=Paper=
{{Paper
|id=Vol-2548/paper-06
|storemode=property
|title=Ontologies and AI in Recruiting. A Rule-Based Approach to address ethical and legal
auditing
|pdfUrl=https://ceur-ws.org/Vol-2548/paper-06.pdf
|volume=Vol-2548
|authors=Carmen Fernández Martínez,Alberto Fernandez
}}
==Ontologies and AI in Recruiting. A Rule-Based Approach to address ethical and legal
auditing==
https://ceur-ws.org/Vol-2548/paper-06.pdf
Ontologies and AI in Recruiting. A Rule-Based Approach
to Address Ethical and Legal Auditing
Carmen Fernández Martínez1[0000-0003-4514-4609] and Alberto Fernández2[0000-0002-8962-6856]
CETINIA, Universidad Rey Juan Carlos, Madrid, SPAIN
1
carmen.urjc@gmail.com
2
alberto.fernandez@urjc.es
Abstract. Artificial Intelligence (AI) domain-specific applications may have
different ethical and legal implications depending on the domain. One of the
current questions of the AI is the challenges behind the analysis of job video-
interviews. The use of semantic descriptions of jobs positions and candidate
profiles could improve Recruiting information management within the
organization and candidate-position matching. There are additional
controversial issues, pros and cons to using AI in recruitment processes, and
potential ethical and legal consequences for candidates, companies and states.
There is a deficit of regulation of these systems, and a need for external and
neutral auditing of the types of matching made in interviews to reduce potential
discrimination, for example on the basis of race or gender, in the job market.
We propose, first, formally define criteria for jobs and candidates using a
candidate desired skills and emotions ontology and job offers ontology to foster
interoperability at company level and a multi-agent system architecture for
neutral auditing to guarantee a fair, inclusive and accurate AI.
Keywords: Domain Specific AI, Ethics, Human Resources, Ontology.
1 Introduction to research question and relevancy
Traditionally, Ai proved very valuable for resume and keywords scanning and for
extraction of candidate skills devoid of bias. There has been a recent trend towards
video-interview analysis in Human Resources. The survey by Personnel Today found
that 38% of enterprises are already using AI in their workplace with 62% expecting to
use it by 2018. In this research, we address such a current issue of the AI, the use of
Machine learning techniques for analysis.
Concerning video-interview systems, there are limitations, some are attributable to
the very nature of the technology (incorrect or biased datasets) and other are related to
the human bias or the specific agenda of the recruiting company. However, the state-
of-the-art in image analysis may allow pre-selecting with respect to age or sexual
orientation or other controversial characteristics. The analyses could lead to ethical
and legal consequences (e.g. in some countries is forbidden to ask for age in
processes). This is why fostering proper auditing of video-interview systems it is
particularly important.
Copyright © 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
The initial entry point of the system and settings will be the set of interview
questions designed by Human Interviewer, later distributed to the hiring candidates.
The output will be a ranking/list of candidates plus reports and warnings. The video
interview produced in subsequent stages will be subjected to general and customized
analysis. Several features will be measured and tracked according to the recruiter
choices, to name but a few: Is there only a person appearing in the video? How long
does it take to answer (candidate time)? Does the candidate look directly into the
camera (eye time)? Does the candidate have a high score of percentage in commercial
profile. As a result, the analyses could track global behaviour traits like attention to
detail (eye contact time, emotions-intonation and body language).
Results will be clustered, later on, to finally being analyzed with AI using
techniques like K-NN, Adaboost or Neural Networks.
2 Problem statement
There has been a recent trend towards video-interview analysis in HR departments.
Traditionally, AI played no more than an assistant role in HR, e.g. resume and CV
scanning. But lately, apps and systems like HireVue 1 , Montage 2 , SparkHire 3 and
WePow 4 have been changing how recruitment is carried out. An AI-based video
interview system analysed with Machine Learning techniques could be programmed
to check, during an interview, features such as age, lighting, tone of voice, cadence,
keywords used (substantial conversation), mood, behaviour (eccentric, movement or
quite calm and not talkative), eye contact and, above all, emotions. AI targets the
specific traits of a customer-oriented role that employers want in their teams.
AI has produced benefits for HR so far, including recruiting time and customised
questions and answers and lack of interviewer bias (physical appearance, tattoos, etc.)
