=Paper=
{{Paper
|id=Vol-3191/paper01
|storemode=property
|title=Harnessing the Power of Responsible Innovation: The Shift towards Human-centered Skills and Competences in AI Engineering
|pdfUrl=https://ceur-ws.org/Vol-3191/paper01.pdf
|volume=Vol-3191
|authors=George Sharkov,Christina Todorova,Pavel Varbanov
|dblpUrl=https://dblp.org/rec/conf/isgt2/SharkovTV22
}}
==Harnessing the Power of Responsible Innovation: The Shift towards Human-centered Skills and Competences in AI Engineering==
https://ceur-ws.org/Vol-3191/paper01.pdf
Harnessing the Power of Responsible Innovation:
The Shift towards Human-centered Skills and
Competences in AI Engineering
George Sharkov 1, 2, Christina Todorova 1, 3 and Pavel Varbanov 1, 4
1
European Software Institute – Center Eastern Europe, 111-G “Tzarigradsko Shosse”
Blvd., Sofia 1784, Bulgaria
2
Plovdiv University “Paisii Hilendarski”, 24 “Tsar Assen” Str., “Plovdiv, 4000,
Bulgaria
3
Cybersecurity Laboratory, Sofia Tech Park, Laboratory Complex, Sofia, 1784,
Bulgaria
4
Institute of ICT of the Bulgarian Academy of Sciences, 2 “Acad. G. Bonchev” Str.,
Sofia, 1113, Bulgaria
Abstract
Particularly against the backdrop of several notable incidents of abuse
and malicious use of technology, in recent years, ethical issues across
all technological fields, including artificial intelligence, have become a
prominent topic for public discussion. The European Union, with its industry
and academia at the forefront, is now viewed as a pioneer on the world stage
for the implementation of an ethical approach toward an ethically-driven,
data-empowered society, with efforts such as the High-Level Expert Group
of the EC on AI and the European Act on Artificial Intelligence (AIA), to
name a few.
The rising set of ethical standards for empowering trustworthy and technically
sound Artificial Intelligence has unavoidably resulted in an industrial
movement toward human-centered skills and competences. The industry
demands at any phase of the AI lifecycle are rapidly changing, posing
considerable difficulties to academia and its ability to quickly design and
implement an industry-relevant curriculum, encompassing both the technical
and the ethical requirements for AI professionals.
This paper will investigate this shift of demands in the context of the rising
body of legal, regulatory, and compliance frameworks, using current research
and a needs and market study done as part of HCAIM – a project aimed at
establishing a Human-Centered Master’s Program on Artificial Intelligence.
Information Systems & Grid Technologies: Fifteenth International Conference ISGT’2022, May 27–28, 2022, Sofia, Bulgaria
EMAIL: gesha@esicenter.bg (G. Sharkov); tina@esicenter.bg (C. Todorova); pavel@esicenter.bg (P. Varbanov)
ORCID: 0000-0001-5086-311X (G. Sharkov); 0000-0002-8061-2941 (C. Todorova); 0000-0002-1868-8638 (P. Varbanov)
© 2022 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
Keywords
Artificial Intelligence, skills, competences, ethics, lawfulness, trustworthy
AI, education
1. Introduction
Technology is a competitive advantage and a critical component of business
success nowadays. This phenomenon is especially valid in advanced informa-
tion societies [1] and IT-intensive sectors of the economy. Artificial Intelligence
(AI) has made quick advances in various application areas, including administra-
tive assistance, healthcare, and finance, to mention a few [2]. Simultaneously,
negative aspects of AI, including multiple privacy issues, bias in decision-making
and fatal mistakes in conclusions made by ungoverned, opaque algorithms, have
sparked public and academic interest in AI ethics [4, 5]. Thus, the current AI
boom has been accompanied by frequent demands for applied ethics, hoping to
limit the disruptive potential of AI technologies.
