Download Competence Gap Analysis Model: Representing and Matching Competences and Profiles and more Lecture notes French Language in PDF only on Docsity!
A MODEL FOR COMPETENCE GAP ANALYSIS ∗
Juri L. De Coi, Eelco Herder, Arne Koesling, Christoph Lofi,
Daniel Olmedilla, Odysseas Papapetrou, and Wolf Siberski
L3S Research Center and University of Hannover, Hannover, Germany { decoi,herder,koesling,lofi,olmedilla,papapetrou,siberski } @L3S.de
Keywords: (^) competence, competency, model, gap analisys, lifelong learning
Abstract: (^) Modeling competences is an integral part of many Human Resource (HR) and e-Learning related activities. HR departments use competence descriptions to define requirements needed for performing specific tasks or jobs. The same competences are acquired by employees and applicants by e.g. experience or certifications. Typically, HR departments need to match such required and acquired competences in order to find suitable candidates. In e-Learning a similar situation arises. Curricula or training programmes need to describe pre- requisites that must be fulfilled before joining and the competences that will be acquired after successful completion. This paper analyses the limitations and extends existing approaches for modeling competences in order to allow (semi-)automatic competence matching.
1 Introduction
Nowadays, people mobility has increased. Learners may study abroad with the benefits of improving their language skills, receiving a better certification, or spe- cializing in a topic not available in their regions. The same applies to the labour market. People do not need to restrict themselves to their city or region while seek- ing for a job but may consider offers in other countries, too. This situation complicates the already difficult job of managers in learning organizations and Human Re- source (HR) departments to decide who may have the right qualifications to join a project or the company it- self. For learning organizations, requirements to join the programme must be taken into account. For exam- ple, an applicant needs to possess a Bachelor degree to apply for Master studies; in order to attend an expert course on a topic, a certification on a basic level may be required. Furthermore, assuming that an applicant ful- fills such requirements, exemptions could be granted for parts of the programme that are similar to earlier followed courses. Imagine a Mathematician starting a Computer Science degree. Most likely, courses like Al- gebra and Statistics could be exempted. In the case of Human Resource departments, the task is equally com- plex. HR experts need to match applicant or employee experience and knowledge with the requirements of a job offer or a project, including both mandatory re- quirements and desired ones (e.g., Business English is required and French would be a plus). Currently, all these competence matches have to be performed man- ually, with hardly any guidelines or support. One im- portant reason for this is that there are currently no sufficiently expressive common formats for the repre- sentation of competences, which is needed for com-
∗This research was partly funded by the
EU TENCompetence project (FP6-IST-02787) (http://www.tencompetence.org) and EU PROLIX project (FP6-IST-027905) (http://www.prolixproject.org/).
plex competence profiles and requirements. Some ini- tiatives, such as the IEEE Reusable Competency Def- inition (IEEE RCD, 2005) and HR-XML (HR-XML, 2004), have done initial steps to define common models and schemas for interoperability, but their current work lacks some important information that is required for competence matching, like proficiency levels, context (cf. Section 3) or mechanisms for increasing reusabil- ity. In this paper, we enhance and extend the work de- veloped under various initiatives and introduce a model for representing competences with their relationships as well as usage profiles (such as profiles for job re- quirements description or for learner achievements de- scription). This model provides the basis for allowing advanced (semi-)automatic competence matching and gap analysis. This paper is organized as follows: section 2 clarifies the terms used throughout the paper and briefly in- troduces our requirements for modeling competences. Section 3 provides an overview of existing modeling specifications, and section 4 describes our modeling approach for competences in a more detailed manner. Section 5 introduces competence profiles (collections of competences) which represent the most visible as- pect of competence modeling in real-life applications. Section 6 gives an example on how a simple profile and related competences can be modeled. Finally, section 7 concludes this paper with a summary and an outlook on future work.
