among
students. In their study of incorporating social issues of computing in a liberal art setting, Davis &
Walker [12] identified ways social issues can be addressed in the technical topics and summative
assessments. The challenges they identified is that framing an exam question related to social issues
can seem awkward and students raise issues that have been raised in other courses. Lack of staff
availability and expertise was cited as a reason for not teaching computing ethics [33] while Wueste
[37] argued that curricular time demands in science and engineering disciplines are major obstacles
in successfully integrating ethics and calls a need for professional development for teachers.
In traditional ethics education for computing, the case method approach, as demonstrated in
related studies, is often presented as a distinct course or topic within a curriculum, lacking practical
training on reflective application in a professional context. This study aims to bridge this gap by
integrating case studies, practical applications, and ethical considerations into the technical content
of the course. By intertwining technical skills and knowledge acquisition with a heightened
awareness of ethical implications, students are provided with a comprehensive learning experience.
The objective is to enhance ethics education by offering a pedagogical approach that effectively
delivers the learning outcomes encompassing both technical proficiency and responsible Al
applications.
2.3. Asynchronous Online Course Design
Asynchronous online is a kind of online learning where students are allowed to study at their own
time. The communication of teaching and learning does not happen at the same [38]. Although
studies have shown that a well-designed course can increase student satisfaction and their learning
experiences, there is no consensus on what the guidelines of a good online course design are [42], A
common approach is to use an online course design template or a checklist to provide consistency to
students in both accessing and navigating the course site, and assists teachers in saving time, reducing
cognitive load, and meet compliance requirements [21][43]. All course content design require a
systematic approach whether taught online or face-to-face. Several studies report that a systematic
approach in course content design and they focus on alginment of course learning outcomes, activities
and assessments [39][41] [42]. Having a a strong online course objectives and the selection of
teaching methods to achieve them is important in an online course design. Wankel [36] suggested
that the course objectives is the key to a successfully online course design and that the design process
can be structured to four essential activities: sharing of information, illustrate skills, guide practice
of skills and ensure that learning occurred.
When sharing information (deliver contents), teachers have variety of choices and are not only
limited to text-based approach. They can curate existing digital resources, use audio/video recordings,
conferencing systems, web-based and learning management system tools or even immersive
technologies (e.g., virtual reality). One strategy employed to increase accessibility [23] and teacher
presence is the use of videos. There are several factors that were considered in the design of videos
including how information should be presented and how it can be supported by providing additional
learning activities [44], These activities may involved additional tasks that a student may peform to
reinforce what has been learned in the videos. Of all the different learning tasks, authentic activities
in online learning has shown many opportunities to increase learning. Authentic activites involve
presenting students with complex and extended cae scenarios that allow them to fully engage in
problem-solving within realistic situations that closely resemble the context where the knowledge
they are acquiring can be practically applied [19],
3. Embedding Principles of Responsible Al in Asynchronous Online
Course
Working on this theoretical background, a course was developed that integrates case studies, practical
applications, and ethical considerations into the technical content of the course. The common
principles of responsible Al were identified. Among the different principles, Microsoft’s grouping of
the Principles of Responsible Al [26] were adapted as it covers all the principles identified by different
Al organizations. The principles were classified as algorithmic fairness and diversity, reliability and
safety, privacy and security, and transparency and accountability.
The technical practical (applied) ethics approach in teaching ethical issues in Al was used. The
course was designed such that the principles of responsible Al are embedded into experiential
learning. Experiential learning theory as synthesized by Kolb [22] defines learning as “a process
where knowledge is created through the transformation of experience”. The experience cycles of
discussion, feedback, and practice and application in real-life context helps students apply and
connect theoretical knowledge with real-life applications. This experience cycle is repeated
throughout the course and was used as the basis of the design of this course. Figure 1 shows the
framework that was used in designing the online course activities.