To address public concerns about AI’s risks and unintended consequences,
governments and international organizations such as the EU, professional asso-
ciations such as the IEEE, and numerous enterprises have shared their vision for
trustworthy AI, including principles, standards, and guidelines [3]. In the same
vein, an EU High-Level Expert Group was established in 2019 [6] to develop eth-
ical standards for AI to assure technically robust, reliable, and resilient AI devel-
opment. The Asilomar Principles of Beneficial Artificial Intelligence [7] and the
IEEE Global Initiative on Ethics [8] are among many other notable initiatives to
better understand the risks of artificial intelligence and develop legal frameworks
and codes of conduct and values-based design methods for autonomous and in-
telligent systems. Similarly, the European Commission also issued a proposal
for a European Act on Artificial Intelligence (AIA) in April 2021 [9]. The AIA
risk-based approach and the proposed legal framework for the implementation of
trustworthy AI-based systems and services in EU extend the human-centered AI
concept with technical and legal requirements depending on the application area
and associated risk levels (for humans). Furthermore, the AIA outlines the respec-
tive schemes and levels of conformity assessments and certifications to transform
all those requirements in practice.
Doubling the number of data science professionals in Europe to more than
10 million, a critical component of the European Data Strategy of 2020 [10],
combined with the growing demand for ethical compliance, poses a significant
challenge to data science and artificial intelligence. Thus, responsible data collec-
tion, processing, and storage requirements and the relevant skills, competences
and knowledge become vital to meet these market needs.
2
To that aim, professionals with a specific combination of skills, knowledge,
and competencies, combining technical competency with societal awareness and
strong ethical understanding, are required for artificial intelligence development,
implementation, and integration [11].
The purpose of this paper is to report on the main findings of a market needs
analysis conducted as part of the “Human-Centred AI Master’s Programme Pro-
ject” – a project dedicated to developing and piloting a Master’s Program in Eth-
ical AI at four European universities: Ireland, the Netherlands, Budapest, and
Italy. The analysis’s key findings provide information on business and academics’
requirements, expectations, and demands regarding AI education and AI profes-
sionals in light of this transition toward responsible AI.
Most significantly, this paper will give an aggregated report on the skills,
knowledge, and competencies recognized by industry and academic experts
across Europe that graduates and young AI professionals need to remain relevant
against the rising demands of the AI field.
2. Steering trustworthy artificial intelligence in Europe
Without trust, the artificial intelligence system may hurt people unintentional-
ly, with the prospect of a range of undesirable outcomes; obstructing the fulfilment
of the enormous social and economic advantages, that artificial intelligence might
bring to the European digital market. Confidence in the creation, deployment and
usage of artificial intelligence systems relates to the intrinsic features of the technol-
ogy and the quality of the socio-technical systems involved in AI applications [6].
With AI progressively being incorporated into areas such as aviation, nu-
clear energy, and defense, the trustworthiness of artificial intelligence systems is
no longer a feature but a must for market entry. As a result, striving for reliable
artificial intelligence requires the reliability of the artificial intelligence system
itself and a holistic and systematic approach that includes the reliability of all par-
ticipants and processes that are part of the system’s social-technical background.
The AI-HLEG [6] specifies three critical components of trustworthy AI that
must be satisfied throughout the system’s life cycle:
• lawful, complying with all applicable laws and regulations;
• ethical, ensuring adherence to ethical principles and values;
• robust and secure, both from a technical and social perspective since,
even with good intentions, AI systems can cause unintentional harm;
Each of the components above is required but insufficient to ensure AI’s
trustworthiness. In an ideal world, these three functions would be synchronized
and overlapping.
Our collective and individual obligation as a society is to work hard to guar-
antee that all three components contribute to the security of dependable artificial
3
intelligence. A reliable strategy is essential for establishing “responsible com-
petitiveness”. It creates a foundation for all persons affected by artificial intel-
ligence systems to guarantee that their design, development, and usage are legal
and ethical.
Following a series of other initiatives in the field of Artificial Intelligence
governance and standardization on a global, European, and national level, the Eu-
ropean Commission published the proposal for the new EU Artificial Intelligence
Act (AIA) [9], which is one of the first initiatives worldwide aimed at providing a
legal framework for AI. It adheres to the risk-based approach to AI development,
implementation, and usage that the Commission outlined earlier in 2020 with the
White paper on AI [12], while also extensively using the AI-findings HLEG and
recommendations on trustworthy AI. As part of AIA, appointing national bodies
in charge of standards and oversight of AI development has emerged to put the
EU in a position of “leading by example” in the global arena as one of the worlds
pioneers in dependable artificial intelligence.
