Received 25 April 2025, accepted 6 May 2025, date of publication 8 May 2025, date of current version 16 May 2025.

Digital Object Identifier 10.1109/ACCESS.2025.3568338

AI and Human Resources in a Literature-Driven

Investigation Into Emerging Trends
CRISTINA IANCU 1,2 AND SIMONA-VASILICA OPREA 1
1 Department of Economic Informatics and Cybernetics, Bucharest University of Economic Studies, 010374 Bucharest, Romania
2 Doctoral School of Economic Informatics, Bucharest University of Economic Studies, 010374 Bucharest, Romania

Corresponding author: Cristina Iancu ([email protected])

This work was supported in part by Bucharest University of Economic Studies during the Ph.D. Program; and in part by the Ministry of
Research, Innovation, and Digitization, CNCS/CCCDI-UEFISCDI, under Project COFUND-CETP-SMART-LEM-1, within PNCDI IV.

ABSTRACT Throughout the years, technology has faced many advancements, the star being the power of
Artificial Intelligence (AI), which continues to strike through. This concept has rapidly gained popularity
and has raised points of concern in almost every government from the European Union (EU) because of
the challenges it possesses in terms of efficiency, decision-making and transparency. This paper revolves
around building up an extensive literature review of the academic landscape surrounding the role of AI
in Human Resources (HR) in the public sector, analyzing Web of Science publication trends and thematic
patterns, spanning 12,121 publications from 2020 to 2024, where both Python and R scripts are applied
to extract insights. The findings highlight the relevance of a human-centric approach to AI adoption by
addressing ethical, cultural and compliance concerns, with the aid of advanced Natural Language Processing
(NLP) techniques, such as Latent Dirichlet Allocation (LDA) for topic modelling and keyword co-occurrence
networks for thematic exploration. Moreover, a Hugging Face Named Entity Recognition (NER) model is
employed to systematically identify and categorize AI techniques within the analyzed abstracts, providing
a foundation for subsequent frequency and trend analyses. The analysis brings out a steady growth in
publication volume, with an average of 2,500 papers annually and a significant concentration of research
within domains such as neural networks, algorithm optimization and digital transformation. Apart from the
interdisciplinary focus of the subject, we aim to shed light on the importance of AI-driven HR strategies in
directing administrative insufficiencies of the public sector.

INDEX TERMS AI role, human resources, latent topics, semantic and bibliometric analysis.

I. INTRODUCTION capabilities continue to expand, driving a growing demand

The advent of Artificial Intelligence (AI) has revolutionized for its applications [2]. Organizations and institutions world-
various sectors and domains, particularly since the onset of wide are increasingly integrating AI technologies to enhance
the COVID-19 pandemic, with Human Resource Manage- productivity and streamline repetitive tasks that traditionally
ment (HRM) being no exception. This shift from manual, require manual effort. This approach allows for more efficient
paper-based systems to AI-driven frameworks aligns with the resource allocation and significantly reduces time consump-
broader progression of HRM, transitioning from rudimentary tion. HRM is certainly a domain that can benefit from all
automation to more advanced technology-enabled processes, the advantages that this new technology brings [3]. It further
and culminating in a fully integrated environment [1]. Over aligns with global trends, drawing attention to the need for tai-
the past decades, AI has increasingly spread through HR lored strategies that balance technological advancement with
functions, offering solutions that automate tasks, improve ethical and social considerations, particularly in workforce
decision-making, enhance employee experience, and its integration and decision-making processes [4].
In the public sector, where administrative efficiency and
The associate editor coordinating the review of this manuscript and transparency are paramount, AI has the potential to sig-
approving it for publication was Mu-Yen Chen . nificantly enhance HR functions by automating routine
2025 The Authors. This work is licensed under a Creative Commons Attribution 4.0 License.
VOLUME 13, 2025 For more information, see https://creativecommons.org/licenses/by/4.0/ 81897
 C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

tasks, while also improving talent management or various ethical use, employee trust and organizational culture remain
administrative processes [5]. AI-driven HR strategies are significant barriers to its adoption, determining many insti-
human-resources approaches and programs that leverage arti- tutions to step back with hesitation at first [13], [14]. The
ficial intelligence technologies to enhance, automate, and importance of an approach which is centered around humans
optimize people-related processes across the employee life- was emphasized, balancing technological advancements with
cycle. In practice, an HR strategy qualifies as ‘‘AI-driven’’ if the ethical and cultural considerations necessary for sustain-
it incorporates one or more of the following core elements: able AI adoption [15]. This desire to sow the seeds of this
data-powered decision-making, machine-learning and pre- implementation came because research on its role within the
dictive analytics, Natural Language Processing (NLP-powers public sector remains fragmented, even though the interest for
resume parsing, sentiment analysis of employee feedback and applying AI in HR is growing [16].
AI-driven chat interfaces for answering HR policy questions), In this context, our research aims to bridge this gap by
automation and intelligent workflow, personalization and conducting a broad semantic and bibliometric analysis of
employee experience, integrated talent and workforce plan- academic literature, providing insights into research trends,
ning, ethical, transparent and compliant design. Furthermore, thematic focuses and influential contributors in the field [17].
AI can streamline functions which will be referred to as AI- The contemporaneous context, deeply characterized by digi-
driven HR strategies, such as recruitment and hiring through tal transformations and a continuously evolving labor market,
automated resume screening, candidate ranking and chatbot- requires a reassessment of HRM strategies to add value to
driven initial interviews [6]. It may enhance employee public administration [18]. Further, this paper aims to lay
performance evaluation by analyzing key performance indi- the foundation of an ample research, which will investigate
cators and providing real-time feedback. Additionally, AI and illustrate how AI can enhance processes within the HR
optimizes workforce planning by predicting staffing needs domain, identifying both the challenges and solutions for
and skill gaps. It can also improve learning and development its implementation. We seek to advance the growing body
by personalizing training programs based on employees’ of work on AI in HRM by integrating state-of-the-art NLP
roles and competencies [7]. AI-driven chatbots further assist methodologies within a bibliometric analysis framework.
employees with HR-related queries, reducing the workload While several prior investigations have explored the con-
on HR personnel [8]. Moreover, AI enhances payroll and ceptual links between HRM and AI applications, our research
benefits administration by ensuring accuracy and compli- extends this discourse by adopting BERT-based entity recog-
ance, as well as support employee engagement and sentiment nition and cutting-edge topic modelling approaches, thereby
analysis through NLP and sentiment tracking. Additionally, uncovering deeper insights into the specific AI techniques,
it improves diversity, equity and inclusion (DEI) initiatives by tools, and trending practices. Consequently, we position our
reducing hiring biases, analyzing workplace diversity metrics work at the intersection of AI and HRM, not only by captur-
and suggesting fair promotion and compensation practices. ing conventional bibliometric indicators, such as publication
AI-driven predictive analytics could help HR professionals counts, co-authorship patterns and keyword co-occurrences,
identify employees at risk of burnout or turnover, allow- but also by applying advanced NLP pipelines to extract
ing for proactive intervention. It may also enhance time and quantify the proliferation of AI methods across the HR
and attendance tracking, reducing errors in shift schedul- domain. Unlike traditional bibliometric studies that often rely
ing and ensuring compliance with labour laws. Succession on basic text matching or manual keyword curation, our NER
planning might be improved with AI models identifying high- approach precisely identifies and categorizes AI terms, tools,
potential employees and recommending career development and concepts within vast corpora of HR-related documents.
pathway [9]. Furthermore, AI can streamline compliance By quantifying these AI techniques, we set forth the contours
management by monitoring regulatory changes and ensuring of current scholarly conversations, highlighting which algo-
HR policies align with the latest legal requirements. However, rithms and application areas are gaining popularity.
the public sector faces unique limitations, including bureau- Most prior reviews either emphasize private sector use
cratic issues or regulatory constraints, which complicate the cases or focus on isolated themes such as recruitment or
integration of AI technologies into HR practices [10]. automation, leaving broader patterns and ethical dimensions
Bibliometric analyses have been employed to map the underexplored. This contribution advances academic dis-
evolution of AI-HRM research. Studies utilizing tools such course by offering both a conceptual synthesis and empirical
as VOSViewer and CiteSpace have systematically identified insight into how AI is evolving in the public HRM domain,
major research trends, co-authorship networks and influential supporting more informed research and policymaking.
studies within the field [11]. The findings indicate that AI is Guiding this by four central research questions (RQs),
increasingly integrated with HR analytics, IoT applications we aim to shape its direction and ensure analytical focus.
and big data management, enhancing HR decision-making RQ1: What are the most prominent themes and technological
processes [12]. approaches in AI-related HR research, particularly within
In contrast, while the potential benefits of AI integration the public sector, over the past five years? RQ2: Which
into HRM processes are undeniable, challenges related to countries, institutions and publication sources are leading the