Figure 1: Integrating Technical (Applied) Responsible Al in the Online Course Design
Course Information Assessment
Course Content and
Instructional Methods
Learning •Assessment
Assessment
Outcomes timelines Topic and subtopics
subtopics
Structure •Assessment
Assessment Content presentations using different T
Learning descriptions instructional strategies (e.g. content
Resources •Assessment
Assessment videos, reading materials, online video
Teaching Team instructions conferencing, code along activities)
Experience cycle
Communications •Assessment
Assessment Practical/Coding Applications
Applications of
(email, forums, criteria Responsible AIAl
video •Marking
Marking rubrics Relevant case scenarios with
conferencing) practical/coding applications
Reflections through discussion forums A
Student-led discussions reflecting on the
the\i
results of their practical/coding activities
The course information contains an overview of the course structure, requirements, learning
outcomes, and resources. A separate section is dedicated the assessments in the course. One of the
learning outcomes specifically states the knowledge, competencies and skills that are expected of the
student to acquire related to Responsible Al. The assessment methods are mapped to these learning
outcomes. The course is divided into unit topics. In the case study, the topics are divided into weekly
topics. The cycle is students learn about the topic followed by practical activities that are related to
the topic. The practical activity often includes coding activities where students apply the principles
of responsible Al in the case scenarios using coding. This activity is often followed by discussions
where students share their reflection on the case scenarios.
3.1 . Case Study: Teaching Responsible Al in an Advanced Data Analytics Course
The course is an undergraduate-level advanced courses where data analytics students are introduced to
Al topics such as reinforcement learning, computer vision and natural language processing. Students who
are enrolled in this course have the basic skills and knowledge in machine learning, specifically, artificial
neural networks and intermediate programming skills in Python. While students learn the technical
theories and applications of Al, the use of responsible Al is emphasized throughout the course. The course
was designed not only to help students assess and plan the human consequences of deploying the Al
applications, but they also get to design and implement changes to mitigate or lower the risks associated
in using these applications.
It was purposely decided that the principles of responsible Al will be the focus of the advanced
data analytics courses where the technical practical ethics approach in teaching ethical issues in Al will
be used. When students enroll in the course, they have a certain “maturity” in coding, they are no longer
taught how to code. The expectation is that even with new code libraries introduced in the coding
activities, students can learn and understand on their own. This is the reason why it was decided to embed
the principles of responsible Al at the later part of the degree, so that students focus their attention on
applying the principles of responsible Al in the topics that they are learning instead of learning how to
code. Attention to the learned material is considered an important factor that influence learning [9].
3.1.1. Practical Ethics Applications
The course integrates responsible Al principles into data analytics education through a set of learning
goals, formative activities, and assessments. Students are tasked with understanding and applying
responsible Al principles, critically evaluating ethical issues in data and algorithms, designing
mitigation strategies for AI-related risks, and effectively communicating ethical considerations. The
curriculum employs a variety of formative activities, including code-along exercises, which include
case study analyses and iterative problem-solving. All designed to reinforce responsible Al principles
using a spaced repetition approach [10]. For instance, students analyze real-world scenarios like bias
in credit scoring models [20], healthcare bias and data privacy issues [14], linkage attacks privacy
issues [24], and data misrepresentation in public records [8], They engage in critical code analysis,
examining transparency, fairness, and privacy aspects of Al systems. Assessments are multi-faceted,
comprising ethical impact analyses of real-world Al systems, practical implementation projects
incorporating responsible Al principles, peer reviews of ethical reasoning, and reflective journaling.
This approach, grounded in experiential learning theory [22], ensures students can bridge the gap
between abstract ethical concepts and real-world applications in data analytics and Al development.
By continually reflecting on the ethical implications of their work and iterating on solutions, students
develop a robust understanding of how to apply responsible Al principles in practice, preparing them
for the ethical challenges they'll face in their professional careers.
An example formative activity is the code-along activities (Figure 2). Students participate in
guided coding exercises that incorporate responsible Al principles.
Figure 2. Example Code-Along Activity.
/ Learning Activity 2.6 - Code Along Activity Part 1: Mitigating Bias using IBM AIF36O
Practical Activity Part 1
We will be using Python using Coogle Colab as a platform and IBM Al Fairness 360 toolkit. IBM AIF36O uses techniques to help detect and mitigate discrimination and bias in machine learning models.