As a result, strengthening governance and certification for trusted artificial
intelligence systems is critical for promoting and sustaining fundamental human
rights across Europe. Only through assuring credibility can European residents,
and by extension, citizens globally, fully appreciate the benefits of artificial in-
telligence systems and be confident that efforts have been made to mitigate any
threats [3].
Considering national initiatives, among those of other European countries,
the Bulgarian government has also been part of this approach towards a human-
centered shift of AI. Bulgaria embraced Bulgarian scientists’ research as an idea
for a national AI capacity development strategy. The concept is based on two
documents, “Framework for a National Strategy for the Development of Artificial
Intelligence in Bulgaria” (2019) and “Strategy for the Development of Artificial
Intelligence in Bulgaria until 2030” [3,13] developed by a team of the Bulgarian
Academy of Sciences and a group of external experts and endorsed by the Bul-
garian Government in 2020.
The Concept for the Development of Artificial Intelligence (AI) in Bulgaria
by 2030 is consistent with current European Commission documents, strategic
initiatives, and policies, which regard AI as one of the primary drivers of digital
transformation in Europe and a critical factor in ensuring the European econo-
my’s competitiveness and high quality of life.
Furthermore, the concept defines particular features of the European vision
of “trustworthy AI”, including methods to gather accessible, high-quality data,
communicate knowledge broadly, and enable fair access to the advantages of AI
technology.
4
3. Examining the need for AI competences shift
The fast pace of AI research has resulted in international initiatives to expand
investment in technology in vital industries such as healthcare, drug testing and
discovery, transportation, cybersecurity, BFSI, education, and energy.
According to one of Grand View Research’s [14] most recent research, the
worldwide artificial intelligence market accounted for 62.35 billion USD in 2020
and is predicted to rise at a compound annual growth rate (CAGR) of 40.2 per cent
from 2021 to 2028. According to the same report, “access to historical datasets is
the essential facet accelerating the rate of innovation in the field of AI”. The last,
along with advancements in other technologies, open data policies, and data sets
provided by IoT devices and sensors, has served as a prerequisite for the develop-
ment and integration of AI solutions in a variety of sectors and disciplines.
According to Gartner [15], the commercial value of AI will reach USD 5
billion by 2025, with 23 per cent of CEOs questioned in their survey claiming
to have already adopted AI in their organizations, either as a separate product or
as an integrated solution to assist their routine business processes. Half of the
respondents emphasize assessing financial and commercial risks before imple-
menting AI initiatives. Consequently, we may deduce that the potential of AI for
company growth and innovation is immense, but a thorough risk assessment must
be undertaken before the project begins.
Similarly, North America will lead the market in 2020, accounting for around
40.4 per cent of global revenue. This high proportion may be ascribed to govern-
ment initiatives that encourage artificial intelligence (AI) in a wide range of en-
terprises and sectors. The American AI Initiative [16], the country’s plan to in-
crease artificial intelligence leadership, is one example. Federal authorities have
developed criteria for the development and real-world application of AI-based
systems across numerous industrial sectors as part of this effort, increasing public
confidence in these systems. There is a strong inclination among the interna-
tional community to examine as much as possible the dangers and consequences
of AI technology adoption, which centers mainly on the technology’s research
components, as opposed to certain favorable features of its rapid delivery to the
worldwide market.
To solve challenges like as power consumption, sluggish processing, and
inefficiency, hardware-based AI solutions are being developed. As the industry
evolves, however, new business models based on predictive, efficient automation
and scalable parallel processing capabilities will become more vital.
In recent years, few companies have invested in developing these hardware
components; however, to gain a competitive advantage, companies such as IBM
and Intel have begun producing artificial intelligence chipsets to achieve high
performance in scaling dynamic operations in parallel.
5
Several key industry participants, including Apple Inc., Google, Samsung
Electronics Co., Ltd., Baidu Inc., and Qualcomm Technologies, Inc., have devel-
oped artificial intelligence processors to handle inference workloads at the edge,
according to Grand View [14]. Amazon.com, Inc., for example, is working on
an artificial intelligence processor based on ASICs to power its Alexa personal
assistant.
The hardware sector is projected to gain another boost from technologies
such as ML, as seen by the rising demand for artificial intelligence applications.