81898 VOLUME 13, 2025

C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

scholarly discourse on the integration of AI into HRM prac- for addressing challenges related to interorganizational AI
tices? RQ3: How is AI being categorized and applied in HR integration through a focus on data protection and equitable
processes, and which techniques or tools are most frequently implementation, particularly in critical areas like healthcare
referenced in literature? RQ4: What trends, research gaps and human resources. These harmonized regulations support
or ethical concerns emerge from the current body of work effective data sharing and tackle privacy concerns, making
that could inform future directions in policy, implementation sure that AI systems function efficiently across member
or governance? These questions form the foundation of the states [21].
research’s methodological design and shape its bibliometric Furthermore, understanding the role and impact of AI in
and semantic analyses, aiming to contribute to the evolving the HR domain is essential to preparing public organizations
intersection of AI and HRM. to navigate ongoing changes and align international standards
This paper is structured in several sections: the next part of efficiency and transparency [22]. Romania aims to adopt
concerns the current state of AI implementation in HR in the and promote the safe and efficient use of AI across all aspects
European Union (EU) and focuses on a bibliometric analy- of daily life, while upholding respect for human rights and
sis of the existing literature regarding this subject. Through social values [23]. This strategy is designed to support both
synthesizing existing literature and recent contributions, this fundamental and applied research in technology by focusing
review identifies prevailing research trends, critical knowl- on the development of authentic and innovative applications
edge gaps and areas requiring further exploration. Section III that contribute to the nation’s economic and social progress.
is designed to showcase the research methodology, while Without such efforts, Romania’s alignment with European
section IV presents the results. Finally, in section V, we draw standards would be significantly slower and more challeng-
several conclusions. ing. The strategy envisions AI as a ‘‘driver of economic
growth and social well-being in Romania’’ [24].
II. LITERATURE REVIEW Germany has also adopted a national AI strategy that
A. CURRENT SITUATION OF THE LEGISLATION includes specific measures for the public sector [25]. The
(REGULATIONS) RELATED TO AI AND HR country has focused on several areas to enhance the efficiency
The increasing integration of AI into various sectors has and transparency of public administration, such as analyzing
spurred significant academic and practical interest, not leav- public data and creating AI and big data application labs.
ing aside its applications within the department of HRM. This This initiative is considered an important step in enabling
section offers an in-depth exploration of the current state of the development and implementation of AI solutions, aiming
research related to the deployment of various AI techniques to increase data accessibility and make it more usable for
applied into human capital management around the world, AI-driven applications, ultimately improving public services.
focusing on its transformative potential and implications, not Furthermore, Germany seeks to enhance the efficiency and
leaving aside its associated challenges. Fortunately, this topic transparency of public administration while promoting col-
is a current priority for the leadership of the EU, as each laboration between the public, private and academic sectors.
member state is encouraged to implement AI in their public Additionally, investing in the continuous training of public
sectors, particularly within the domain of HR. As a result, sector employees is essential to ensure a smooth and efficient
numerous recent studies and publications from EU member transition to AI adoption. This investment can facilitate the
states have begun to explore this subject, while some coun- testing and implementation of innovative solutions while con-
tries have already implemented aspects of AI in practice. tributing to the modernization of public administration [26].
The EU’s Artificial Intelligence Act [19], enacted on Because AI is promising enhanced efficiency and effec-
August 1, 2024, is the first comprehensive global legal frame- tiveness, its rapid pace of AI’s development has led to
work for AI, designed to ensure safety, transparency and the challenges concerning ethical governance, societal align-
protection of fundamental rights. It adopts a risk-based cate- ment and implementation complexity [27]. The Norwegian
gorization of AI applications into four levels: unacceptable, government explores these dynamics in the public sector,
high, limited and minimal risk. Systems used in employment, emphasizing that a responsible approach to AI innovation is
such as CV-sorting tools for recruitment, are classified as significant for addressing these challenges [28]. The authors
high-risk, subjecting them to stringent requirements includ- identified the need for frameworks that adaptively align
ing detailed conformity assessments, transparency obliga- policy and strategy with technical advancements, ensuring
tions and mandatory human oversight to mitigate associated that AI initiatives meet societal and ethical expectations.
risks. These measures aim to balance technological progress It has been stated that the success of AI in public services
with safeguarding individual rights, particularly in HR hinges on its ability to stabilize these dimensions effec-
practices [20]. tively, along with a balanced focus on technical and societal
The EU has further advanced its regulatory agenda with needs.
frameworks that address AI-related risks while encourag- The public sector faces growing pressure to adapt and
ing ethical adoption. Policies emphasizing transparency and integrate continuous innovations to improve efficiency and
accountability align with the EU’s broader objectives of deliver higher-quality services to citizens [29]. As a broad
inclusivity and public trust. The EU establishes a model field that impacts and involves a significant number of
VOLUME 13, 2025 81899
 C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

stakeholders, the adoption of diverse and emerging digital collaboration [39]. These centers focus on diverse sectors
tools has become imperative, as this sector represents another such as healthcare, agriculture, renewable energy and urban
key area strongly marked by the fasted pace of AI develop- development, providing a platform for public-private part-
ment [30]. Countries from all around the world have already nerships to generate high-impact solutions. Funding and
adopted techniques that are reshaping government opera- institutional support from agencies like FAPESP and partner-
tions, allowing better citizen interaction. Beyond service ships with entities like IBM highlight efforts to strengthen
efficiency, the effects of AI in public services depend not only the country’s AI capabilities. Moreover, the Brazilian AI
on its technical capabilities but also on the public’s percep- Strategy (EBIA) outlines 73 strategic actions targeting gov-
tion of its reliability [31]. For instance, AI-powered chatbots ernance, R&D, and workforce training, promoting both the
and digital assistants are increasingly used for administrative decentralization of research and the internationalization of
tasks, enhancing service delivery efficiency, but research sug- AI development. This approach boosts innovation and also
gests that public trust in such systems is contingent on their aligns with global standards, ensuring that Brazil remains
ability to provide clear and unbiased responses. While AI competitive in AI research and applications. These initiatives
chatbots can improve responsiveness and accessibility, they demonstrate a balance between addressing domestic needs
may also trigger concerns about algorithmic decision-making and participating in the global AI landscape [40].
replacing human discretion, emphasizing the importance of HR analytics systems now allow for real-time tracking
designing AI with some accountability measures that aim to of employee performance, reducing inefficiencies and
prevent resistance from citizens [32]. improving recruitment transparency [41]. These systems are
Certainly, digital transformation in governance is not solely supported by integrated IT platforms, combining personnel
a technological shift but requires strategic alignment with data with predictive analytics to optimize talent acquisi-
ethical and social frameworks [33]. It has been stated that tion and workforce planning. One of the notable trends
government social media platforms are being utilized to in this domain is the increasing adoption of gamification
engage citizens while promoting AI initiatives and increas- and augmented reality tools used for employee training and
ing digital literacy. Even if these efforts improve public engagement, which improve the effectiveness of learning and
awareness and participation, they also present concerns about also help to motivate and retain employees by making training
data privacy, which need to be addressed through well- interactive and tailored to individual needs. For example,
structured communication strategies [34]. Moreover, fairness gamified learning platforms have shown remarkable success
and accountability must move beyond basic regulatory com- in developing both soft and technical skills, while AR/VR
pliance, because AI tensions, including bias, privacy risks technologies simulate real-world scenarios for skill enhance-
and algorithmic opacity, remain major challenges that require ment. This approach is particularly effective in addressing the
structured policy responses [35]. routine nature of public sector jobs, boosting the mood and
Some studies have explored the role of conversational AI, productivity of employees [42]. Additionally, AI-powered
such as chatbots, in transforming educational services and systems improve workplace safety and ergonomics. Sensors
university programs, by being increasingly adopted in global and IoT devices integrated with AI algorithms can identify
universities such as Georgia State University and Arizona potential hazards, monitor health parameters, and rec-
State University to streamline communication, administrative ommend interventions to prevent injuries. This proactive
efficiency and student engagement [36]. Because chatbots approach not only ensures compliance with safety regulations
assist in reducing workload for university staff by addressing but also contributes to employee well-being [43].
repetitive inquiries and allows staff to focus on more complex The rise of e-HRM (Electronic Human Resource Manage-
tasks, they provide real-time, personalized support, enhanc- ment) signifies the fusion of HR practices with IT systems
ing the student experience and increasing retention rates. to deliver efficient, scalable, and integrated HR solutions.
Concerning education and workforce development, AI tools E-HRM facilitates automation in performance evaluations,
like ChatGPT [37] have proven instrumental in reimagining recruitment, and employee engagement, reducing administra-
learning methodologies. They enable interactive, customized tive burdens and enhancing real-time decision-making capa-
educational experiences that address skill deficiencies bilities. Furthermore, it promotes global talent acquisition
and promote ongoing professional growth. By automating through tools like e-recruitment and e-learning platforms,
mundane tasks, AI frees HR professionals to focus on as seen in Indian industries like SBI’s Gyanodaya Gateway,
strategic initiatives, such as cultivating a supportive work which demonstrate the necessity for organizations to adopt
culture [38]. technologically advanced HR practices to remain competitive
Diving more into what other countries have already done, in a globalized market [44]. Moreover, the development of
Brazil has made significant strides in advancing AI by creat- human-inspired networking and collaboration between edge,
ing a robust infrastructure to foster innovation and address network, and cloud systems has significantly transformed
national challenges. The establishment of 11 Centres for AI service deployment, by utilizing the fusion between
Engineering and Applied Research (CPE/CPA) for AI reflects deep neural networks (DNNs) and advanced optimization
Brazil’s commitment to integrating research and industrial algorithms [45]. These systems enhance the efficiency and