How
1. Read the data description to have an insight into the features collected in the data set (link: CermanDataSetDescription.pdf)
2. Follow the Instructions (link: Wk2_6_CodeAlongActivity_Part1RemovingBlas.html)
3. In this notebook, you need to do the following:
• Install and import packages and modules
• Use AIF 360 toolkit to assess and mitigate bias using in-processing (during data processing)
• The data set that we will be using can be accessed via the AIF36O toolkit
The Wk2_6_CodeAlongActivity_PartlRemovingBias.html in the activity contains the details of the
walk-through of the code. In this example, students write a code and perform a critical observation
during a code-along activity. Students learn how to use the toolkit to help enforce fairness and remove
bias. Fairness metrics was used to check for bias in machine learning workflows, and bias mitigators
was used to overcome bias in the workflow to produce a fairer outcome. Students are expected to
apply their knowledge of responsible Al to assess these aspects critically.
As shown in Figure 3, students are introduced the tutorial (a), followed by sample code that they
can code-along (b), and then some parts where students have to code (c).
Figure 3(a). Extract of code-along introduction.
Code Along Activity: Detecting and Mitigating Age Bias
on Credit Decisions
The goal of this tutorial is to introduce the basic functionality of Al Fairness 360, an open source toolkit
developed by IBM for bias mitigation.
This example is adapted from
:httpst//nbviewer.org/github/IBM/AIF360/blob/master/examples/tutorial_credit_scoring.ipynb
Biases and Machine Learning
In your predictive and machine learning courses, you learned how to create models to predict an o u t c o m e given
a particular instance. For example, given an instance of a demographics, we may use a model to predict whether
the applicant will buy or not; or whether the person will default in a home loan or not. The model makes
predictions based on a training dataset, and observed (target) outcomes. A machine learning algorithm or
predictive models will attempt to find patterns, or generalizations, in the training dataset to use when a
prediction f o r a new instance is needed.
Figure 3(b). Extract of code-along instructions provided.
Step 1: Install the libraries
In l 1: “>pip i n s t a l l numpy m a t p l o t l i b seaborn
! pip i n s t a l l numba==0.48
! pip i n s t a l l aif360==0 . 2 . 2
! python -m pip i n s t a l l B l a c k B o x A u d i t i n g
I pip i n s t a l l t e n s o r f l o w = l . 13 . 1
Step 2: Import all necessary packages
We will also import the GernnanDataset that is part of aif360.datasets. A description separate description of the
data set is provided in the practical activity.
In [ ]: # import a l l necessary packages
import sys
import numpy as np
from a i f 360. d a t a s e t s import GermanDataset
from a i f 360. m e t r i c s import B i n a r y L a b e l D a t a s e t M e t r i c
from a i f 360. a l g o r i t h m s . i n p r o c e s s i n g import A d v e r s a r i aI D e b i a s i n g
from a i f 3 6 0 . e x p l a i n e r s import M e t r i c T e x t E x p l a i n e r , Met r i e l SONExplainer
Figure 3(c). Extract of code-along activity where students were asked to reflect and code.
p r i n t ( " K e y : " , d a t a _ o r i g i n a l . metadata [ ' label_maps ' ] )
df [ ‘ c r e d i t ' ] . v a l u e _ c o u n t s ( ) . p l o t ( k i n d = ' bar ‘ )
p i t . x l a b e l C ' C r e d i t ( 1 = Good C r e d i t , 2 = Bad C r e d i t ) " )
pit .ylabel("Frequency" )
Take a minute to explore the relationship between age category and credit. Is the mean of credit different for
young and older people? What does the difference in means indicate?
# STUDENT CELL
# w r i t e code to check the mean c r e d i t score by age category
After completing this activity, students reflect on and share this in the discussion forums:
•• Ethical implications of bias in credit scoring
•• Trade-offs between fairness and model performance
•• Potential societal impacts of such models
The above example is the typical format of practical activities. Students are expected to repeatedly
apply reflections on their comprehension of the data and the algorithmic fairness and diversity in
data processing. Afterward, they reflect, plan, and utilize other principles of responsible Al that relate
to the current topic. At the conclusion of each activity, students are requested to reflect on and discuss
their feedback regarding some or all the principles, depending on their applicability. For instance, the
reinforcement learning topic placed particular emphasis on transparency in addition to privacy and
fairness. In contrast, discussion and reflections in the computer vision topic encompassed a broader
range of principles, including reliability and safety.
The experiential learning concept was also adapted in the design of the assessment instruments.