One such example is public surveillance technologies, which have also been the
topic of several debates and discussions in recent years. According to Stanford
HAI, the technology necessary for large-scale surveillance is fast evolving, with
image classification technologies, face identification, video analysis, and speech
recognition all making substantial development in 2020.
In the midst of the COVID-19 epidemic, various governments have begun
using big data technologies to gather and monitor people’s COVID status and
quarantine compliance, raising many ethical problems.
Aside from ethical concerns, the massive demand for image recognition sys-
tems has many key obstacles that impede the industry’s development. The neces-
sity for vast volumes of data to train artificial intelligence systems for biometrics
and image recognition is one such issue. Furthermore, stacking such an enormous
volume of data improves data traceability. Artificial intelligence is used by com-
panies such as Google and Facebook in picture identification systems that need
access to massive volumes of data. However, these datasets are abundant com-
pared to data availability for AI developments in the healthcare industry, where
enormous datasets for AI solutions are still uncommon.
In healthcare, one such example is the data required to identify cancer in
scans and X-rays. The fundamental issue that emerges in artificial intelligence
due to a lack of data availability is using existing data to make appropriate judg-
ments. One cause for this lack of data is the capacity to acquire, handle, and ethi-
cally dispose of such artifacts.
According to Stanford HAI, the number of papers containing ethics-related
keywords in AI conference titles has grown since 2015. However, the average
number of article titles matching ethics-related keywords in major AI confer-
ences has remained low. Fortunately, various organizations have developed to
address the absence of ethical criteria in AI.
Against this backdrop, Europe is a part of the worldwide effort to address
the ICT skills shortage [17, 18, 19]. By 2021, the global talent shortfall will have
reached 40 million skilled employees. The worldwide talent shortfall will reach
85.2 million employees by 2030 [20]. Companies globally risk losing $8.4 tril-
lion in sales due to a shortage of competent people [20]. Computer and informa-
tion technology employees are expected to rise 11 percent from 2019 to 2029,
6
substantially faster than the overall average [20]. These vocations are expected to
generate 531,200 additional employments. The increased focus on cloud comput-
ing, extensive data collecting and storage, and information security will increase
the demand for these personnel [21].
The demand for ICT specialists in businesses grows in tandem with the de-
velopment of the digital economy. Still, factors such as the aging population,
accelerated migration processes, the difficulties of academia in keeping up with
the technological development pace, and differences in educational quality (even
across national borders) impede solutions.
The scarcity of AI-related workers in Europe may be due to a mismatch
between supply and demand for skills. The academic sector develops and pro-
duces graduates with primarily scientific knowledge in AI rather than competent
professionals capable of working on high-end projects that mix technology with
medical, mechanical engineering, biology, social sciences, and other disciplines.
Similarly, educating limited experts such as system engineers, software de-
velopers, network/system administrators, and electrical engineers complicates
the effective fulfillment of IT workers. The interdisciplinary character of current
jobs, such as data scientists, IoT engineers, cybersecurity specialists, and robot-
ics engineers, necessitates a more holistic and complicated approach to program
design in educational institutions [21]. Knowing the demands of the industry is a
keen start on developing new requirements for higher education, vocational train-
ing, and professional development.
4. Primary outcomes of the HCAIM AI market analysis
Our research began by establishing the goal of creating a Master’s Program
curriculum in human-centered AI to train ethical AI architects who perform in
keeping with the European Union’s vision of the future of AI. We used the Eu-
ropean e-Competence Framework for the technical interpretation of the profile
(e-CF) [22]. CEN (European Committee for Standardization) has created 30 ICT
profiles based on the e-CF [22]. The Human-Centered AI expert primarily fo-
cuses on the Data Scientist, Data Specialist, and Systems Architect positions.
Although the new e-CF definition includes a data scientist profile, this de-
scription lacks the competences that match specific data scientist positions
in enterprises. The e-CF3.0 integrates with the organization’s workflow, yet data
science requires connecting multiple organizational roles and departments. As a
result, the e-CF3.0 provided an overly broad framework for defining IT compe-
tencies and allocating them to data science competencies.
However, the EDISON framework [23] provides a solution. It may be viewed
as an extension of the e-CF framework, especially for the professional profile of
the data scientist [24]. This framework has been used to form the data science
7
competences of the Human-Centered AI Master’s Program. However, we had to
make a more complex decision regarding the ethical competences.