81900 VOLUME 13, 2025

C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

precision of hiring decisions while mitigating the human B. BIBLIOMETRIC ANALYSES ON AI AND HR
biases typically associated with traditional recruitment [46]. AI stands out as one of the most promising emerging
Beyond recruitment, AI is revolutionizing performance technologies, with the potential to radically transform pro-
management by enabling real-time monitoring and predictive cesses within the field of HR by automating repetitive tasks,
analytics. AI offers unprecedented opportunities for person- improving decision-making and increasing transparency and
alization, such as learning programs based on individual fairness in recruitment, while also playing a significant role
employees’ needs by analyzing data about their perfor- in supporting performance management [58]. By drawing
mance, improving learning outcomes [47]. For instance, on a systematic review of global case studies and factual
AI-powered coaching systems simulate real-world scenarios, research, this section synthesizes insights from all the the-
enabling employees to develop complex skills in a risk-free oretical frameworks and empirical findings gathered so far.
environment. These tools, however, should augment rather The decision to address this topic stems from the pressing
than replace human trainers, as the latter provide the emo- need to explore how AI can be effectively integrated into
tional intelligence and contextual understanding necessary HR processes in the public sector, where challenges such as
for effective mentorship [48]. bureaucracy, limited resources and the need to attract and
As these systems become more pervasive, organizations retain talent are especially pronounced [59], [60].
must address concerns about surveillance and ensure trans- In the past years, numerous studies have been conducted
parency in how performance data is collected and uti- on this new phase of technology, revealing a progressive
lized [49], but they also must ensure that these tools are trajectory to AI adoption in various processes that imply
transparent and adhere to ethical standards to maintain trust human activity of any kind, ranging from basic automation to
among stakeholders, as they remain paramount in integrating fully embedded AI-driven systems. For instance, researchers
AI into HRM and its administration [50]. Special atten- emphasize the dual role of AI as a tool for operational
tion is given to the intersection of AI with human-centric efficiency and strategic decision-making, capable of revolu-
design principles, highlighting how organizations can achieve tionizing recruitment, performance evaluation and employee
a harmonious balance between technological innovation and engagement [61], [62].
human values [51]. The concept of ‘‘human-in-the-loop’’ AI HRM has become a strategic direction in organiza-
highlights the critical role of human oversight in enhanc- tions, driven by economic, political, social, and especially
ing machine learning systems [52]. It becomes mandatory technological transformations, as well as the demands for
to incorporate expert feedback into algorithmic training, competitiveness and operational excellence [63]. Highlight-
so that organizations can reduce biases and improve the ing its faced challenges, it is clear that this area has been
interpretability of AI decisions [53]. Such a collaborative among the most affected by new trends and is in a state of
approach ensures that AI complements human expertise, constant transformation, redefining its role and significance
maintaining the nuanced judgment required in HR functions, within organizations [64]. The integration of HR with busi-
while also benefiting from the efficiency that automated ness strategies is achieved by positioning its management at
systems can bring [54]. the highest decision-making level within the organization.
Employee apprehension regarding AI adoption, particu- The more reliable the information, the more accurate the
larly concerns related to trust, ethics and job displacement, decision-making process becomes [65]. For instance,
require a proactive approach to facilitate seamless integra- AI-driven recruitment systems like Manatal and SeekOut
tion [55]. Establishing confidence in AI-driven processes analyze candidate data to predict their suitability for roles,
depends on fostering transparency, maintaining open com- enhancing the quality and speed of hiring [66].
munication channels, and equipping employees with the Table 1 provides a consolidated overview of eight studies
necessary skills to adapt to evolving work environments [56]. that examine the intersection of AI and HRM through biblio-
To navigate these complexities effectively, organizations metric and systematic review methods. Each entry specifies
must implement ethical governance structures that uphold the primary objectives, ranging from mapping intellectual and
fairness and accountability while simultaneously building up conceptual structures to identifying emerging research trends,
employee engagement. Ensuring that employees understand and describes the bibliometric techniques used. The databases
AI’s role and potential benefits through continuous dialogue and the number of documents is included. Notably, the studies
and training remains essential for securing their coopera- span publication periods up to 2023 or 2024, reflecting the
tion and minimizing resistance [57]. Overall, AI continues most recent trajectories in AI-driven HRM research.
to shape HRM in profound ways, with research advancing
toward a deeper understanding of its implications, benefits,
and challenges. Taken together, these studies underscore the C. CONTRIBUTIONS TO STATE-OF-THE-ART LITERATURE
dual necessity of harnessing AI for administrative efficiency Our research makes several distinct contributions to the exist-
while addressing its ethical and societal implications. Public ing research on AI in HR within the public sector. Unlike
trust is still a significant factor in adopting AI, needing more previous studies that primarily focus on AI applications in
transparency. private sector HRM, our research specifically targets HR

VOLUME 13, 2025 81901

 C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

TABLE 1. Comparing previous studies.

practices in government institutions, addressing a relatively This study further advances the field by categorizing AI
unexplored area. By analyzing Web of Science publications techniques within HRM research, providing a granular anal-
from 2020 to 2024, our research provides a timely and up- ysis that distinguishes between different AI applications such
to-date perspective on AI trends in HRM, whereas some as decision-support systems, recruitment automation and
previous works focus on broader timeframes or pre-pandemic workforce analytics. This fills a methodological gap in prior
data. studies that often treat AI as a monolithic concept without
A methodological contribution lies in the integration differentiating its varied implementations in HRM. Another
of advanced NLP techniques, including LDA for topic significant contribution is the emphasis on a human-centric
modelling, keyword co-occurrence networks for thematic approach to AI adoption, addressing ethical, cultural and
structuring, and a Hugging Face Named Entity Recognition compliance concerns alongside technological advancements.
(NER) model to systematically classify AI techniques used Additionally, our research offers a trend analysis of the
in HRM research. This approach offers a more sophisticated steady rise of AI-related HR publications, with an average of
and automated method for identifying thematic structures 2,500 papers per year, and identifies thematic clusters such
in AI-related HR literature, surpassing the capabilities of as neural networks, digital transformation and algorithmic
traditional bibliometric analysis tools such as VOSViewer decision-making. The study also suggests emerging research
and CiteSpace. directions, including AI’s role in mitigating administrative

81902 VOLUME 13, 2025

C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

inefficiencies in the public sector, optimizing workforce plan- criteria, a total of 12,121 records were extracted on January 2,
ning and ensuring ethical AI integration that maintains human 2025. The decision to retain such a large number of records
oversight while imposing automation. using the proposed query was methodologically grounded:
the use of unsupervised machine learning techniques benefits
III. METHODOLOGY AND DATA from diverse datasets to ensure a proper pattern extraction
In this section, we present the input dataset, and the main and a wide thematic coverage, compared to a smaller one
methods used to extract insights from the collected infor- that could risk overfitting to narrow trends and missing
mation. The analysis is based on 12,121 publications and emerging or less frequently discussed themes. For integrity
articles from the Web of Science, a widely recognized and reasons, a manual verification of relevance was conducted
trustworthy database known for its large amount of academic every 100 papers, where one paper from each batch was man-
literature. The bibliometric dataset was extracted with the aid ually reviewed by the authors to ensure its alignment with the
of query that captures relevant publications at the intersection intended query. Also, given the interdisciplinary nature of AI
of AI with HR and the public sector: adoption in public sector HR, it became essential to capture
cross-domain references from fields like computer science,
((AF = ‘‘artificial intelligence′′ ∨ AF public administration, organizational behavior and ethics. For
= ‘‘AI ′′ ) ∧ (AF = ‘‘human resources′′ ∨ AF the used tools that follow this section, the large dataset will
= ‘‘HR′′ )) ∨ ((AF suffer refinements through rough filters and semantic post-
= ‘‘artificial intelligence′′ ∨ AF processing, so that the results remain relevant despite the
breadth of the initial input. The dataset includes bibliometric
= ‘‘AI ′′ ) ∧ (AF metadata such as titles, abstracts, keywords, publication years
= ‘‘public sector ′′ ∨ AF and citation data, which will become the starting point for the
= ‘‘public administration′′ ∨ AF following analyses in this study.
= ‘‘government ′′ )) After collecting the data and extracting key bibliographic
information, we processed and cleaned the dataset using a
∧ Publication Years Python script (as in Fig. 1). This resulted in a refined file
∧ Chosen Research Areas containing the desired data, making it suitable for further
∧ Filtered Categories analysis.
∧ Filtered Publishers The data was downloaded in batches of 1,000 records
each, resulting in multiple.bib files which were program-
where: matically merged into a single file using R language. This
∨ → OR step simplified data handling and ensured that no records
∧ → AND were inadvertently omitted, providing a solid foundation for
AF → All Fields (title, abstract, keywords). bibliometric analysis.
Publication Years → Includes 2020, 2021, 2022, 2023, 2024. Afterwards, the process continued with data cleaning:
Chosen Research Areas → Includes fields like Computer Sci- missing or incomplete metadata were excluded to maintain
ence, Engineering, Public Administration, Social Sciences, the dataset’s integrity and duplicate entries were removed
etc. to prevent skewed analyses. Optimizing the dataset was
Filtered Categories → Covers Web of Science categories necessary for conducting advanced bibliometric analysis,
relevant to AI, HR and Public Sector studies. as the pre-processed dataset was imported into the R environ-
Filtered Publishers → Elsevier, Springer Nature, IEEE and ment, where it was analyzed using the bibliometrix package.
MDPI. Bibliometric analyses included examining trends in key-
Due to its rapid development, this topic has been predom- word usage, visualizations such as publications per country
inantly studied over the past five years, and so we restricted plots or collaboration networks or keyword co-occurrence
the publication years to the range from 2020 to 2024 to focus maps. Additionally, conceptual and thematic maps provided
on the most recent advancements. Additionally, the research insights into emerging themes and areas of research focus.
areas were narrowed down to those most relevant to the topic, The interactive biblioshiny application was used to refine
including Computer Science, Engineering, Public Adminis- the analysis, offering user-friendly tools for generating the
tration, Social Sciences and Business & Management. These mentioned plots.
fields cover the technical foundations of AI, the policy con- To sum up, our research employed both Python and R, each
text of public administration and the organizational focus of chosen for its specific strengths. Python is primarily used
HR practices. The query was further refined to include arti- for advanced NLP tasks such as NER and LDA, utilizing
cles issued by publishers such as Elsevier, Springer Nature, libraries like transformers, gensim and pyLDAvis for their
IEEE and MDPI, enhancing the quality and credibility of flexibility and deep integration with machine learning work-
the results. We chose these publishers as they are known for flows. In contrast, R was used for conducting bibliometric
editorial standards and host leading journals in AI, gover- analysis through the bibliometrix package and its companion
nance and management research. After applying all filtering application biblioshiny, which provide interactive tools for

VOLUME 13, 2025 81903

 C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

visualizing results such as co-authorship networks, keyword and reporting. Using the words from abstracts rather than
trends and thematic clusters. the field ‘‘keywords’’ from the bibliometric dataset has pro-
vided richer contextual data for the LDA analysis, allowing
A. LATENT DIRICHLET ALLOCATION (LDA) for more nuanced topics. Abstracts are a valuable input for
To explore the latent thematic structure of the academic topic modelling, as they typically contain sentences and full
literature, we implemented a LDA model using the gensim explanations of a paper’s scope, methodology, findings and
library in Python. Fig. 2 depicts a structured flowchart that implications.
visualizes the process of training and analyzing a LDA model,
in a detailed step-by-step explanation of the flowchart and B. NAMED ENTITY RECOGNITION (NER)
the associated procedure. Starting from raw textual data, The present research employs a multi-stage methodology to
the process involves a few preprocessing steps to clean and identify and catalogue AI techniques across a substantial
structure it, which included lemmatization, stopword removal input of scientific abstracts. Initially, a dataset of over 12,121
and lowercase transformation, with the id2word dictionary items was divided into batches of approximately 500 each,
constructed from tokenized abstracts. Firstly, abstracts were resulting in 29 separate CSV files. The text within each
converted to lowercase and stripped of punctuation using reg- abstract was preprocessed by converting it to lowercase,
ular expressions and each sentence was tokenized into words. removing punctuation and filling missing values to standard-
We removed standard English stopwords from the NLTK ize the input, being already ready after previous section,
corpus to retain only meaningful terms, and then the results which prepared the data for LDA.
were lemmatized using spaCy. The final cleaned tokens were A Hugging Face NER model, specifically ‘‘dbmdz/bert-
used to generate a dictionary and a corpus, which formed the large-cased-finetuned-conll03-english’’, was then applied to
basis of the LDA model input. For training and generating each cleaned abstract, extracting potential AI-related entities.
the model itself, a few parameters had to be set: the number The model is optimized for the CoNLL-2003 NER task,
of topics and other model parameters, including alpha and and it was used in its pre-trained form without additional
beta, were left at their default symmetric settings for baseline fine-tuning, ensuring broad generalizability across the large
comparison. After the initial run of the LDA model, a tuning academic dataset and preserving reproducibility.
process was performed to optimize the number of topics, To further refine these preliminary results, a predefined
aiming to maximize the coherence score and thereby identify lexicon of core AI and HR-specific terms (e.g., ‘‘machine
the most interpretable and meaningful topic structure. In the learning’’, ‘‘deep learning’’, ‘‘NLP’’ and ‘‘AI-driven work-
next step, the results are saved and visualized using the force planning’’) was utilized in conjunction with fuzzy
python package pyLDAvis, offering interactive exploration of matching. This method employed Levenshtein-based simi-
topics and their relationships. The final outputs, including the larity thresholds to identify partial or misspelled variants,
trained model and visualization, are stored for further analysis thereby capturing meaningful variants and spelling deviations

FIGURE 1. Research flow methodology.