The criteria used for assessing students align with the course learning outcomes. Students are
assessed in the different data analytics and Al concepts introduced in the courses. In each of these
concepts, the assessments have the following format: a real-life case scenario is provided where
students will be asked to write and submit a code and report. In the report, students are asked to
explain the problem, discuss their approach to the problem, and discuss strategies in implementing
the principles of responsible AL The assessment was intentionally designed to require students to
draw upon their understanding of abstract concepts in principles of responsible Al and apply that
knowledge to their concrete experiences working with data, designing solutions, and engaging in
actual coding. Additionally, students were tasked with resolving any conflicts they encountered
between what they observed in the learning activities (e.g., code-along activities) and what they were
doing (i.e., designing and coding solutions to problems). While fairness, diversity and privacy
principles were integral components in all assessments) different principles were applied in various
assessments. Figure 4 illustrates example instructions for assessing students' understanding and
application of transparency and accountability in machine learning. In this assessment, students were
presented with a case study scenario and tasked with designing and developing an optimized
reinforcement learning model. Following this, they were required to discuss how their chosen
reinforcement learning approach addresses principles of Responsible Al, specifically focusing on
transparency and accountability. This exercise aimed to evaluate students' ability to not only
implement advanced machine learning techniques but also to critically consider the ethical
implications of their work within the framework of Responsible Al.
Figure 4. Example assessment instruction that focusses on the transparency and accountability
principles.
5. Transparency and Accountability
1. Discuss how the model addresses the transparency and accountability principles of responsible Al. For example,
you may want to consider the following questions:
1. Assuming that bias has been removed from the environment where the model learns, is the model
designed to prevent bias to be introduced?
2. Is the model transparent?
3. Is it clear who will be accountable if bias is introduced?
2. If the model is not addressing these issues, suggest steps that should be taken to address them in future
developments of the recommender system.
4. Challenges
Integrating responsible Al principles into Al courses presents a multifaceted set of challenges that
educators must navigate carefully. A primary concern is finding the right balance between technical
content and ethical considerations. While it is crucial to prepare students with Al skills, there's an
equally pressing need to instill a deep understanding of the ethical implications of these technologies.
For instance, when teaching reinforcement learning, teachers must not only cover complex
algorithms but also discuss potential biases in reward functions and their societal impacts. To address
these issues, we integrated ethical discussions directly into technical topics using real-world
examples, developed case studies combining technical implementation with ethical analysis, and
implemented project-based learning in assessments.
Another significant challenge lies in maintaining the relevance of case studies and ethical scenarios
in a rapidly evolving field. An example that was pertinent last year, such as facial recognition in
public spaces, might be superseded by more pressing concerns like AI-generated deepfakes in political
campaigns. We continue to search for current news articles to maintain relevance, though a more
scalable approach is needed.
Assessing ethical reasoning poses its own set of difficulties. Unlike technical skills that can often
be evaluated through quantitative metrics, judging the quality of ethical decision-making requires
multifaceted criteria. Educators might struggle to develop rubrics that objectively measure a student's
ability to identify and reason through ethical dilemmas in Al development. Assessing ethical
reasoning posed unique difficulties, which we tackled by developing rubrics focused on the process
of ethical decision-making.
Student resistance can also be a hurdle, as some may perceive ethical training as less valuable than
technical prowess in the job market. This attitude might be reinforced by the tech industry's historical
focus on innovation over ethical considerations, though this is gradually changing. To combat student
resistance and demonstrate the importance of ethical skills, we highlighted job postings specifically
mentioning ethical Al requirements.
Instructor expertise presents another challenge. Many Al professors come from technical
backgrounds and may not feel equipped to lead discussions on complex ethical issues. Conversely,
ethics professors might struggle with the technical intricacies of advanced Al systems. This gap
necessitates either extensive cross-training or collaborative teaching models. The expertise gap
among instructors was currently addressed by associating IT instructors with professional bodies
offering ethics-related development opportunities. Finally, time constraints in already packed
curricula can make it difficult to give both technical and ethical aspects their due attention. A course
on natural language processing, for example, must cover a vast array of algorithms and techniques,
leaving little room for in-depth discussions on the ethical implications of language models in areas
like content moderation or automated customer service. While we've made significant progress, we
recognize that this is an evolving process requiring ongoing adaptation. Addressing these challenges
requires a thoughtful, multidisciplinary approach to curriculum design and a commitment to ongoing
adaptation as the field of Al continues to advance and evolve.