The awareness of ethical and societal factors in the professional competences
as specified in e-CF is a distinguishing feature of the human-centered AI profes-
sional profile. Only a pre-release version of the EU ICT Ethics initiative [25] was
available when this analysis was conducted. The EU ICT Ethics has compiled a
list of ethically linked essential human aspects and components, which, coupled
with e-CF knowledge, skills, and behavior, can serve as a pattern for a proactive
approach to ethics. Thus, for the study of the needs of the AI market, we took
competences related to societal, organizational and technical aspects of AI from
the e-CF and let industry and academia representatives rate their importance ac-
cording to their company’s context and experience.
The project’s assessment strategy includes categorization and mapping
methodologies executed through an online survey and focus group interviews.
The methods used allow us to assess subsequently each implementation of the
Master’s Program in sufficient depth to discover the best practices that can be
applied to improve the HCAI Master’s programs. The online survey and focus
group interviews conducted with program recipients from the sector served to
verify and validate the research implications of the previous 2-3 years.
Following best practices in educational research, it was required to design
data collecting procedures that could be deployed consistently by both experi-
enced and young researchers across numerous focus groups, nations, and organi-
zational contexts.
As a result, several data gathering technologies have been established.
Quantitative data is mainly collected using questionnaires, while qualitative
information is gathered through various methods, including facilitator-led
focus group observations and literature studies.
To shorten the time spent responding to the surveys, we used Likert ques-
tions as much as feasible. However, the survey’s broad questions enabled us to
grasp market requirements thoroughly. As a result, whereas questionnaires allow
us to obtain many surface data from many participants, we require observation
and interviews to collect more in-depth data from fewer individuals.
We must comprehend specific corporate representatives’ experiences to ap-
preciate better, why a particular topic or ability is vital to the industry. Data is
gathered from numerous sources to improve efficacy in achieving that goal.
Data mainly from the Industry Survey Questionnaire will be used for
this paper.
The Industry Survey was divided into four major categories to get a detailed
grasp of the respondents’ perspectives and profiles, as follows:
1. Definitions and Understanding. The primary purpose of this section of
the survey was to obtain a broad profile of the respondents and their grasp of
8
the many concepts in human-centered artificial intelligence.
2. Societal Aspects. This section aims to gather insight into the themes,
skills, knowledge, and competencies that respondents deem essential for
practical work on human-centered artificial intelligence initiatives.
3. Organizational Considerations. This section aims to provide a clear
picture of the relevance of numerous business and management-related sub-
jects, skills, knowledge, and competencies perceived by respondents.
4. Technological Aspects. This section of the survey intends to structure
essential technical ideas, skills, and talents in the field of human-centered
artificial intelligence based on the responses.
All competency descriptions were derived from the European e-Competence
Framework and classed as social, organizational, or technological.
The participants were provided with the competences descriptors when
answering the questions related to them and during the focus group inter-
views when those were discussed.
The goal is to present a comprehensive picture of respondents’ thoughts on
the worth of each ability in each of the three specified elements: technological,
organizational, and societal. The subsequent prioritization of the skills suggested
by the replies helped to a deeper understanding of the comprehensive and mul-
tidisciplinary nature of the AI knowledge area. The survey findings were then
crosschecked with focus group data and augmented with feedback from par-
ticipants in the focus groups.
The HCAIM Project Industry Survey was completed by an overall 49
people, representing companies from eight countries, namely:
• Ireland (17 people);
• Italy (10 people);
• Hungary (6 people);
• Bulgaria (4 people);
• Netherlands (3 people);
• USA (3 people);
• Belgium (1 person);
• Russia (1 person).
It is worth noting that one person responded that they had not yet registered
their corporation. Another survey respondent did not submit a response to this
question. Finally yet importantly, two participants claimed that their firms operate
globally and are not located in a single nation.
Participants in the poll primarily represent four types of organizations: in-
dustry, research, education, and public (government) agencies.
The poll respondents were primarily senior experts (18) who represented
mainly large enterprises (23), as shown in Figure 1.
9
Figure 1: [HCAIM Survey] Participants by organization size
Furthermore, the participants engaged have been preselected so that they
were sufficiently aware, and their organizations worked on AI-related initiatives.