But there are several problems that accompany the use of these technologies; for
example, candidates are unfamiliar with video-interview analysis (for example,
lighting, settings), which could affect global performance. It is necessary to point out
the Gender and racial bias and the imprecision of technology. Traditionally, machine
learning algorithms were trained with data from white people or biased datasets.
We studied several potential controversial characteristics, among them, facial
symmetry, race, gender, sexual orientations in voice and image recordings. The
problem of racial-bias in AI is not new, just like the detection of mixed race in bad
lighting conditions according to Siyao et al. [1]. As an illustration of the advances in
sexual orientation recognition both in images and sound, one study [2] needed ethical
supervision due to the opaque invasive nature of the research and the use of real user
data from dating applications. Researchers argue that there is a relationship among
homosexuality, morphological features e.g. jawline and exposure to particular
concentrations of hormones in the womb.
With reference to Ethical and legal aspects of AI, whilst the use of AI in this
context may have its benefits, it also strips away aspects of humanity, reducing a
human recruit to a set of descriptors. The automation of HR processes could lead to
potential ethical and legal implications that cannot be ignored. In some countries,
companies are not allowed to ask a candidate's age during recruitment. Especially in
1
https://www.hirevue.com/
2
https://www.montagetalent.com
3
https://www.sparkhire.com
4
https://wwww.wepow.com/es
the United States where is forbidden to "improperly classify or segregate employees
by race". (US Civil Rights Act, 1964).
Given the illustration of the problem above mentioned, we want to achieve
automatisation of analyses in Human Resources that guarantee compliance and
auditing of the questions posed to the candidates considering differences in legislation
and national scenarios. Secondly, we aim for accuracy and suitability of the data sets
used for training and test of learning algorithms, looking if it is fair to extrapolate
their classification results to different regions and backgrounds. If not possible, the
option of notifying by means of warning the irregularities in datasets or in an
interview format. We then will elaborate on the assumption that Agents are the right
paradigm to support this problem. Agents could be defined as distributed nodes of
information that could carry out the automatic analysis independently from one
another. The software agents of our proposal assist the Human Resources employees
and candidate in the automatisation and completion of tasks. Finally, we strongly
believe in the benefits of adding semantic Web knowledge to the agents. For instance,
HR domain-specific ontologies could not improve interoperability or reuse of
information. The could also very well boost the ability of learning and therefore better
and more accurate reasoning and automatisation, helping human employees in their
knowledge discovery and decision process, duties that could be taken over by a new
Discovery agent in the MAS architecture.
3 Related work
The complicated issue of handling of unstructured data in big organisations and the
ontology-based solutions to deal with it is a recurring topic in previous work. In the
specific case of Human Resource data, the proposed solutions are mainly related to
the extraction of information from Resumes [5] and ontology-based information
extraction system for matching résumés to job openings [6].
Regarding the implementation of the auditing system, Multi-agent systems pose
similar challenges concerning information formats that traditional complex distributed
systems and the problem is more acute in firms.
The current approach is the definition of agents tasks and implementations and the
exploration of both MAS and ontologies and shared lexicons, which foster domain
specification and interoperability, an idea supported in previous works [3] so as to the
introduction of MAS in complex corporate settings such as manufacturing industry.
The idea of coordination and interoperability of agents in heterogeneous domains
has been widely used in many different domains, such as health care, emergencies due
to natural disasters [8], smart cities [9], etcetera. Multi-agent systems (MAS) have
also been used to improve business processes in complex organisations. Practicalities
of big multinational corporations, the difference in formats and regulations make it
necessary to create vocabularies and ontologies, architectures and models to rule over
changing specific agent populations. Architectures have been proposed to deal with
the random and quick changes of a particular productive section, such as the
manufacturing industry but could be applied to other dynamic corporate environments
[10]. The recruiting scenario in a multinational context is also quick and complex and
needs models. So far there is no much literature related to applications of HR and
MAS as enabling technology, but these types of architectures have been extensively
used, as noted above, in manufacturing and corporate control production (e.g. Ciortea
et al. [10]).
Additionally, we supported our research on previous works in legal formalization
[4]. The domain of Law or Legal formalization is open to MAS applications and has
been addressed over the last decade by Law scholars such as Walker. It is important
for rules-based systems a proper model of the legal rules. Legal requirements
modelling (as knowledge acquisition) is a specialized field nowadays, celebrating
conferences such as International Conference on Legal Knowledge and Information
System (JURIX).