81904 VOLUME 13, 2025

C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

FIGURE 2. LDA methodology flow.

that might otherwise have been overlooked. Terms that fell Fig. 3 presents an overview of the methodological work-
below the similarity threshold but appeared unique or con- flow, depicting the segmentation of the initial dataset, the
textually significant were set aside for further review, and so preprocessing steps structured for clarity, and the analysis
a striking balance was set between recall and precision in the stages such as NER, fuzzy matching, CSV merging, semantic
detection of AI concepts. filtering and the final step of frequency calculation.
Following the initial extraction and fuzzy matching pro-
cedures, all 29 initial CSV files were merged into one
dataset, where each abstract was associated with sets of exact
matches, fuzzy matches and potential new terms. Afterwards,
we improved result accuracy by applying multiple filters and
validation measures. First, we used a Sentence Transformer
model (‘‘all-MiniLM-L6-v2’’) to generate vector embed-
dings for both the discovered terms and the reference AI/HR
term list, allowing a cosine-similarity check to discard seman-
tically irrelevant entries. Second, a series of lexical and
semantic constraints was imposed, removing short words
and verifying each candidate term against an extended set
of AI-related keywords or linguistic patterns that indicate
genuine AI methods. Consequently, off-topic tokens were
filtered out, leaving a set of refined terms for a following
frequency analysis. Finally, with the aid of a customized FIGURE 3. NER methodology flow.
dictionary of closely related terms, we managed to group
similar expressions under standardized labels. For example,
‘‘neural networks’’, ‘‘neural network’’ and ‘‘DNNs’’ were IV. RESULTS
unified as ‘‘neural networks’’. This allowed us to construct a In this section, we employ R and Python scripts to provide
clean and structured term frequency table, which served as the insights extracted from the input dataset at the intersec-
basis for analyzing the prevalence of specific AI techniques tion of AI and HR, especially found in public institutions.
across the corpus. Fig. 4 illustrates a summarized overview over the loaded

VOLUME 13, 2025 81905

 C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

data. The dataset comprises a substantial body of research,

which consists of 12,121 documents that are published
from 2020 to 2024. This time span is particularly relevant for
the emerging trends in AI these recent years, and the interest
for the researched fields seems to grow year by year with an
average growth rate of 12.5%.

FIGURE 5. Annual scientific publications and the average citations per

year.

FIGURE 4. Overview of data found. subsequent zones where they contribute progressively with
fewer articles. The grey shaded ‘‘Core sources’’ area high-
Fig. 5 illustrates the annual scientific publications (bar lights the most influential journals in the dataset, such as
chart) and the average citations per year (line chart) for ‘‘IEEE Access’’ and ‘‘Sensors’’, which play a central role in
research publications in the dataset, having as publication shaping the literature written over AI in HR sector. The steep
years the period ranging from 2020 to 2024. The blue bars slope in the core zone indicates a concentration of knowledge
represent the total number of publications per year, demon- within a select group of sources, emphasizing their central-
strating a consistent output of scientific work, with a peak ity and influence. As the curve gets flatter, it demonstrates
observed in 2023, followed by a gradual decline toward the diminishing contributions of peripheral sources, aligning
2025. This trend proposes an initial surge in research interest, with Bradford’s principle of information dispersion.
driven by evolving topics, which diminished in subsequent Displaying the most frequently occurring terms in the
years. Conversely, the red line represents the average citations dataset through a cloud of words provides an extensive per-
per year, showcasing a steep decline from 2021 to 2024. spective over the thematic landscape within the analyzed
While 2020 and 2021 exhibit relatively high average cita- literature. As seen in Fig. 7, dominant terms like ‘‘classifica-
tions, an indicator of impactful studies during these years, tion’’, ‘‘model’’, ‘‘algorithm’’, ‘‘system’’ and ‘‘framework’’
the decreasing trend suggests that more recent publications indicate a strong emphasis on computational methods and
(2022–2024) may not yet have accumulated significant cita- structured approaches in literature. Similarly, terms such
tion counts due to their recency. At the same time, the as ‘‘networks’’, ‘‘design’’, ‘‘optimization’’ and ‘‘neural net-
decrease could also suggest that while more research is being works’’ suggest an alignment with technical and machine
published, not all of it has the same level of impact or learning-based methodologies. Moreover, the presence of
visibility. This highlights the need for higher-quality, more terms like ‘‘prediction’’, ‘‘performance’’ and ‘‘recognition’’
focused studies that address key challenges and offer real- showcases the applied nature of the research, aimed at solving
world solutions, rather than just increasing the quantity of real-world problems.
publications. This image brings out the focal points of research efforts,
While our analysis focuses on the 2020–2024 period, exist- demonstrating the importance of systems thinking and data-
ing reviews indicate that AI-related HR research between driven approaches in AI and its related disciplines. It also
2015 and 2019 was more limited in scope and primarily reflects connections to HR topics, characterized by words like
centered on private sector use cases such as recruitment ‘‘communication’’, ‘‘resource-allocation’’, ‘‘management’’,
automation or workforce analytics. The notable increase in ‘‘privacy’’ or ‘‘security’’. These terms highlight critical
publications from 2020 onward aligns with the onset of aspects of HR, including effective interaction, strategic
the COVID-19 pandemic, which significantly accelerated planning, data protection and workplace safety.
digitalization across both private and public institutions. Affiliations that are the most prolific from the ones con-
In particular, the shift to remote work, increased demand for tributing to research in this field are shown in Fig. 8. The chart
virtual HR solutions and administrative pressure to maintain ranks institutions based on the number of articles authored,
service delivery under constrained conditions likely con- with ‘‘Seoul National University’’ and ‘‘Korea University’’
tributed to the spike in research interest. These contextual leading the field, each contributing with over 1,000 articles.
factors help explain the sharp upward trend captured in the This dominance by South Korean institutions indicates their
post-2020 literature. leadership in AI research related to HR. The presence of
Fig. 6 represents the core sources that are significant to a other global institutions, such as ‘‘University of Granada’’
particular research domain by the Braford’s Law bibliometric and ‘‘Katholieke Universiteit Leuven,’’ underscores the inter-
principle. The sources are split into a ‘‘core zone’’, where national collaborative nature of the research. The bubble
a small number of sources contribute to most articles, and sizes represent article contributions, where is clear that the

81906 VOLUME 13, 2025

C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

FIGURE 6. Core sources by Bradford’s law.

strong focus on AI and technological innovation, as they have

already showed us in the past decades.
The middle column shows the scientific categories or
research areas, such as computer science, engineering,
telecommunications and automation & control systems. The
width of the links connecting these categories to the coun-
tries on the left and the keyword on the right shows their
importance and relevance in the analyzed dataset. Figure 8
shows that computer science appears as a central hub, linking
multiple countries and keywords.
In the rightmost part of the plot, the third column stands for
the most frequently occurring keywords, such as deep learn-
ing, artificial intelligence, machine learning, training and
FIGURE 7. Keywords word cloud. neural networks. These keywords highlight the core topics
explored in the research. For instance, ‘‘deep learning’’ and
‘‘artificial intelligence’’ appear prominently, suggesting their
centrality in AI-related studies. The connections between
ones from Asia are significantly over others by visually ‘‘computer science’’ and these keywords emphasize how this
highlighting this. field is fundamental in driving research on these innovative
The three-fields or Sankey plot in Fig. 9 displays the rela- topics, as it includes AI techniques within its growing fields
tionships between three interconnected dimensions: countries of research.
(AU_CO), scientific categories (SC) and keywords (DE). Table 2 unveils an ordered list of the most frequently occur-
It provides insights into the contributions of specific countries ring terms in the dataset, providing a summary that quantifies
to scientific categories and how are they associated with key the dominating key concepts from our research. Summing the
research themes or topics. Countries such as Korea, China, occurrences of the term network in both singular and plural
Spain and the USA are shown as prominent contributors. The form, it appears with the highest frequency (448), indicating
width of the links originating from these countries points out its centrality to the discourse, reflecting the significance of
the strength of their connections to different scientific cate- networks in computational systems, neural networks and
gories in the central column. For instance, Korea has strong connectivity in AI applications. Other highly frequent terms
ties to ‘‘computer science’’ and ‘‘engineering’’, reflecting include model (397), classification (374), and algorithm
its significant contribution to these domains, driven by its (303), emphasizing the focus on developing and applying

VOLUME 13, 2025 81907

 C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

FIGURE 8. Most relevant affiliations.