The majority of survey responses came from organizations that are actively de-
veloping or implementing (or both) AI solutions. Most participants stated that
their job primarily focuses on the infrastructure and smart cities sector (13 per-
sons) or the healthcare sector (12 people), as shown in Figure 2.
Figure 2: [HCAIM Survey] Domains in which participants’ organizations have
developed or integrated AI-related solutions
From the participants, who answered with “Other”, we received the follow-
ing clarifications:
• Finance (3);
• Education (2);
• Identity Verification;
• Telecommunication;
• Cybersecurity;
• IT;
• Physical Security;
• Enterprise Software;
• Biotech.
10
The majority of the participants (41 out of 49) state that they plan to design,
to develop or use AI-related solutions within the following year.
The participants in the HCAIM Industry Survey took part in the rel-
evant focus groups as well. To obtain a deeper understanding of the needs and
expectations of the internal and external stakeholders, especially of the program’s
labor market, the HCAIM project consortium set out to conduct focus group in-
terviews.
The grouping criteria, core topics for discussion, and the basic parameters
for forming and implementing the focus group interviews have been identified
and discussed, leveraging the industry survey’s preliminary results in April 2021.
The HCAIM project defines a focus group interview as a market research
method, supporting the consortium to achieve a complete market view of the
needs of the industry related to human-centered artificial intelligence masters’
programs.
Focus Group Size. 6–12 participants
Format. Online video discussion
Focus Group Moderation. Dual moderation. Two moderators moderate each
focus group from the project consortium.
Based on the industry survey results, three focus groups were interviewed,
namely 1) small and medium-sized enterprises, 2) large enterprises, and 3) re-
search and academia. The focus group interviews were conducted virtually, in
English, July 6–7, 2021.
4.1. Societal aspects of AI
As shown in Figure 3, we can conclude that the knowledge and skills related
to technology innovation, needs, requirements analysis, and design are of utmost
importance for the business.
Figure 3: Preferred competences in the field of societal aspects
11
Those competence domains can be connected and related to the following
requirements from the Ethical Guidelines for Trustworthy AI by the AI HLEG:
• Human agency and oversight;
• Transparency;
• Diversity, non-discrimination, and fairness;
• Societal and environmental well-being.
The respondents assessed that during the ideation and design phases (but
not only) of an AI solution development cycle, professionals must also be able to
evaluate continuously the effects of the solution on society and the environment.
4.2. Organizational aspects of AI
Regarding the organizational aspects, the respondents classified the compe-
tences as presented in Figure 4.
Figure 4: The rating of desired competences, related to organizational aspects.
The competence labels and encoding refer to the e-CF [22]
The competences with higher importance are:
• Information security (strategy) management/ governance;
• Risk management;
• Problem management;
• Information and knowledge management;
• Process improvement;
• Project and portfolio management.
The two most valued sectors required as credentials and abilities in the in-
dustry are information security and risk/problem management and information
and knowledge management.
Skills and competencies related to the evaluation of the business implica-
tions of research results, cost analysis at each phase for ensuring security and
12
privacy of personal data and communications, competitive advantages provided
by thrust, resilience, and accuracy of advanced solutions, transparency, and ex-
plainability of technical issues are among the essential requirements of the EU
market for educational institutions.
4.3. Technological aspects of AI
Finally yet importantly, the respondents rated the importance of technologi-
cal competences. They comprise accountability (testing and validation), software
development skills (incommutable when talking about ICT professionals), and
system / engineering thinking:
• Testing;
• Application development;
• Solution development;
• System engineering.
Figure 5: The rating of competences in the technological area
The needs and requirements analysis outcomes should be linked with the
technological capabilities and procedures and any restrictions. The AI HLEG
suggests doing this using a set of “whitelist” rules (behaviors or states) that the
system can follow and “blacklist” limits on behaviors or states that the system
should never permit. Furthermore, the AI HLEG proposes how to modify the
system architecture to ensure that the Trustworthy AI follows these three phases
of the cycle:
• at the “sense”-step, the system should be developed such that it recog-
nizes all environmental elements necessary to ensure adherence to the re-
quirements;
• at the “plan”-step, the system should only consider plans that adhere to
the requirements;
• at the “act”-step, the system’s actions should be restricted to behaviors
that realize the requirements.