4 Research question
RQ1: How to achieve an improvement of interoperability in Recruiting. Is it
applicable to a use case/controlled business scenario?
RQ2: How to achieve accurate automatization in domain Human Resources.
RQ3: Is it possible to implement a comprehensive legal auditing or compliance
system given the current state-of-the-art?
The first research question is related to the improvement of interoperability, full
automatization and better ethical and legal auditing in Recruiting. To promote the
interoperability we will follow an ontology-based approach for information extraction
from video-interview systems analyzed with Artificial Intelligence. Then it will be
possible to extract the information of interest regarding the candidate and the traits
and competencies analyzed in the video-interview and reformulate all in a structured
document to make easier the auditing and compliance process with Labour Law.
Similarly, for the third research question, the introduction of ethical and legal
auditing test in the domain we will consider Multi-Agent System, rules engines to
prove the appropriateness of this approach.
5 Hypothesis and Proposal
The hypotheses behind our research is that automatization of tasks, the use of
ontologies and semantic descriptions and the introduction of proper auditing could
improve the business processes in Human Resources. The research questions outlined
in the previous section will be addressed following these hypotheses:
(RQ1) H1: Describing semantically interview formats, CVs and job positions
makes easier attempting to do ontology matching and fostering interoperability in a
cross-cultural international business scenario.
(RQ1) H2: The extraction of facts, entities and relations from unstructured data
like video-interview stream or images is a manageable problem, given the current
state-of-the-art in visual analyses.
(RQ2) H1: The automated analyses and reasoning in Human Resources are
tangible.
(RQ2) H2: The inclusion of search and discovery, learning capabilities in HR
software agents are tangible.
(RQ3) H1: The rule-based approach can produce a fair ethical and legal test. The
rule-based approaches are a more efficient solution to implement a compliance
system.
(RQ3) H2: MAS is the right paradigm to handle legal auditing in a Human
Resources environment.
To prove the above-mentioned hypotheses and solve the research problem of
interoperability I am going first to approach the creation of ontologies for jobs
positions, résumés and characteristics related to the analysis of facial expressions in
video-interview. The inclusion of semantic technology, the better definition of
international corporate terms, in general, the knowledge representation, could indeed
be helpful in discovery and searches of information. The software agents could,
therefore, be able to show learning capabilities, e.g inclusion of discovery agents in
the architecture.
Most AI data analytics applications for Recruiting are based totally on unstructured
or not categorized data. Once the proposal of ontology for this domain is finished, it
will favor the automatization without human intervention and the automatic legal
auditing of the recruitment processes. The most relevant innovation of this proposed
work will be the attempt to automatise part of this legal and ethical auditing as well as
some analyses carried out by the different distributed agents. It has been attempted in
previous work in corporate business processes, not HR. Many sectors have adopted an
intelligent process automatisation technology so far. Activities like data extraction,
creation of reports without human intervention, access to databases are now usual.
The specific contribution and progress made thus so far is mainly the proposal of
semantic descriptions for Human Resources and a Multi-agent systems architecture
for auditing HR [Fig 1]. Additionally, there has been prototyping of some parts -legal
rules engine.
With reference to the ontologies for Human Resources, we have found essential to
define the traits in relation to competencies assessed in a recruitment process. For
example, the trait Engagement analysed by emotions detector Affectiva and others is
the result of very different candidate features. The systems test the candidates on
grounds of different emotions that could be a potential match for the jobs positions,
emotions such as Anger, Contempt, Disgust, Engagement, Joy, Sadness, Surprise and
Valence.