FIGURE 9. Three-fields (Sankey) plot.

predictive models, classification systems and algorithmic intersection of AI with connectivity and organizational
methodologies. Terms such as system (272), design (244), applications.
and framework (190) highlight the importance of structural Fig. 10 shows the top 10 research areas based on the
and architectural considerations in AI implementations. In an number of publications. The size of each tile is pro-
interesting way, optimization (198) and performance (164) portional to the number of publications in each research
underscore the emphasis on refining AI systems for efficiency area, with ‘‘Computer Science’’ dominating as the largest
and effectiveness, while prediction (174) and recognition contributor, accounting for 9,914 publications. Following
(166) suggest ongoing efforts in predictive analytics and this, ‘‘Engineering’’ ranks second with 6,292 publications
pattern recognition technologies. Additionally, terms like and ‘‘Telecommunications’’ occupies a notable third posi-
internet (189) and management (115) reflect the growing tion with 2,151 publications. Smaller but still relevant

81908 VOLUME 13, 2025

C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

TABLE 2. Frequency of most used terms. relevance and frequent connections to other terms. The net-
work is divided into clusters, with distinct colors representing
thematic groupings. The thickest link is the one between
‘‘internet’’ and ‘‘things’’, because Internet of Things (IoT)
is a gigantic domain within the realm of technology and
informatics. One cluster focuses on machine learning models
(red), while another centers on internet technologies and
challenges (blue), and a third highlights prediction and neural
network techniques (green). This network illustrates how
the literature integrates diverse fields, connecting theoretical
foundations, algorithmic design, and practical applications.
The visualized interplay between terms offers insights into
the convergence of multiple domains regarding AI advance-
ments in the already existing research.

contributions come from fields like ‘‘Physics’’ (407), ‘‘Oper-

ations Research & Management Science’’ (361), ‘‘Mathemat-
ics’’ (350) and ‘‘Chemistry’’ (512). The distribution reflects a
strong emphasis on technology-related fields, indicating the
prominence of artificial intelligence, systems engineering, FIGURE 11. Words co-occurrence network.
and communication technologies in current research. The
significant representation of areas like ‘‘Imaging Science’’ Fig. 12 depicts the global distribution of scientific publica-
and ‘‘Automation’’ suggests an interdisciplinary focus, con- tions based on the analyzed bibliographic data, highlighting
necting computer science and engineering to applications significant contributions from South Korea, China, and Spain.
in fields such as robotics and diagnostics. The inclusion Using a custom R script, the data was processed to count
of ‘‘Physics’’ and ‘‘Mathematics,’’ though smaller, high- unique country contributions for each article, displaying each
lights their supporting role in underpinning theoretical and nation’s involvement. The map reveals that South Korea
computational advances in these domains. leads with 1,956 articles, followed by China (1,094) and
Spain (1,092). These high outputs can be attributed to robust
research funding, strategic focus on AI, and strong insti-
tutional frameworks. South Korea’s ‘‘National Strategy for
Artificial Intelligence’’ [67] and China’s ‘‘New Generation
Artificial Intelligence Development Plan’’ [68] have pro-
pelled their research output, particularly in AI applications
for governance, urban studies, and public sector innova-
tions. Their national AI strategies and substantial government
funding are clearly directed toward innovation in public
administration and such initiatives reflect broader geopo-
litical priorities which position themselves as AI leaders
globally. However, the underrepresentation of developing
regions raises concerns about the global equity of AI adoption
in HRM, indicating a potential digital divide that may rein-
FIGURE 10. Plot for top 10 research areas. force existing administrative inefficiencies and limit inclusive
growth. Spain, benefiting from its involvement in the Euro-
The co-occurrence network in Fig. 11 reveals the rela- pean Union’s Horizon programs, focuses on AI’s societal and
tionships and connectivity between key terms from the administrative impacts. The map further highlights moderate
analyzed abstracts. Larger nodes such as ‘‘classification’’, contributions from nations like the United States and India,
‘‘model’’, and ‘‘internet’’ signify central themes with high reflecting their global academic prominence.

VOLUME 13, 2025 81909

 C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

FIGURE 13. Country collaboration network.

uncover key insights from the thematic focus areas, thus offer-
ing a foundation for deeper exploration and interpretation of
FIGURE 12. Country scientific publications.
the current literature review.
Moreover, to assess topic quality, we tuned the LDA model
and computed the coherence scores, for different numbers
The previous map pairs well with the collaboration map of topics and hyperparameters, which measure the semantic
shown in Fig. 13, which visually demonstrates the global similarity between top-ranked words in each topic. The high-
scientific collaborations between nations in the field of est score of 0.7079, associated with the three topics depicted
study that we have analyzed. Larger nodes, such as those below, reflects a good degree of topic coherence, acceptable
representing South Korea, China, and the United States for exploratory bibliometric purposes.
or even Spain, signify these countries’ prominent roles in Topic 1 ‘‘Data-Driven Modeling and Performance Opti-
engaging international research partnerships. For instance, mization’’ is visible in Fig. 14 and has key terms like
strong collaborations can be observed between China and ‘‘data’’, ‘‘model’’, ‘‘proposed’’, ‘‘learning’’, ‘‘models’’, ‘‘per-
South Korea, as well as between European countries such formance’’, ‘‘method’’, ‘‘training’’, ‘‘features’’. It represents
as Spain, Germany and Italy, showing significant clustering 44.3% of all tokens, indicating its dominant importance
tendencies based on different regions. Furthermore, countries across the dataset. The high frequency of terms like ‘‘data’’,
like the United States act as bridges, connecting multiple ‘‘model’’ and ‘‘performance’’ suggests a strong focus on
clusters across continents. Moreover, this plot emphasizes the involving data to develop and optimize machine learning
importance of international collaborations in shaping scien- models and methods, which emphasize data-centric
tific frameworks, whereas also revealing potential geographic approaches, and the methodologies used to improve pre-
inequalities, as less active participation is observed from dictive performance. It represents research focused on the
regions such as Africa and South America. development and refinement of models and methods in
However, this regionalization also signals critical dispari- AI for tasks involving data collection, preprocessing and
ties in how AI in HR is being studied and shaped. Countries model performance evaluation. The prominence of ‘‘train-
in Africa, South America and parts of Southeast Asia are ing’’, ‘‘features’’ and ‘‘accuracy’’ indicates interest in the
minimally represented in these collaborations, which raises training phase of machine learning models, where datasets
concerns about knowledge monopolies dominated by high- are structured to yield accurate predictions.
income countries. If collaboration networks remain uneven,
the resulting AI systems and governance models risk being
geopolitically biased, less sensitive to local labor markets,
and less aligned with the administrative realities or ethical
norms of underrepresented regions. Moving forward, build-
ing inclusive frameworks that enable broader participation
through open data, capacity building and equitable funding
mechanisms will be essential for shaping AI systems that
reflect global needs and values in HRM.
The LDA analysis conducted on the corpus of documents
revealed three distinct topics that encapsulate the thematic
structure of the data [69]. These were identified based on the
frequency and relevance of terms within the dataset, enabling
a nuanced understanding of the underlying themes. The inter- FIGURE 14. Topic 1 from LDA analysis.
active pyLDAvis visualization provides a dual perspective:
an intertopic distance map that illustrates the relationships The second topic from Fig. 15, ‘‘Artificial Intelligence in
between them, and bar charts highlighting the most prominent Systems and Applications’’ is marked by the terms ‘‘ai’’,
and relevant terms for each group. This way, we manage to ‘‘data’’, ‘‘intelligence’’, ‘‘artificial’’, ‘‘proposed’’, ‘‘system’’

81910 VOLUME 13, 2025

C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

and accounts for 31.2% of tokens, making it the second most

dominant theme. The frequent appearance of terms like ‘‘ai’’,
‘‘data’’ and ‘‘systems’’ suggests a focus on AI’s integration
into broader systems and its application to various domains,
including public or private administration and technology
development.

FIGURE 16. Topic 3 from LDA analysis.

The presence of ‘‘iot’’ and ‘‘smart’’ implies the convergence

of AI with emerging technologies like IoT to deliver innova-
tive solutions. It may describe how AI systems are adapted for
FIGURE 15. Topic 2 from LDA analysis. real-world applications, succeeded by the inclusion of terms
from the government realm that suggests AI’s growing role
Following up, the third topic from Fig. 16 centers on in decision making and policy implementation.
how AI technologies are embedded in systems, such as We identified distinct clusters of research topics, each
smart networks and various human related services. Words characterized by a unique set of terms. These reflect the
like ‘‘government’’, ‘‘public’’ and ‘‘services’’ indicate that interdisciplinary nature of the research, concerning areas such
this cluster includes discussions on AI’s role in mod- as technological innovations in HR, challenges in both private
ernizing public administration and enhancing efficiency. and public administration and ethical considerations in AI

TABLE 3. Excerpt results from NER analysis.

VOLUME 13, 2025 81911

 C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

TABLE 4. AI techniques frequency found in abstracts. Coming behind a series of additional checks and consol-
idation steps, such as semantic filtering using the Sentence
Transformer model and applying the dictionary of synonyms,
these AI-related terms were assembled into a final two-
column table, mapping each AI technique to its frequency
of occurrence across all abstracts. This output is presented
in Table 4, which enumerates well-established methods such
as ‘‘neural networks’’, ‘‘machine learning’’ and ‘‘deep learn-
ing’’ and also a spectrum of less frequent methods (e.g.,
‘‘chatbots’’, ‘‘emotion recognition AI’’, ‘‘object detection’’)
that represent either niche applications or possess a potential
for further investigation.
The high-frequency terms point to the continued dom-
inance of foundational AI paradigms, particularly those
related to supervised learning and advanced neural archi-
tectures, across a diverse range of research applications.
Meanwhile, the lower-frequency items draw attention to areas
that may be on the rise or are currently less common but
could experience growing attention in the near future. By this
study, we intended to offer a thorough snapshot of the evolv-
ing AI landscape and its intersection with HR or broader
organizational contexts.