5. Best Practices
Implementing best practices for integrating responsible Al principles into Al education also requires
a thoughtful approach. Rather than treating ethics as a standalone topic, it's crucial to embed these
principles throughout the entire curriculum. For instance, when teaching machine learning
algorithms, teachers can consistently highlight potential biases in data sets and discuss the ethical
implications of model choices. This integrated approach ensures that students view ethical
considerations as an inherent part of Al development rather than an afterthought.
Utilizing real-world case studies is another vital strategy. For example, teachers might analyze the
ethical concerns surrounding OpenAI's GPT models, discussing issues like potential misuse for
disinformation, copyright infringement, and the amplification of biases. Such current and relevant
examples make ethical dilemmas tangible and demonstrate their immediate relevance to the field.
Hands-on ethical coding exercises further reinforce these principles. Students might be tasked with
implementing fairness constraints in a credit scoring algorithm or designing transparency measures
for a recommendation system, thereby gaining practical experience in translating ethical principles
into code.
Collaborative learning plays a crucial role in developing a well-rounded understanding of Al
ethics. Group projects could involve designing an Al system for a sensitive application, such as
healthcare diagnostics, requiring students to collectively navigate technical challenges while
addressing ethical concerns like patient privacy and algorithmic transparency. An interdisciplinary
approach, possibly involving collaboration with ethics or philosophy areas can provide deeper
insights into ethical frameworks and their application to Al. For instance, a joint seminar between
computer science and philosophy students could explore the ethical implications of autonomous
vehicles, bringing together technical knowledge and ethical reasoning. Continuous assessment of
ethical reasoning is also essential to ensure that students are internalizing these principles. This could
involve regular reflection papers on the ethical implications of topics covered in class, or project
milestones that require ethical impact assessments.
6. Lessons Learned
The experience in embedding responsible Al principles into the Al technical course has resulted to
valuable lessons that can significantly enhance the learning experience and outcomes for students.
One crucial insight is the importance of timing in introducing these concepts. By incorporating ethical
considerations into courses were students have the coding experience, students can focus on applying
these principles to complex Al systems rather than grappling with basic coding challenges. For
instance, in an intermediate level machine learning course, students might explore the ethical
implications of using Al in criminal justice predictive systems, analyzing potential biases and fairness
issues in real-world applications.
Practical application plays an important role in ethical Al education. When students can
immediately apply ethical principles to their coding projects, engagement and understanding deepen
significantly. For example, a project on developing a recommendation algorithm for a streaming
service could include requirements for transparency in the algorithm's decision-making process and
considerations for diverse representation in content suggestions. This hands-on approach allows
students to see firsthand how ethical considerations shape technical decisions.
The development of ethical reasoning skills has been observed to require consistent practice and
reflection. Regular opportunities for ethical decision-making, such as weekly case study discussions
or ethical impact assessments for each major project, help students hone their ability to identify and
navigate complex ethical dilemmas in Al. For instance, students might be asked to regularly update
an "ethical journal or discussion forum" throughout the course, reflecting on how each new Al
technique they learn could be used or misused from an ethical standpoint. In addition, relevance to
current Al developments and potential career scenarios has proven to significantly enhance student
engagement. Discussing recent controversies, such as the ethical implications of AI-generated art or
the role of large language models in spreading misinformation, helps students see the immediate
relevance of ethical considerations in their field. Another activity that has shown to be important in
teaching responsible Al is peer learning. Group discussions on ethical dilemmas often lead to rich
insights and perspectives that individual reflection might not yield.
Flexibility in curriculum design has emerged as a crucial factor, given the rapidly evolving nature
of Al and its ethical challenges. Teachers must be prepared to adapt their teaching materials to address
emerging issues, such as the ethical considerations surrounding new Al technologies. Assessment
strategies have evolved to combine evaluation of technical skills with assessment of ethical reasoning.
This might involve projects where the technical implementation is judged alongside an ethical impact
report, providing a holistic view of a student's capabilities as a responsible Al practitioner.
Emphasizing the relevance of responsible Al principles to future careers has significantly
increased student appreciation for these courses. When students understand how ethical
considerations will impact their work in industry or research, they engage more deeply with the
material. This could involve assignments that mimic real-world scenarios.