𝑒𝑛𝑔𝑎𝑔𝑒𝑚𝑒𝑛𝑡 = ∃ℎ𝑎𝑠𝐵𝑟𝑜𝑤𝑅𝑎𝑖𝑠𝑒. 𝐵𝑟𝑜𝑤𝑅𝑎𝑖𝑠𝑒∃ℎ𝑎𝑠𝐵𝑟𝑜𝑤𝐹𝑢𝑟𝑟𝑜𝑤. 𝐵𝑟𝑜𝑤𝐹𝑢𝑟𝑟𝑜𝑤 ∩
∃ℎ𝑎𝑠𝑁𝑜𝑠𝑒𝑊𝑟𝑖𝑛𝑘𝑙𝑒. 𝑁𝑜𝑠𝑒𝑊𝑟𝑖𝑛𝑘𝑙 … ∃ℎ𝑎𝑠𝐿𝑖𝑝𝐶𝑜𝑟𝑛𝑒𝑟𝐷𝑒𝑝𝑟𝑒𝑠𝑠𝑜𝑟. 𝐿𝑖𝑝𝐶𝑜𝑟𝑛𝑒𝑟𝐷𝑒𝑝𝑟𝑒𝑠𝑠𝑜𝑟
… ∃ℎ𝑎𝑠𝐶ℎ𝑖𝑛𝑅𝑎𝑖𝑠𝑒. 𝐶ℎ𝑖𝑛𝑅𝑎𝑖𝑠𝑒 … ∃ℎ𝑎𝑠𝐿𝑖𝑝𝑃𝑢𝑐𝑘𝑒𝑟. 𝐿𝑖𝑝𝑃𝑢𝑐𝑘𝑒 … ∃ℎ𝑎𝑠𝐿𝑖𝑝𝑃𝑟𝑒𝑠𝑠. 𝐿𝑖𝑝𝑃𝑟𝑒𝑠𝑠
… ∃ℎ𝑎𝑠𝑀𝑜𝑢𝑡ℎ𝑂𝑝𝑒𝑛. 𝑀𝑜𝑢𝑡ℎ𝑂𝑝𝑒𝑛 … ∃ℎ𝑎𝑠𝐿𝑖𝑝𝑆𝑢𝑐𝑘. 𝐿𝑖𝑝𝑆𝑢𝑐𝑘 … ∃ℎ𝑎𝑠𝑆𝑚𝑖𝑙𝑒. 𝑆𝑚𝑖𝑙𝑒
Engagement implies gestures such as brow raise, brow furrow, nose wrinkle, lip
corner depressor, chin raise, lip pucker, lip press, mouth open, lip suck and smile.
Other features such as the estimated age of the candidate age are listed in
categories or rank like enumerated types. Face.Age (18-24, 25-34, 35-44, 45-54, 55-
64, 65_PLUS, under_18, unknown)
5.1 Towards automated support for auditing HR. Rules-based approach
The main objective of the auditory will be the legal check carried out by Labour Law
agent when necessary. We consider a rules-based approach to achieve this. It is
assumed that the interviews will have a great amount of confidential information and
private responses but when necessary, they will be processed by the external Law
agent to make sure there are no infringements of Labour Law of the countries
carrying out the process.
In the first place, with the aim of simplifying our reasoning, it was proposed to
write up simple Labour Law measures existing in a particular country using the
format of rule and include/exclude as a result of executing the rule. The first analysis
for simplicity will be checking the legal age of the candidate and then the analysis
will focus more specifically on advanced criteria to avoid discrimination. For
example, biased decisions tracked in the Natural Language processing of interview
towards hiring individuals of certain age <40 or native speakers only, or women only
or individuals of certain sexual orientation. The bias will be displayed in the form of
informative warnings.
E..g Spanish Labour Law rule format
Age>18=>include
Age<=18 AND emancipated AND consent parents/tutor=>include
Age<=18 AND not emancipated AND non authorized =>exclude
Age <=18 AND not emancipated AND authorized mother AND consent
father=>include
Age <=18 AND not emancipated AND authorized father AND consent
mother=>include
Age <=18 AND not emancipated AND total orphan AND authorized legal
tutor=>include
Age <=18 AND not emancipated AND total orphan AND not authorized legal
tutor=>exclude
Age <=18 AND not emancipated AND partial orphan AND authorized
mother=>include
Age <=18 AND not emancipated AND partial orphan AND authorized father
=>include
Age <=18 AND not emancipated AND partial orphan AND not authorized
mother=>exclude
Age <=18 AND not emancipated AND partial orphan AND not authorized father
=>exclude
Age>60 AND legally retired=>exclude
Age>60 AND not legally retired=>include
Age>60 AND legally retired AND active retirement status=>include
The Ethical check carried out after pre-selecting individuals by gender or age
merely based on fiscal benefit for the employers and positive discrimination.