V. CONCLUSION
This bibliometric analysis examines the diverse academic
nature of AI research across various HRM processes. In this
paper, we have brought to light how advanced NLP tech-
niques and complex AI applications are involved into HR
systems, also not setting aside the already existing and future
potential challenges associated with this trend. While current
research emphasizes the significant impact of these technolo-
gies in areas such as automated recruitment and employee
implementation. Despite the wealth of research identified, engagement analytics, the results also point to important gaps
several gaps may remain. For instance, from the ‘‘where?’’ in ethical governance and policy integration.
perspective, much of the research focuses on developed A central theme emerging from the examined litera-
nations, with limited studies examining the impact of AI in ture is the ethical dimension of AI integration within
HRM of developing countries. By examining the ‘‘when?’’ HRM. As technological advancements continue to shape
part, we can conclude that most studies are cross-sectional, HR processes, stakeholders must carefully navigate the
lacking longitudinal insights into the long-term impacts of legislative background to ensure that cognitive computing
AI practices applied into HR sector and employee outcomes. initiatives do not undermine employee trust, neither com-
Also, the analysis also underscores the need for more com- promise an organization’s compliance with legal and ethical
prehensive studies addressing regional disparities, long-term standards.
impacts and ethical frameworks, because still there is a lack The evolving European regulatory framework, particularly
of standardized guidelines to govern AI deployment in HR, the EU’s Artificial Intelligence Act and various national
particularly within public administration. strategies, marks the key influence of legal and policy
Next is the NER analysis conducted on all the abstracts structures in shaping AI adoption within HRM. National
found for the WoS dataset. Upon completion of the pro- initiatives, such as those in Romania and Germany, reflect a
cessing pipeline that was previously mentioned, two primary collective recognition of AI’s increasing influence on work-
outputs were generated. First, a CSV contained four main force dynamics. Romania’s approach emphasizes a balance
columns, reflecting each abstract’s text alongside its detected between ethical responsibilities and the practical imple-
AI-related terminology. An excerpt from this file is shown in mentation of AI, aligning with broader EU objectives to
Table 3. As illustrated, each row provides the original abstract promote social well-being while safeguarding data protection
text, any terms that perfectly matched the reference list, terms and human-centric values. Meanwhile, Germany prioritizes
that partially matched (above the similarity threshold) and cross-sector collaboration, with public, private and academic
new or uncategorized phrases that emerged from the NER institutions working together to advance AI-driven HR appli-
process. cations. The establishment of application labs and continuous

81912 VOLUME 13, 2025

C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

employee training programs further demonstrate how inte- AI’s real-world effectiveness and its alignment with organi-
grated governance structures can support the development of zational goals and the values of today’s society. These actions
scalable AI solutions that uphold ethical standards whereas may support the trustworthy and effective integration of AI in
effectively serving organizational and societal interests. public HRM environments.
Furthermore, our analysis reveals substantial variation Strengthening the link between theory and practice
in the extent to which technological progress is adopted becomes essential, as these findings are strongly supported
and institutionalized, particularly in public sector HR cir- by real-world implementations within public sector HR sys-
cumstances. In this context, factors such as bureaucratic tems across several countries. Germany’s national AI strategy
issues, budgetary constraints and conflicting policy objec- integrates AI applications such as public data analysis and
tives may impede the assimilation of even well-established AI-driven administrative processes through innovation labs.
AI tools. These barriers emphasize the need for deeper, Brazil has involved AI development through applied research
context-specific research into the institutional and cultural centers that bridge academic research and government needs,
conditions that either facilitate or hinder AI adoption. contributing to AI deployment in services affecting public
Ultimately, a more integrated approach, wherein technology personnel. Furthermore, the use of AI chatbots for internal
design aligns explicitly with the legal and ethical mandates of HR support and citizen interaction, as seen in public admin-
public institutions, could help bridge this gap. istrations across Europe and Asia, demonstrates a growing
Equally critical is the practical implementation of AI tech- institutional trend toward intelligent HR automation. More
nologies in real-world HR scenarios. While the literature relevant examples are provided across the U.S. related to
abounds with theoretical models and pilot studies, long- faster citizen services, higher productivity and more proactive
term evaluations of AI effectiveness remain insufficient. decision-making and long-term planning. Ten use cases for
Assessing user acceptance, measuring cost-effectiveness, and AI in local government are showcased (Public safety and law
quantifying productivity gains across diverse organizational enforcement; Traffic management and transportation; Citizen
structures will be essential for translating AI’s promise into services; Urban planning and development; Healthcare and
sustainable performance improvements. These future inves- social services; Environmental monitoring and management;
tigations stand to benefit significantly from longitudinal data Emergency response and disaster management; Financial
and robust methodological frameworks, both quantitative and management; Community engagement; Education and work-
qualitative, that enable a nuanced view of AI’s impact on force development) [70]. U.K. Ministry of Defense has begun
employee well-being and organizational culture. using an AI-powered writing tool called Textio to improve
After these analyses, our research reveals that while the its recruitment advertisements. Textio’s system analyzes the
research community is heavily focused on technical AI tools wording of job postings and suggests revisions to make them
and performance metrics, themes such as fairness, bias miti- more inclusive and appealing to a broader talent pool [71].
gation and accountability appear less frequently and unevenly Moreover, U.K. Civil Service Guidance and Oversight is
across studies. This imbalance highlights a need for greater proactively addressing the benefits and risks of AI in recruit-
interdisciplinary collaboration, where technological innova- ment across its agencies. In 2023, the Department for Science,
tion is continuously informed by ethical principles and public Innovation and Technology (DSIT) published ‘‘Responsible
policy frameworks. In the public sector of HRM, this is AI in Recruitment’’ guidance for public sector and private
particularly essential, as AI adoption directly affects civil recruiters [72]. This official guidance notes that AI tools can
service transparency, citizen trust and institutional legitimacy. automate and streamline hiring processes (yielding greater
The findings of our research carry important implications efficiency and consistency). For example, AI might be used
for public sector institutions that are aiming to incorporate for sourcing candidates, screening CVs or even conduct-
AI into their HR functions. Practical applications include the ing AI-scored interviews. However, it also cautions against
deployment of AI tools for automating repetitive adminis- ‘‘novel risks’’ such as perpetuating existing biases or digital
trative tasks, enhancing recruitment accuracy and improving exclusion of those less familiar with the technology. These
employee engagement through predictive analytics and per- examples show that the bibliometric patterns uncovered in our
sonalized training. HR personnel should be equipped with research are reflected in active government reforms, illustrat-
training in data literacy and ethical AI principles, enabling ing the tangible policy relevance of AI-HR convergence.
them to interpret AI-generated insights responsibly, as they By mapping research trends and identifying unaddressed
possess many concerns. These capabilities can help public gaps, this bibliometric analysis provides a foundation for
organizations address inefficiencies, improve transparency advancing the dialogue around AI’s role in modern HR prac-
and deliver services that are centered to their citizens more tices, including those within the public sector. The findings
effectively. highlight both the immense opportunities for efficiency ben-
Moving forward, institutions should consider investing in efits and strategic insights, as well as the need to confront
AI-based HR modules that can upskill their HR professionals, unresolved questions regarding ethics, transparency and prac-
considering that an ethical framework should be established tical feasibility. The urgency for deeper policy integration and
to ensure responsible AI use. Future research should assist cross-sector collaboration becomes imperative, because even
these initiatives through practical case studies that evaluate if the academic momentum is evident, its impact on actual HR

VOLUME 13, 2025 81913

 C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

transformation remains uneven. Without robust mechanisms [3] A. M. Mohamed and A. S. Hameed, ‘‘Trends and opportunities of
to translate research into practice, particularly in the public artificial intelligence in human resource management: Aspirations for
public sector in Bahrain,’’ Int. J. Sci. Technol. Res., vol. 9, no. 1, pp. 1–16,
sector, there is a risk of theoretical progress outpacing prac- Jan. 2020. Accessed: Jan. 21, 2025. [Online]. Available: https://www.
tical implementation. Future work must therefore prioritize researchgate.net/publication/340460650_Trends_And_Opportunities_Of_
interdisciplinary studies that bridge technical advancements Artificial_Intelligence_In_Human_Resource_Management_Aspirations_
For_Public_Sector_In_Bahrain
with organizational realities and public sector constraints. [4] D. N. Roopalatha, ‘‘Artificial intelligence on human resource management-
Bridging these dimensions will require collaborative efforts innovation, challenges and path forward,’’ Educ. Admin., Theory Pract.,
among researchers and policymakers, thus ensuring that AI’s vol. 2024, pp. 13686–13698, May 2024, doi: 10.53555/kuey.v30i5.5980.
[5] E. Farrell et al., ‘‘Artificial intelligence for the public sector,’’ Publications
integration into HR is not only innovative but also just, Office Eur. Union, 2023. [Online]. Available: https://data.europa.
inclusive, and sustainable. eu/doi/10.2760/91814 and https://op.europa.eu/en/publication-detail/-
This research is subject to several limitations. The use /publication/11923574-fdd1-11ed-a05c-01aa75ed71a1/language-en
[6] N. Kaushal and S. Ghalawat, ‘‘Research perspective of artificial intelli-
of pre-trained language models, like the ones used in this gence and HRM: A bibliometric study,’’ Int. J. Bus. Innov. Res., vol. 31,
paper (LDA, BERT for NER) may introduce biases based no. 2, pp. 168–196, 2023, doi: 10.1504/ijbir.2023.131432.
on their training data, potentially affecting the classification [7] N. Kaushal, R. P. S. Kaurav, B. Sivathanu, and N. Kaushik, ‘‘Artificial
of domain-specific terms. Even if large datasets are compre- intelligence and HRM: Identifying future research agenda using systematic
literature review and bibliometric analysis,’’ Manage. Rev. Quart., vol. 73,
hensive, they may contain noisy or tangential entries despite no. 2, pp. 455–493, Nov. 2021, doi: 10.1007/s11301-021-00249-2.
filtering efforts, which could influence topic modeling out- [8] J. J. G. Hernández, G. M. Díaz, and J. L. G. Salvador, ‘‘Artificial intelli-
puts. Also, while the input spans multiple countries and gence applied to human resources management: A bibliometric analysis,’’
in Proc. Int. Conf. Inf. Technol. Syst., in Lecture Notes in Networks and
institutions, the identified trends may not fully generalize Systems, Jan. 2024, pp. 269–277, doi: 10.1007/978-3-031-54235-0_25.
across all public sector contexts due to regional policy differ- [9] P. Mathushan, A. S. Gamage, and V. Wachissara, ‘‘Human resource
ences or varying levels of AI maturity. Addressing these gaps management and artificial intelligence: A bibliometric explorations,’’ Vidy-
odaya J. Manag., vol. 9, no. 1, pp. 1–29, May 2023.
requires the integration of qualitative methods and region- [10] S. B. Devi, ‘‘Digitization in human resources,’’ J. Inventive Sci. Res. Stud.,
specific case studies to ensure that findings are actionable and vol. 2, no. 2, Oct. 2024. Accessed: Jan. 26, 2025. [Online]. Available:
reasonable. https://www.jisrs.com/special
[11] M. Sahoo and S. Mishra, ‘‘A bibliometric analysis of artificial intelligence
and human resource management,’’ in ICT Systems and Sustainability
DECLARATIONS (Lecture Notes in Networks and Systems), vol. 765. Singapore: Springer,
DISCLOSURE STATEMENT (COMPETING INTEREST) 2023, pp. 33–40, doi: 10.1007/978-981-99-5652-4_4.
The authors have no relevant financial or non-financial [12] D. M. Obreja, R. Rughiniş, and D. Rosner, ‘‘Mapping the concep-
tual structure of innovation in artificial intelligence research: A bib-
interests to disclose. liometric analysis and systematic literature review,’’ J. Innov. Knowl.,
vol. 9, no. 1, Jan. 2024, Art. no. 100465, doi: 10.1016/j.jik.2024.
DATA AVAILABILITY STATEMENT 100465.
The data will be made available upon reasonable request. [13] U. Murugesan, P. Subramanian, S. Srivastava, and A. Dwivedi, ‘‘A study
of artificial intelligence impacts on human resource digitalization in
industry 4.0,’’ Decis. Anal. J., vol. 7, Jun. 2023, Art. no. 100249, doi:
ETHICAL APPROVAL 10.1016/j.dajour.2023.100249.
Not applicable. [14] Y. Zhai, L. Zhang, and M. Yu, ‘‘AI in human resource management:
Literature review and research implications,’’ J. Knowl. Economy, vol. 15,
pp. 1–37, Jan. 2024, doi: 10.1007/S13132-023-01631-Z.
INFORMED CONSENT [15] P. R. Palos-Sánchez, P. Baena-Luna, A. Badicu, and J. C. Infante-Moro,
Not applicable. ‘‘Artificial intelligence and human resources management: A bibliometric
analysis,’’ Appl. Artif. Intell., vol. 36, no. 1, pp. 1–28, Dec. 2022, doi:
10.1080/08839514.2022.2145631.
AUTHOR CONTRIBUTIONS [16] A. Alsaif and M. S. Aksoy, ‘‘AI-HRM: Artificial intelligence in human
Cristina Iancu: Methodology, Formal analysis, Investigation, resource management: A literature review,’’ J. Comput. Commun., vol. 2,
Resources, Writing–Original Draft, Writing–Review and no. 2, pp. 1–7, Jul. 2023.
[17] A. Saxena, ‘‘The growing role of artificial intelligence in human resource,’’
Editing, Visualization. Simona-Vasilica Oprea: Conceptual- EPRA Int. J. Multidisciplinary Res., vol. 6, pp. 152–158, Aug. 2020, doi:
ization, Validation, Formal analysis, Investigation, Writing– 10.36713/epra4924.
Original Draft, Writing–Review and Editing, Visualization, [18] D. S. Dixit, K. Kathuria, N. Kumari, K. K. Raina, R. Scholar, and A. Pro-
fessor, ‘‘Role of artificial intelligence in human resource management,’’
Supervision, Project administration.
Int. J. Novel Res. Develop., vol. 7, no. 12, pp. 1–8, Dec. 2022. Accessed:
Jan. 21, 2025. [Online]. Available: www.ijnrd.org
REFERENCES [19] European Union. AI Act | Shaping Europe’s Digital Future. Accessed:
[1] D. Q. B. G. Zetty. (2025). The Role of Artificial Intelligence in Transform- Jan. 25, 2025. [Online]. Available: https://digital-strategy.ec.europa.eu/
ing Human Resource Management Post-COVID-19: Automation, Data en/policies/regulatory-framework-ai
Analytics, and Digitalization. Accessed: Jan. 26, 2025. [Online]. Available: [20] B. D. Sai. The EU is Proposing Regulations on AI- and The Impact on
https://www.researchgate.net/publication/388221729_Article_Title_The_ Healthcare Could be Significant. Accessed: Jan. 25, 2025. [Online].
Role_of_Artificial_Intelligence_in_Transforming_Human_Resource_ Available: https://www.forbes.com/sites/saibala/2021/04/25/the-eu-is-
Management_Post-COVID-19Automation_Data_Analytics_and_ proposing-regulations-on-ai-and-the-impact-on-healthcare-could-be-
Digitalization significant/?sh=16cd73519be6
[2] L. M. T. D. S. Barreto, M. P. D. Silva, A. L. Fischer, L. G. D. Albuquerque, [21] A. Campion, M. Gasco-Hernandez, S. Jankin Mikhaylov, and M. Esteve,
and W. A. C. D. Amorim, ‘‘Emerging issues in human resources manage- ‘‘Overcoming the challenges of collaboratively adopting artificial intel-
ment: An analysis of academic production,’’ Revista de Administração da ligence in the public sector,’’ Social Sci. Comput. Rev., vol. 40, no. 2,
UFSM, vol. 4, no. 2, pp. 215–232, Sep. 2011, doi: 10.5902/198346592401. pp. 462–477, Apr. 2022.