7. Conclusion and Recommendation
The rapid emergence and adoption of Artificial Intelligence technologies has led to increased
opportunities, risks, and challenges, highlighting the need to teach the principles of responsible Al.
In this paper, a case study was presented, showcasing the effective embedding of responsible Al
principles in an advanced technical online course using the theory of experiential learning and applied
ethics as a foundation. The course design incorporated practical and code-along activities that moved
ethical discussions to specific topics, emphasizing the importance of understanding and checking data
prior to modeling and evaluating models and algorithms, with the Al principles consistently
integrated. The consistent presentation and reinforcement of the principles of responsible Al through
the design of practical activities have proven beneficial in fostering students’ comprehension and
application of these concepts. It not only enhances awareness of the ethical challenges but also
facilitates the practical applications of ethical concepts. Moreover, this approach fosters a conducive
environment that stimulates critical thinking among students regarding ethical considerations and
the far-reaching implications of their professional actions.
Despite the success of the course design, challenges were identified, such as finding relevant up-
to-date case scenarios and lack of technical and responsible Al expertise among teachers. Developing
practical activities and assessments for the course presented a significant challenge for the course
designers. One of the main hurdles was finding case studies that were relevant and up to date with
the latest technology-related events, followed by the task of modifying them to include the concepts
being taught while also highlighting the principles of responsible Al. Additionally, finding suitable
data to illustrate both the concepts and principles of responsible Al was also a challenge. In many
cases, the data had to be modified to suit the purpose of the topic. Another challenge faced in the
courses design was the lack of teachers who possessed both the necessary technical expertise and an
understanding (or interest) in embedding responsible Al principles into their courses. This issue
required extensive training and support to equip the educators with the necessary skills and
knowledge to effectively integrate these principles into their teaching.
Best practices for addressing these challenges involve embedding ethics throughout the
curriculum, utilizing real-world case studies, implementing hands-on ethical coding exercises,
fostering collaborative learning, and adopting interdisciplinary approaches. Lessons learned highlight
the importance of timing in introducing ethical concepts, the value of practical application, the need
for consistent practice in ethical reasoning, the benefits of peer learning, and the necessity of flexible
curriculum design. Additionally, emphasizing the relevance of responsible Al to future careers has
been shown to increase student engagement and appreciation for these principles.
This research, while providing valuable insights into integrating responsible Al principles into
technical courses, has several limitations and areas for future work. A key limitation is the lack of
formal student feedback, which would provide crucial data on the effectiveness of the approach from
the learners' perspective. Future research should prioritize collecting and analyzing student feedback
to refine the curriculum design. Additionally, the study's scope was limited to a single course,
potentially limiting its generalizability. Future work should focus on scaling up the integration of
responsible Al principles across entire curricula, including non-AI related courses. This expansion
would require developing comprehensive teacher training programs and establishing partnerships
with industry to ensure ongoing relevance of case studies. Creating a dynamic, regularly updated
database of ethical case studies and fostering multidisciplinary collaborations between computer
science and philosophy departments could further enhance the approach.
Declaration on Generative Al
The author has not employed any generative Al tools.
References
[1] ABET, “Criteria for accrediting computing programs 2023-2024,” 2023, URL:
https://www.abet.org/accreditation/accreditation-criteria/criteria-for-accrediting-computing-
programs-2023-2024/
[2] Association of Computing Machinery, “ACM Code of Ethics and Professional Conduct,”
version 1,2023, URL: https://www.acm.org/code-of-ethics
[3] Australian Computer Society, “ACS Code of Ethics,” 2014, URL:
https://www.acs.org.au/content/dam/acs/rules-and-regulations/Code-of-Ethics.pdf
[4] Google, “Artificial Intelligence at Google: Our Principles,” version 11, 2022, URL:
https:// ai.google/principles/
[5] D. Azim and M. S. Shamim, “Educational theories that inform the educational strategies for
teaching ethics in undergraduate medical education,” Journal of the Pakistan Medical
Association, 70.1(2020): 123-128.
[6] B. Blundell, “Ethics in Computing, Science, and Engineering: A Student's Guide to Doing
Things Right,” 2020, Springer Nature, Switzerland.
[7] F. Bott, “Professional Issues in Information Technology,” 2nd ed. (2014), British Computing
Society.