Age<30 AND first employment=> Warning
Age>45 AND long-term unemployment=>Warning
Gender=Woman AND domestic violence victim=>Warning
Terrorism victim=>Warning
As previously mentioned, we have resolved to include a rule engine to implement
the LegalClassifier. Rule engines are used nowadays to filter spam emails or to
restart passwords, subscribe/unsubscribe user.
We intend to codified first Laws from Codes, Statutes and formal sources, resulting in
up to 100 rules organized in rule files specific for every country. The rules files are
based on the salience of rules, that means that the very basic rules like legal age to
enter the workplace will have greater salience values than the rules that constitute
legal and ethical warnings. Any other different sources of Law or Regulation, like
specific collective bargain agreements, could be considered in future releases.
The implementation is just for interview text format and is supported by files
including Labour Law of Spain, US or others. The interview information will be
integrated with candidate data and interview responses in the later stages of our
research. The output of the classifier will be include, exclude meaning if an interview
format is not compliant with the basic legal rules and needs to be discarded.
The warning option is essential to contemplate the cases that are being
discriminatory or biased to a certain minority group- but that not break any existing
law. The idea is the implementation of neutral auditing but it is understandable that
countries promote and work towards access to the job market of women, unemployed
youth and minorities.
It is evident that the legal issues concerning candidates, such as the identification of
race and sexual orientation in the selection process due to the advances in image
processing, entails ethical questions that cannot be ignored. Careful analysis from
auditing bodies, governments, ethics committees and psychologies is needed. In this
section, we describe a proposal of a multiagent software architecture for auditing,
which is depicted in Fig. 1.
Fig. 1 shows an abstract MAS architecture that must be adaptable to different
international corporate environments and by recruiters of different nationalities in
search of international compliance. The challenge of storing legal knowledge and
doing sound checks and reasoning become complicated in cross-jurisdictional cases.
Fig. 1. Multi-agent systems architecture
The core of the architecture comprises three different parties that must collaborate:
(i) a recruiter/company, (ii) external auditor, and (iii) government/authorities. An
Interview design agent, based at the company central headquarters, is responsible for
designing a general interview. The first action in the use case would be the opening of
a new general position by the HR manager. The interview design agent will later
translate the requirement to design customized interviews applying the general
interview format to a regional scenario of the country where the recruiting is taking
place. The Interview auditing agent is based in company branches and will watch the
specific national requirements necessary for the process. The Selection process agent
can cancel the process due to controversies or give back a list of candidates to the
central office if the process is fair. It is also capable of running checks with authorities
and auditors. If the features analysed in the recruiting process break any law or if the
process contravenes basic civil rights, the interview process agent would ask for the
approval of the Labour Law Agent or Ethical Agent if necessary. If the recruiting
process is dealing with a candidate’s personal information, it would require the
candidate’s approval.
Interview generator agent. In order to design and compare different and fair
interviews, an interview design agent is required. It would be located next to the
company main headquarters. According to company requirements, some interview
profiles could be designed to be recruited thereafter internationally
(e.g. software engineer, marketing expert). The input of the agent would be certain
characteristics of the open position, controversial or not, (experience, commercial
role, age>40, gender=woman, English native speaker, Spanish native speaker). The
output would be the interview questions in natural language for a particular role
together with scores and answers and rankings expected for the role (optimism 50%,
commercial profile 60% etcetera). It is expected that the total implementation will
require the use of technologies such as RDF or ontologies as a starting point to define
and harmonize the formats. Once the interview has been designed is submitted to the
database to be reviewed thereafter by the in-company auditing agent.
Interview in-company auditing agent. The Interview in-company auditing agent
manages the selection processes in different branches. The human recruiters select a
particular generalist interview from the corporate database for a concrete role, for
instance, analyst. It applies the general format to every regional scenario, e.g. Europe,
Asia Pacific, and the Middle East, adapting the candidate criteria to be selected to
culture. The agent controls interactions among human recruiters and decides if it is
essential to pass interviews with additional stages if they handle controversial data.
The Input would be a generic interview format (later redistributed among HR team)
and output Boolean true/false if it needs auditing. In the cases when the information
needs further validation it is passed to other agents. All the legal and ethical
transactions are hosted outside of the company, in external auditing bodies and
governmental premises. The recruiting process could end at this very point if the
questions are simple and do not need legal handling.