81914 VOLUME 13, 2025

C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

[22] W. G. D. Sousa, E. R. P. D. Melo, P. H. D. S. Bermejo, R. A. S. Farias, and [40] L. C. Kubota and M. B. Rosa, ‘‘Technical note n. 125 (Diset): Artificial
A. O. Gomes, ‘‘How and where is artificial intelligence in the public sector intelligence in Brazil? Adoption, scientific production and regulation,’’
going? A literature review and research agenda,’’ Government Inf. Quart., Federal Government Brazil, Ministery Planning Budget, Tech. Rep. 125,
vol. 36, no. 4, Oct. 2019, Art. no. 101392, doi: 10.1016/j.giq.2019.07.004. 2024, doi: 10.38116/TNDISET125-ENG.
[23] Joint Research Centre. Romania AI Strategy Report—European [41] L. C. Huang, P. Wu, R. J. Kuo, and H. C. Huang, ‘‘A neural network mod-
Commission. Eur. Commission. Accessed: Dec. 2, 2024. [Online]. elling on human resource talent selection,’’ Int. J. Human Resour. Develop.
Available: https://ai-watch.ec.europa.eu/countries/romania/romania-ai- Manage., vol. 1, no. 2/3/4, p. 206, 2001, doi: 10.1504/ijhrdm.2001.001006.
strategy-report_en [42] A. Kankanhalli, Y. Charalabidis, and S. Mellouli, ‘‘IoT and AI for smart
[24] Government of Romania. Strategia Naţională în Domeniul Inteligenţei government: A research agenda,’’ Government Inf. Quart., vol. 36, no. 2,
Artificiale. Accessed: Dec. 2, 2024. [Online]. Available: https://www. pp. 304–309, Apr. 2019, doi: 10.1016/j.giq.2019.02.003.
adr.gov.ro/wp-content/uploads/2024/03/Strategie-Inteligenta-Artificiala- [43] M. S. A. Bhat and D. B. A. Sheikh, ‘‘Impact of digitization on human
22012024-1.pdf resources: Challenges and opportunities,’’ J. Prod., Operations Manage.
[25] M. Oschinski, ‘‘Assessing the impact of artificial intelligence on Econ., vol. 4, no. 42, pp. 1–7, Feb. 2024, doi: 10.55529/jpome.42.1.7.
Germany’s labor market: Insights from a ChatGPT analysis,’’ Univ. Library [44] D. T. Rana, ‘‘Digitization of human resource practices—An emerging
Munich, Munich, Germany, MPRA Paper 118300, 2023. trend,’’ in Proc. 10th Int. Conf. Digit. Strategies Organizational Success,
[26] OECD Artificial Intelligence Review of Germany, OECD Publishing, Paris, Jan. 2019, pp. 1–11, doi: 10.2139/ssrn.3323740.
France, Jun. 2024, doi: 10.1787/609808d6-en. [45] Y. Hu, Q. Li, Y. Chai, D. Wu, L. Lu, N. Shi, Y. Teng, and Y. Zhang, ‘‘AI
[27] A. Fenwick, G. Molnár, and P. Frangos, ‘‘Revisiting the role of HR in service deployment and resource allocation optimization based on human-
the age of AI: Bringing humans and machines closer together in the like networking architecture,’’ IEEE Internet Things J., vol. 11, no. 14,
workplace,’’ Frontiers Artif. Intell., vol. 6, Jan. 2024, Art. no. 1272823, pp. 24795–24813, Jul. 2024, doi: 10.1109/JIOT.2024.3384546.
doi: 10.3389/frai.2023.1272823. [46] P. Mathushan, A. S. Gamage, and V. Wachissara, ‘‘Human resource
[28] L. Z. Knutsen, J. E. Hannay, and M. A. Riegler, ‘‘Artificial intelli- management and artificial intelligence: A bibliometric exploration,’’ Vidy-
gence in the public sector—An agenda for responsible innovation through odaya J. Manage., vol. 9, no. I, pp. 1–29, May 2023, doi: 10.31357/vjm.
learning,’’ in Proc. 7th ACM/IEEE Int. Workshop Softw.-Intensive Bus., v9ii.6370.
Apr. 2024, pp. 1–6, doi: 10.1145/3643690.3648235. [47] S. Ri Na, ‘‘Application of artificial intelligence in recruitment and selec-
[29] A. S. Goga. Inteligenţa Artificiala în Serviciul Public: O Speranţă şi O tion,’’ Academic J. Sci. Technol., vol. 9, no. 2, pp. 56–58, Feb. 2024, doi:
Provocare—JURIDICE.ro. Accessed: Dec. 2, 2024. [Online]. Available: 10.54097/f4gvxp61.
https://www.juridice.ro/716060/inteligenta-artificiala-in-serviciul-public- [48] S. R. Na, ‘‘Application of artificial intelligence in employee training and
o-speranta-si-o-provocare.html development,’’ Math. Model. Algorithm Appl., vol. 1, no. 1, pp. 26–28, Jan.
[30] O. Bobrovskyi, O. Niema, O. Domsha, D. Zayats, and N. Vasiuk, 2024, doi: 10.54097/gg5eemnb.
‘‘Elements of digitalization of public management and administration: [49] B. Biswabhusan and K. Anshul. (Oct. 2023). Impact of Artificial
Information technologies and artificial intelligence in the field of HR,’’ Intelligence on Human Resource Management. Accessed: Jan. 24,
AD ALTA, J. Interdiscipl. Res., vol. 13, no. 2, pp. 151–159, Jul. 2023, doi: 2025. [Online]. Available: https://www.researchgate.net/publication/
10.33543/j.130235.151159. 374418190_Impact_of_Artificial_Intelligence_on_Human_Resource_
[31] D.-F. D. Niloaia, ‘‘Exploring the landscape: A literature review of AI’s Management
impact on human resource management,’’ Eur. J. Public Admin. Res., [50] K. Ishii, ‘‘Comparative legal study on privacy and personal data protection
no. 3, pp. 15–22, 2024, doi: 10.47743/ejpar.2024-3-6. [Online]. Available: for robots equipped with artificial intelligence: Looking at functional and
https://www.researchgate.net/publication/384022271_EXPLORING_TH technological aspects,’’ AI Soc., vol. 34, no. 3, pp. 509–533, Sep. 2019, doi:
E_LANDSCAPE_A_LITERATURE_REVIEW_OF_AI’S_IMPACT_ON 10.1007/s00146-017-0758-8.
_HUMAN_RESOURCE_MANAGEMENT [51] R. Krishnamurthy and K. C. Desouza, ‘‘Big data analytics: The case of the
[32] N. Aoki, ‘‘An experimental study of public trust in AI chatbots in the public social security administration,’’ Inf. Polity, vol. 19, no. 3,4, pp. 165–178,
sector,’’ Government Inf. Quart., vol. 37, no. 4, Oct. 2020, Art. no. 101490, Dec. 2014, doi: 10.3233/ip-140337.
doi: 10.1016/j.giq.2020.101490. [52] C. G. Harris, ‘‘Making better job hiring decisions using ‘human in the loop’
[33] A. Zuiderwijk, Y.-C. Chen, and F. Salem, ‘‘Implications of the use of techniques,’’ in Proc. HumL@ISWC, 2018, pp. 1–10.
artificial intelligence in public governance: A systematic literature review [53] Y.-C. Chen and J. Lee, ‘‘Collaborative data networks for public
and a research agenda,’’ Government Inf. Quart., vol. 38, no. 3, Jul. 2021, service: Governance, management, and performance,’’ Public
Art. no. 101577, doi: 10.1016/j.giq.2021.101577. Manage. Rev., vol. 20, no. 5, pp. 672–690, May 2018, doi:
[34] Y.-P. Yuan, Y. K. Dwivedi, G. W.-H. Tan, T.-H. Cham, K.-B. Ooi, 10.1080/14719037.2017.1305691.
E. C.-X. Aw, and W. Currie, ‘‘Government digital transformation: Under- [54] S. Bankins, P. Formosa, Y. Griep, and D. Richards, ‘‘AI decision making
standing the role of government social media,’’ Government Inf. Quart., with dignity? Contrasting workers’ justice perceptions of human and AI
vol. 40, no. 1, Jan. 2023, Art. no. 101775, doi: 10.1016/j.giq.2022. decision making in a human resource management context,’’ Inf. Syst.
101775. Frontiers, vol. 24, no. 3, pp. 857–875, Feb. 2022, doi: 10.1007/s10796-
[35] R. Madan and M. Ashok, ‘‘AI adoption and diffusion in public admin- 021-10223-8.
istration: A systematic literature review and future research agenda,’’ [55] M. Janssen and G. Kuk, ‘‘The challenges and limits of big data algo-
Government Inf. Quart., vol. 40, no. 1, Jan. 2023, Art. no. 101774, doi: rithms in technocratic governance,’’ Government Inf. Quart., vol. 33, no. 3,
10.1016/j.giq.2022.101774. pp. 371–377, Jul. 2016, doi: 10.1016/j.giq.2016.08.011.
[36] R.-I. Popescu, O. M. Sabie, and M. I. Truşcă, ‘‘The contribution of artificial [56] M. K. Ganeshan, ‘‘Impact of artificial intelligence on human resource
intelligence to stimulating the innovation of educational services and uni- management in the information technology sector,’’ Tech. Rep.,
versity programs in public administration,’’ Transylvanian Rev. Administ. 2022.
Sci., vol. 2023, no. 70, pp. 85–108, Oct. 2023, doi: 10.24193/tras.70e.5. [57] J. Soni, ‘‘A study on the impact of artificial intelligence on human resource
[37] M. Allgood and P. Musgrave, ‘‘Leveraging ChatGPT in public sector management,’’ Int. J. Res. Anal. Rev., vol. 9, no. 2, pp. 149–168, 2022.
human resource management education,’’ J. Public Affairs Educ., vol. 30, Accessed: Jan. 25, 2025. [Online]. Available: http://www.ijrar.org
no. 3, pp. 494–510, Jul. 2024, doi: 10.1080/15236803.2024.2370455. [58] K. Maheswari, P. William, G. Sharma, F. T. M. Ayasrah,
[38] N. Nawaz, H. Arunachalam, B. K. Pathi, and V. Gajenderan, ‘‘The adoption A. Y. B. Ahmad, G. Ramkumar, and A. Shrivastava, ‘‘Enterprise
of artificial intelligence in human resources management practices,’’ Int. human resource management model by artificial intelligence to get
J. Inf. Manage. Data Insights, vol. 4, no. 1, Apr. 2024, Art. no. 100208, befitted in psychology of consumers towards digital technology,’’
doi: 10.1016/j.jjimei.2023.100208. J. ReAttach Therapy Developmental Diversities, vol. 6, no. 10,
[39] R. Brandão. (Apr. 2024). The Current Scenario of Artificial Intelligence pp. 209–220, Sep. 2023. Accessed: Jan. 24, 2025. [Online]. Available:
Development in Brazil Mapping Artificial Intelligence Centers in https://jrtdd.com/index.php/journal/article/view/1338
Brazil: Initiatives, Actions, and Projects. Accessed: Jan. 21, 2025. [59] J. B. Bullock, ‘‘Artificial intelligence, discretion, and bureaucracy,’’
[Online]. Available: https://bv.fapesp.br/en/604/engineering-research- Amer. Rev. Public Admin., vol. 49, no. 7, pp. 751–761, Oct. 2019, doi:
centersapplied-research-centers-cpecpa/ 10.1177/0275074019856123.