13
Although the proposal is somewhat generic, it provides the basic principles
of aligning the technology development part with the requirements analysis and
avoiding human and natural well-being risks.
5. Conclusions
We have recognized that artificial intelligence systems hugely positively influ-
ence business and the economy. To guarantee the long-term viability of this influ-
ence, we must also be concerned with the hazards and other adverse impacts con-
nected with these technologies and ensure that they are effectively implemented.
The growing set of ethical norms for supporting trustworthy and technically
sound AI has inevitably led to an industry shift toward human-centered skills
and competencies. Industry expectations at any stage of the AI life cycle are con-
tinuously changing, offering significant challenges to academia and its ability to
swiftly create and implement an industry-relevant curriculum that addresses AI
professionals’ technical and ethical requirements.
In today’s culture, simply understanding how AI works is insufficient. AI
developers must also be able to analyze and comprehend the impact of AI solu-
tions on organizations, users, and society. Our research and other work in the field
examined in this paper show a need to expand the number of ethically minded AI
engineers to enable the widespread implementation of AI. With many current AI
initiatives still concentrating only on the technical elements of AI, there is a criti-
cal need for engineers who realize that while building and implementing AI, no
compromises with the ethical integrity of the solutions can be made and who have
the skillset to ensure this. Furthermore, the engineers should also understand and
follow the AI-associated threats and risks, the requirements coming from techni-
cal and legal frameworks derived from the AIA [9], and relevant implementation
and certification schemes to achieve the respective level of technically robust and
trustworthy AI-based systems and services.
The purpose of this paper was to report on the findings of a recent analysis of
the skills, competencies, and knowledge relevant to the AI industry to contribute
to the body of knowledge related to shifting AI market demands and thus inspire
educational leaders to reassess their AI curricula. Lastly, as the discussion on AI
ethics and reliability grows more widespread and practical, we anticipate that our
effort will help provide the groundwork for future research.
6. Acknowledgments
The HCAIM Project [26] has received funding from the European Union’s
Connecting Europe Facility under grant agreement №CEF-TC-2020-1 Digital
Skills 2020-EU-IA-0068.
14
7. References
[1] Floridi, Luciano. “Mature information societies – A matter of expecta-
tions”. Philosophy & Technology. (2016). 29(1), 1-4. doi:10.1007/s13347-
016-0214-6.
[2] Bartneck, Christoph, Christoph Lütge, Alan Wagner and Sean Welsh. “Ap-
plication Areas of AI”. (2021). doi: 10.1007/978-3-030-51110-4_9.
[3] Sharkov, George, Christina Todorova, and Pavel Varbanov. “Strategies,
Policies, and Standards in the EU Towards a Roadmap for Robust and
Trustworthy AI Certification.” Information & Security: An International
Journal 50, no. 1 (2021): 11-22. doi:10.11610/isij.5030.
[4] Benbya, Hind, Stella Pachidi and Sirkka Jarvenpaa. “Artificial Intelligence
in Organisations: Implications for Information Systems Research,”. Journal
of the Association for Information Systems (2021): 22:2. Atlanta: Associa-
tion for Information Systems, p. 10. doi:10.17705/1jais.00662.
[5] Thilo Hagendorff. “The Ethics of AI Ethics: An Evaluation of Guidelines”.
Minds & Machines. (2020): 30, 99–120. doi:10.1007/s11023-020-09517-8.
[6] European Commission. High-Level Expert Group on Artificial Intelli-
gence. “Ethics Guidelines for Trustworthy AI”. April 2019. URL: https://
ec.europa.eu/futurium/en/ai-alliance-consultation.1.html.
[7] The Future of Life Institute. Asilomar AI Principles, 2017. URL: https://
futureoflife.org/ai-principles.
[8] IEEE. The IEEE Global Initiative on Ethics of Autonomous and Intelligent
Systems, 2018. URL: https://standards.ieee.org/industry-connections/ec/
autonomous-systems.html.
[9] European Commission, “Proposal for a regulation of the European Par-
liament and the Council laying down harmonised rules on Artificial In-
telligence (Artificial Intelligence Act) and amending certain Union leg-
islative acts.” Artificial Intelligence Act. (21 April 2021). EUR-Lex
– 52021PC0206. URL: https://eur-lex.europa.eu/legal-content/EN/
TXT/?uri=CELLAR:e0649735-a372-11eb-9585-01aa75ed71a1.