Selection Process agent. The Selection Process agent is the main element of the
proposed Multiagent System. It is hosted near the company branch where the
selection process is being held. It is in charge of processing the different events that
occur in the system (new interviews, new auditing and legal checks, authorities, etc.),
triggering new events when necessary and providing a ranking of candidates to the
human interviewer. It could occasionally reuse interviews from other companies, it
coordinates different companies, external neutral auditors and authorities, but only if
necessary. In case no external auditing is needed –no controversial issues are found-
the Selection Process agent provides direct feedback of the candidate.
Candidate data Check Agent. Once the candidate dossier arrives at the external
auditors it may happen that, due to the sensitivity or inaccuracy of the data, some
confirmation or permission from the candidate is required to proceed with the data
handling. Since the main purpose of this Multi-Agent System is to assess the selection
process in each and every phase the Candidate data Check Agent would inform
promptly the company side if a candidate denies confirmation or does not give
consent to the data.
Labour Law external auditing Agent. The human auditor could optionally proceed
with legal tests. The analyses are very convenient since the Labour Law agent runs
different checks in different legislation, specifically focusing on Employment Laws.
The Labor Law agent allows to easily adapting to new legal scenarios considering
multiple countries and jurisdictions. In the case of law infringement, it would inform
the main agent, i.e. the Selection process agent.
Ethical agent. In the case the user data are very controversial and address ethical
issues, e.g. the interview is designed to check for age or sexual orientation it would be
necessary to complete a check by the Ethical agent, which would support the analysis
with ethics committees and experts to make a verdict on the excessive invasion of the
interviewee privacy. The output of this agent would be ethical approval or
disapproval. The ethical agent would be addressed by the interview auditing agent. In
case it checks any subtle irregularity in candidate criteria database, both positive and
negative discriminations, the auditing agent launches a check for the ethical agent.
E.g. candidate and homosexual=>exclude, candidate and heterosexual=>include.
Candidate and transsexual=>include.
Authorities agent. In the case a company is expatriated and recruiting international
candidates, an Authorities agent coordinates different government’s decisions and
must decide whether a company must comply with additional proceedings or if it is
needed to check register of a foreign candidate in a city census.
Once the final check is completed the information flows back to the Selection process
agent who is responsible for closing the selection process or cancelling it on
legal/ethical grounds.
6 Preliminary Results and/or Evaluation Plan
Concerning limits, so far I am in the process of producing theoretical work and
submitted automatization of legal auditing/ prototyping of legal rules engine
concerning Spanish and US Law. The proposal has been widely accepted by my
scientific department and a brief description of the project published in ERCIM news,
a publication gathering relevant research carried out by European research groups.
As the doctoral work advances, we will see if we could address in all detail the
architecture or just the automatization of the legal auditing if it involves hundreds of
rules. There is an absence of affordable and limited testing opportunities of real
corporate applications so we are counting on prototyping on a simulated corporate
scenario. The testing is very dependent on data availability, use of different ontologies
and the real attributed measured by video-interview systems. Legal reasoning entails
the correct formalization of laws and technical supervision of jurists.
The refinement of the algorithm for detection of bias would seem necessary here in
our evaluation plan. For instance, if a candidate is excluded for valid reasons but
happens to also be homosexual, the system should decipher whether exclusion is a
result of bias or not.
The rules engine so far includes rules for detection of candidates allowed to work
according to Spanish legislation, considering age, underage workers with a permit to
work, active retirement and the query of race and sexual orientation for unethical
purposes. In Spanish legislation, the candidate is not often asked about personal issues
not even for survey purposes, like in the United Kingdom. One supposes that an
expert in Law should translate the legal language of various legislations into rules. In
addition, I have a background in Law but I consider this process does not entail major
difficulty because laws are reduced to a very simple set of preconditions (age>18 and
so on).
At this point of research, we are using a case study to demonstrate the workflow
through the entire system just as prototyping in a simulated corporate environment.
We will soon begin the test with real corporate data transferred from other academic
departments. The main emphasis for improvement should be on the ethical agent and
more detailed analyses that extrapolate the results from systems Affectiva and
HireVue, that documented very well the traits measured, to all HR products.
7 Approach
There is not much work directly in this area at the moment and we are trying to
consolidate a broad range of thinking linked to the topic. In an attempt to address the
research questions, we are keen on breaking new ground in critically examining both
current and potential issues with the use of AI in HR.