[8] City of New York, "DOHMH New York City Restaurant Inspection Results," NYC OpenData.
URL: https://data.cityofnewyork.us/Health/DOHMH-New-York-City-Restaurant-Inspection-
Resuits/ 43nn-pn8j.
[9] M. M. Chun and N. B. Turk-Browne, "Interactions between attention and memory," Current
Opinion in Neurobiology, vol. 17 2 (2007):177-184, doi: 10.1016/j.conb.2007.03.005
[10] D. Ausubel and M. Youssef, "The Effect of Spaced Repetition on Meaningful Retention," The
Journal of General Psychology, 73.1 (1965): 147-150, URL:
https://doi.org/10.1080/00221309.1965.9711263.
[11] J. Davis and H. M. Walker, "Incorporating Social Issues of Computing in a Small, Liberal Arts
College: A Case Study," in Proc. 42nd ACM Technical Symposium on Computer Science
Education, 2011, pp. 69-74.
[12] M. Davis, Ethics and the University, Taylor and Francis, 2022.
[13] Meta Al, “Facebook’s five pillars of responsible Al,” version 11, 2022, URL:
https://ai.facebook.com/blog/facebooks-five-pillars-of-responsible-ai.
[14] Fairlearn, “Fairlearn: A Python package to assess and improve fairness in Al,” version 11, 2022,
URL: https://fairlearn.org.Ferrarello, L. (2019).
[15] L. Ferrarello, "Social awareness in design & engineering education and practice: the value of
ethics in postgraduate education," in Proc. 21st International Conference on Engineering &
Product Design Education, Glasgow, 2019, doi:10.35199/epde2019.83.
[16] C. Frauenberger, M. Rauhala, and G. Fitzpatrick, “In-Action Ethics, Interacting with
Computers,” 29.2 (2016): 220-236, doi: 10.1093/iwc/iww024.
[17] K. S. Hansen, "Literature for Specific Purposes: A Literary Approach to Teaching Ethics in
Science and Technology," Configurations (Baltimore, Md.), 26.3 (2018):337-343, doi:
10.1353/con.2018.0030.
[18] A. Harris, M. Lang, D. Yates, and S. E. Kruck, "Incorporating Ethics and Social Responsibility
in IS education," Journal of Information Systems Education, 2.3 (201 l):183-190.
[19] J. Herrington, R. Oliver, and T. Reeves, "Patterns of engagement in authentic online learning
environments," Australian Journal of Educational Technology, 19.1 (2003): 59-71, doi:
10.14742/ajet.l701.
[20] H. Hofmann, "German Credit Risk," 2016, version 2, 2022, URL:
https://www.kaggle.com/datasets/uciml/german-credit.
[21] H. Kathuria and D. W. Becker, "Leveraging Course Quality Checklist to Improve Online
Courses," Journal of Teaching and Learning with Technology, 10.1 (2021), doi:
10.14434/jotlt.vlOil.31253.
[22] D. A. Kolb, "Part 1: Experience and Learning,” Experiential Learning: Experience as the Source
of Learning and Development, 2nd ed., Pearson Education, 2014.
[23] S. Kolber, "Enhancing accessibility with instructional video," Teaching Online Master Class,
2020.
[24] P. Ohm, "Your secrets live online in databases of ruin," Ars Technica, (2009). URL:
https://arstechnica.com/tech-policy/2009/09/your-secrets-live-online-in-databases-of-ruin/.
[25] J. Saltz, M Skirpan, C Fiesler, M Gorelick, T Yeh, R Heckman, N Dewar, Integrating Ethics within
Machine Learning Courses, ACM Transaction on Computing Education, 19.4 (2019) 1-26.
doi:10.1145/1188913.1188915.
[26] Microsoft, Microsoft Responsible Al Standard, v2, version 11, 2022. URL:
https://www.microsoft.com/en-us/ai/responsible-ai
[27] IBM, Al Ethics, version 11, 2022. URL: https://www.ibm.com/impact/ai-ethics
[28] Accenture, Responsible Al : Al Ethics & Governance, version 11, 2022, URL:
https://www.accenture.com/au-en/services/applied-intelligence/ai-ethics-governance
[29] Open Al, OpenAI Charter, version 10,2022, URL: https://openai.com/charter/
[30] D-W. Olayinka, Teaching Ethics in Decision Making: Embedding Moral Reasoning in the
Management of Information Systems, Teaching Ethics Across the Management Curriculum, K.