[10] European Commission. A European Strategy for Data. February 2020.
Communication from the Commission to the European Parliament, the
Council, the European Economic and Social Committee and the Commit-
tee of the Regions. URL: https://eur-lex.europa.eu/legal-content/EN/TXT/
HTML/?uri=CELEX:52020DC0066&from=EN.
[11] Schiff, Daniel, Bogdana Rakova, Aladdin Ayesh, Anat Fanti and Michael
Lennon. “Principles to Practices for Responsible AI: Closing the Gap”.
(2020). doi: 10.48550/arXiv.2006.04707.
[12] European Commission, “White Paper on AI – A European approach to
excellence and trust” (2020). URL: https://ec.europa.eu/info/sites/default/
15
files/commission-white-paper-artificial-intelligence-feb2020_en.pdf.
[13] Ministry of Transport, Information Technology and Communications.
“Strategy for the Development of Artificial Intelligence in Bulgaria until
2030”. (2020). URL:https://www.mtc.government.bg/sites/default/files/
koncepciyazarazvitienaiivbulgariyado2030.pdf.
[14] Grand View Research. Artificial Intelligence Market Size, Share & Trends
Analysis Report By Solution, By Technology (Deep Learning, Machine
Learning, Natural Language Processing, Machine Vision), By End Use, By
Region, And Segment Forecasts, 2021 – 2028. June 2021. URL: https://
www.grandviewresearch.com/industry-analysis/artificial-intelligence-ai-
market/methodology.
[15] Gartner. Information Technologies Insights. Artificial Intelligence. (2021)
URL: https://www.gartner.com/en/information-technology/insights/artifi-
cial-intelligence.
[16] The White House Office of Science and Technology Policy. “American
AI Initiative: Year One Annual Report” (2020). URL: https://trumpwhite-
house.archives.gov/wp-content/uploads/2020/02/American-AI-Initiative-
One-Year-Annual-Report.pdf.
[17] ICIMS Report on US Tech Talent. URL: https://www.icims.com/company/
newsroom/employers-hired-only-60-of-their-required-tech-talent-in-the-u-s-
over-the-last-three-years-based-on-icims-analysis-of-25-million-applicants.
[18] Gartner Inc. Survey (2019) on Global IT Talent Shortage. URL: https://
www.gartner.com/en/newsroom/press-releases/2019-01-17-gartner-sur-
vey-shows-global-talent-shortage-is-now-the-top-emerging-risk-facing-
organizations.
[19] Forbes Report (2021). Analysing The Software Engineer Shortage. URL:
https://www.forbes.com/sites/forbestechcouncil/2021/04/13/analyzing-
the-software-engineer-shortage/?sh=738c3e32321c.
[20] Daxx Company Report. (2021). The Software Developer Shortage in the
US and the Global Tech Talent Shortage in 2021. URL: https://www.daxx.
com/blog/development-trends/software-developer-shortage-us.
[21] The US Bureau of Labor Statistics. Occupational Outlook Handbook on
Computer and Information Technology. URL: https://www.bls.gov/ooh/
computer-and-information-technology/home.htm.
[22] CEN. CWA-16458-1: European ICT professionals role profiles – Part 1:
30 ICT profiles. Brussels. (2018). CEN-CENELEC Management Centre.
URL: https://itprofessionalism.org/about-it-professionalism/competences/
where-to-buy-the-e-cf-standard.
[23] EDISON Community Initiative. EDISON Data Science Framework EDSF
(2017). Release 2. URL: https://edison-project.eu/edison/edison-data-sci-
ence-framework-edsf.
16
[24] EDISON Community Initiative. Data Science Competence Framework
(CF-DS). (2017). EDISON Data Science Framework: Part 1. doi: 10.5281/
zenodo.1044346.
[25] EU ICT Ethics. European Professional Ethics Framework for the ICT Pro-
fession. (2020). URL: https://www.ict-ethics.eu.
[26] Project HCAIM. The Human-Centred AI Master’s Programme and Com-
munity. URL: https://humancentered-ai.eu.
17