The possibility of reusing ontologies was examined in the early stages. The
existing standards and popular ontologies in Human Resources describe many aspects
of job positions, sectors and activities [11] but need to be up to date to the current
application of technology in the sector. We finally consider not to implement them
from scratch. Nevertheless, the field of investigation we consider, the analysis of
video-interview through Machine learning techniques, the interview formats and so
on, requires the extension of these ontologies.
8 Reflections
As a conclusion, we summarize the expected results. The contribution of this PhD
is, therefore, the automatization. The benefits of our research will be a step forward in
the accurate automatization of tasks in Human Resources and ethical and legal
analyses of video-interview techniques. The trend is towards full automatization. The
work could break new ground in critically examining and auditing the video-interview
systems so as in providing semantic description and ontologies for job
positions/interviews and candidate criteria and competencies. Due to the global nature
of the companies, this could mean a step forward in interoperability in an organization
and third parties, states and auditing bodies.
Taken into account all that have been stated, it should be pointed that my approach
gathers evidence of real video-interview systems analyzed with AI and includes is
based on a thorough analysis of HR business processes, going beyond other
approaches that deal only with the information extraction of résumés, Linkedin pages
and personal blogs and matchmaking of job positions/candidate résumés. Considering
the additional detailed study of Labour Law in different legislation it is to be expected
that the rules engine is precise and likely to be applicable to real scenarios. Finally,
we will prove the effectiveness of a rules-based approach for this system.
Acknowledgments.
Work partially supported by the Spanish Ministry of Science, Innovation and
Universities, co-funded by EU FEDER Funds, through grants TIN2015-65515-C4-4-
R and RTI2018-095390-B-C33.
References
1. Siyao F., Haibo H., Zeng-Guang H. Learning from Face: A Survey. IEEE Trans. Pattern
Anal. Mach. Intel 36(12): 2483-2509 (2014).
2. Kosinski M. and Wang. Y. Deep neural networks are more accurate than humans at
detecting sexual orientation from images. Journal of Personality and Social Psychology,
114(2):246–257, (2018).
3. DiLeo J., DeLoach S. Integrating ontologies into multiagent systems engineering. Air
Univ Maxwell AFB Al centre for Aerospace Doctrine Research and Educ. (2006).
4. Walker, V.R.. A default-logic framework for legal reasoning in multiagent
systems. AAAI Fall Symposium - Technical Report. 88-95. (2006).
5. Darshan, P. M. Ontology-based information extraction from resume. In 2017
International Conference on Trends in Electronics and Informatics (ICEI) (pp. 43-
47). IEEE. (2017).
6. Celik, D. Towards a semantic-based information extraction system for matching
résumés to job openings. Turkish Journal of Electrical Engineering & Computer
Sciences, 24(1), 141-159 (2016).
7. Ossowski, S., Omicini, A,. Coordination knowledge engineering. Knowledge
Engineering Review, 17(4), 309–316. (2002)
8. E. Mas, A. Suppasri, F. Imamura, S. Koshimura. Agent-based simulation of the 2011
great east Japan earthquake/tsunami evacuation: An integrated model of tsunami
inundation and evacuation, Journal of Natural Disaster Science, 34(1), 41-57 (2012).
9. Billhardt, H., Fernández, A., Lujak, M., Ossowski S., Agreement Technologies for
Coordination in Smart Cities. Applied Science, 8(5), 816. (2018)
10. Debenham, J., A Multi-Agent Architecture for Business Process Management Adapts to
Unreliable Performance, In: Parmee I.C. (eds) Adaptive Computing in Design and
Manufacture, V. Springer, London, DOI: 10.1007/9780-85729-345-9_31 (2002)
10. Ciortea, A., Mayer, S., Michahelles, F., Repurposing Manufacturing Lines on the fly
with Multi-agent Systems for the Web of Things, In AAMAS ’18: Proceedings of the 17th
International Conference on Autonomous Agents and MultiAgent Systems, Proceedings of
the 17th International Conference on Autonomous Agents and Multiagent Systems, (pp.
813-822), (2018)
11. Gómez-Pérez, A., Ramírez, J., Villazón-Terrazas B,. An ontology for modelling
human resources management based on standards. International Conference on
Knowledge-Based and Intelligent Information and Engineering Systems. Springer,
Berlin, Heidelberg (2007)