VOLUME 13, 2025 81915

 C. Iancu, S.-V. Oprea: AI and HR in a Literature-Driven Investigation Into Emerging Trends

[60] Q. Neshat, ‘‘The impact of artificial intelligence in human resource deci- [71] R. Booth. Amazon-hosted AI Tool for U.K. Military Recruitment ‘Carries
sion making processes,’’ J. Pharmaceutical Negative Results, vol. 13, Risk of Data Breach’ | Artificial Intelligence (AI) | The Guardian.
no. 7, pp. 1–6, Dec. 2022. Accessed: Jan. 24, 2025. [Online]. Available: Accessed: Apr. 25, 2025. [Online]. Available: https://www.theguardian.
https://pnrjournal.com/index.php/home/article/view/5724/7049 com/technology/2024/dec/17/amazon-hosted-ai-tool-for-uk-military-
[61] R. Anuradha and D. Uma, ‘‘The impact of artificial intelligence in recruitment-carries-risk-of-data-breach
HRM: Assessing recruitment, performance management, and employee [72] U.K. Government. Responsible AI in Recruitment—GOV.UK. Accessed:
experience,’’ J. Jilin Univ., vol. 42, no. 10, pp. 17–32, 2023, doi: Apr. 25, 2025. [Online]. Available: https://www.gov.uk/government/
10.5281/zenodo.10012057. publications/responsible-ai-in-recruitment-guide/responsible-ai-in-
[62] M. Marius, D. Marin, S. A. Gabriela, and S. Cristian, ‘‘The technolog- recruitment
ical transition of European manufacturing companies to industry 4.0. is
the human resource ready for advanced digital technologies? The case
of Romania,’’ Econ. Comput. Econ. Cybern. Stud. Res., vol. 55, no. 2,
pp. 23–41, Jun. 2021, doi: 10.24818/18423264/55.2.21.02.
[63] M. Jatobá, J. Santos, I. Gutierriz, D. Moscon, P. O. Fernandes, and
CRISTINA IANCU received the B.Sc. degree in
J. P. Teixeira, ‘‘Evolution of artificial intelligence research in human
resources,’’ Proc. Comput. Sci., vol. 164, pp. 137–142, Jan. 2019, doi:
economic cybernetics, statistics, and informatics,
10.1016/j.procs.2019.12.165. in 2022, and the M.Sc. degree in database admin-
[64] M. Z. Shaikh, M. S. S. Raina, K. Jayapriya, and A. R. N. Chander, istration for business from Bucharest University of
‘‘Artificial intelligence and public sector human resource management: Economic Studies, Romania. She is currently pur-
Opportunities, challenges,’’ J. Electr. Syst., vol. 20, no. 4s, pp. 98–104, suing the Ph.D. degree with the Doctoral School of
Apr. 2024, doi: 10.52783/jes.1858. Economic Informatics, researching AI in human
[65] B. Hmoud, ‘‘Will artificial intelligence take over human resources recruit- resources. She is a Java Developer. She has won
ment and selection?’’ Dept. Editorial, Network Intell. Stud., Romanian multiple awards in digitalization competitions and
Found. Bus. Intell., Tech. Rep. 13, pp. 21–30, Jul. 2019. published her first research paper in the IE2024
[66] K. M. Gasana, ‘‘The impact of artificial intelligence on human Conference (Springer).
resource management,’’ Res. Output J. Arts Manage., vol. 3,
pp. 14–18, Jan. 2024. Accessed: Jan. 24, 2025. [Online]. Available:
https://www.researchgate.net/publication/383084162_The_Impact_of_
Artificial_Intelligence_on_Human_Resource_Management
[67] OECD. AI Strategies and Policies in Korea. Accessed: Jan. 6, 2025.
[Online]. Available: https://oecd.ai/en/dashboards/countries/SouthKorea SIMONA-VASILICA OPREA received the M.Sc.
[68] G. Webster, R. Cree, E. Kania, and P. Triolo. (2017). Full Translation: degree in infrastructure management program
China’s ‘New Generation Artificial Intelligence Development Plan’
from Yokohama National University, Japan,
(2017). Accessed: Jan. 6, 2025. [Online]. Available: https://digichina.
in 2007, the Ph.D. degree in power system engi-
stanford.edu/work/full-translation-chinas-new-generation-artificial-
intelligence-development-plan-2017/ neering from Bucharest Polytechnic University,
[69] P. S. K. Pahul. Interpreting Topic Model Visualization—LDAvis Package in 2009, and the Ph.D. degree in economic infor-
| Pahul Preet Singh Kohli. Accessed: Jan. 18, 2025. [Online]. Available: matics from Bucharest University of Economic
https://pahulpreet86.github.io/interpreting-topic-model-visualization- Studies, in 2017. She is currently an Associate Pro-
ldavis-package/ fessor with the Faculty of Cybernetics, Statistics,
[70] M. Jackley. Using AI in Local Government: 10 Use Cases. Accessed: and Economic Informatics, Bucharest University
Apr. 25, 2025. [Online]. Available: https://www.oracle.com/artificial- of Economic Studies, involved in several research projects.
intelligence/ai-local-government/?utm_source=chatgpt.com

81916 VOLUME 13, 2025