Ogunyemi, Ed. New York: Business Expert Press, 2014, pp. 179-197.
[31] A. Scott and S. Barlowe, "How software works: Computational thinking and ethics before CS1,"
in Proc. IEEE Frontiers in Education Conference (FIE), 2016, pp. 1-9, doi:
10.1109/FIE. 2016.7757436.
[32] M. Skirpan et al., "Ethics Education in Context: A Case Study of Novel Ethics Activities for the
CS Classroom," in Proc, of the 49th ACM Technical Symposium on Computer Science
Education, 2018, pp. 940-945.
[33] I. Stavrakakis, D. Gordon, B. Tierney, A. Becevel., E., Murphy, G. Didg-Crnkociv, D. O’Sullivan,
"The teaching of computer ethics on computer science and related degree programmes: A
European Survey," Int. J. Ethics Educ., 7.1 (2022): 101-129, doi: 10.1007/s40889-021-00135-l.
[34] T. Stratton, "Incorporating ethics content throughout an integrated pharmacy curriculum," Am.
J. Pharm. Educ., 84.11, 2020, doi: 10.5688/ajpe7865.
[35] SUNY Online, "The SUNY Online Course Quality Review Rubric (OSCQR)," version 09, 2022,
URL: https://oscqr.suny.edu/.
[36] C. Wankel, "Design considerations," in Redesigning Courses for Online Delivery Design,
Interaction, Media, and Evaluation, vol. 8, R. Parker, Ed. Emerald Group Publishing Limited,
2013, doi: 10.1108/S2044-9968(2013)0000008003.
[37] D. E. Wueste and D. Scibilia, "Cognition and Conation: A Potent Alliance in Teaching Ethics,"
in Instructional Models, Methods, and Modalities for University Studies, pp. 5-20, Rowman &
Littlefield Publishers, 2021.
[38] R. Zhang, N. C. Bi, and T. Mercado, "Do zoom meetings really help? A comparative analysis of
synchronous and asynchronous online learning during Covid-19 pandemic," Journal of
Computer Assisted Learning, 9.1 (2023): 210-217, 2023, doi: 10.1111/jcal.l2740.
[39] N. Lee and S. J. Kim, "A Systematic Course Design Approach to Guide the Development of a
Construction Engineering and Management Capstone Course," in 2020 ASEE-NE Annual
Conference, pp. 1-5, doi: 10.1109/ASEENE51624.2020.9292652.
[40] A. Lewis and J. Stoyanovich, "Teaching Responsible Data Science: Charting New Pedagogical
Territory," Int. J. Artificial Intelligence in Education, doi: 10.1007/s40593-021-00241-7, 2022.
[41] J. Macdonald and A. Campbell, "Demonstrating online teaching in the disciplines: A systematic
approach to activity design for online synchronous tuition," Br. J. Educ. Technol., 43.6, (2012):
83-891, doi: 10.1111/j.1467-8535.2011.01238. x.
[42] F. Martin, A. Ritzhaupt, S. Kumar, and K. Budhrani, "Award-winning faculty online teaching
practices: Course design, assessment and evaluation, and facilitation," Internet High. Educ., 42,
(2019): 34-43, doi: 10.1016/j.iheduc.2019.04.003.
[43] S. J. McGhan, C. M. Jackson, and K. Premer, "Online course quality assurance: Development of
a quality checklist," InSight: A J. Sch. Teach., 10.1 (2015): 126-140.
[44] O. Poquet, L. Lim, N. Mirriahi, and S. Dawson, "Video and learning: A systematic review (2007-
2017)," in Proc, of the 8th Int. Conf, on Learning Analytics and Knowledge—LAK, 2018, doi:
10.1145/3170358.3170376.
[45] "Ethics of Artificial Intelligence," UNESCO.org, 2023. URL:
https://www.unesco.org/en/artificial-intelligence/recommendation-ethics.
[46] "Ethics guidelines for Trustworthy Al," Shaping Europe's digital future, 2019. URL:
https://digital-strategy.ec.europa.eu/en/library/ethics-guidelines-trustworthy-ai.