2 What is a competence?
In this work we adopt the definition of competence as “effective performance within a domain/context at different levels of proficiency”, as given in (Cheetam and Chivers, 2005). Note that there exists some confu-
sion on the term competency^2 in the literature. (IEEE RCD, 2005; IMS RDCEO, 2002) define the stricter term of competency as “any form of knowledge, skill, attitude, ability, or learning objective that can be de- scribed in a context of learning, education or training”. This definition is insufficiently expressive for compe- tence gap analysis. For example, it is not clear if “pi- loting” covers both the ability to pilot a small plane and to pilot a big passenger airplane. Or if the competency “English writing skills” represents a specific level such as intermediate, fluent, native or simply the existence of the competency. In fact, if that information becomes part of the competency definition, its reusability is dras- tically reduced (with the consequence of, e.g., hav- ing different competency definitions for each context in which a competency is applied, and for any profi- ciency level and proficiency level scale). The definition given in (HR-XML, 2004) tries to extend the previous one: “A specific, identifiable, definable, and measur- able knowledge, skill, ability and/or other deployment- related characteristic (e.g., attitude, behavior, physi- cal ability) which a human resource may possess and which is necessary for, or material to, the performance of an activity within a specific business context”. In this case, “measurable” indicates a relationship with a specific proficiency level^3 and competency now applies only to the business context. In any case, since context is implicit, the models proposed from these specifica- tions do not include context information. As stated above, current approaches to modeling competencies do not explicitely address proficiency level and context. On the contrary, we believe that com- petency, proficiency level and context are three differ- ent dimensions that should be modeled separately in order to maximize their reuse. For example, the same competencies may be used in different contexts, or the same proficiency level scales may be reused among different certifications. The same applies to contexts (or “domain models”), which in many situations al- ready exist and therefore may be reused by compe- tences. Therefore, according to what stated above, we model competence (plural:competences) as a three- dimensional variable, made up of a competency (plu- ral:competencies), a proficiency level and a context (see figure 1). For example, “Fluent Business English” would be composed of the competency “English”, the proficiency level “Fluent” and the context “Business”.
For sake of clarity, and in order to avoid confusion between the terms competence and competency, we may use competency and skill interchangeably here- after. However the reader should be aware that skill is not a synonym for competency, as it only covers part of its scope.
(^2) The reader is alerted for the distinction between the two
terms, competence and competency (^3) Although they later refer to it as “grade”, which is differ-
ent from proficiency level - see section 5.
Competence
Proficiency Level
Context
Competency
Figure 1: Competence as a combination of competency, pro- ficiency level and context
3 Related Work
There exist some standardization efforts on mod- elling competenc ies. These efforts focus on different aspects related to competency: competencies as such, competency profiles and relationships among compe- tencies. The IMS Reusable Definition of Competencies or Educational Objective (IMS RDCEO, 2002) and the later IEEE Reusable Competency Definition (IEEE RCD, 2005) (based on IMS RDCEO) focus on reusable competency definitions. The primary idea is to build central repositories which define competencies for cer- tain communities. These definitions can be referenced by external data structures, encouraging interoperabil- ity and reusability. However, IEEE RCD lacks infor- mation on context and proficiency level and does not allow relationships or recursive dependencies among competencies. HR-XML focuses on the modeling of a wide range of information related to human resource tasks (like contact data or aspects of the curriculum vitae). The work performed in HR-XML Measurable Competen- cies (HR-XML, 2004) tries to define profiles in order to use such competency definitions. It specifies data sets like job requirement profiles (which describe the com- petencies that a person is required to have) or personal competency profiles (which describe the competencies a person has). Such profiles are composed of evidences (either required or acquired) referring to competency definitions (e.g., IEEE RCD). Unfortunately, the pro- posed model does not clearly separate required and ac- quired profiles. The consequence is that an acquired competency could have mandatory and optional ele- ments according to the model. Furthermore, it is un- clear why a competency is composed of several evi- dences: since a competency is a reusable object, evi- dences should rather represent a requirement or demon- strate the acquirance of a competency. Hence, the ev- idences should refer to or contain competency defini- tions and not vice versa. The Simple Reusable Competency Map (SRCM,
- tries to model relationships between competen- cies. A map can contain information about depen-
they will be needed for competence matching. For in- stance, a job requiring someone with intermediate En- glish skill typically has implicit the quantifier “with at least”, meaning that anyone with advanced English would also be accepted (and maybe even preferred ). In order to represent this relationships, an ordered list provides a reasonable means to represent a proficiency level scale (see figure 2). In such a list, the minimum value (subsumed by any other in the list) is given by the first element and the maximum is given by the last one. Therefore, the order in the list represents subsumption relationships, that is, the first element is subsumed by the second one which is as well subsumed by the third one and so on. In order to improve interoperability and matching among scales, an optional field is included for mapping to a universal scale (e.g., [0,1]). The reason why this mapping field is optional is that even though it would be useful to include it, in some contexts it may not be possible to find a suitable mapping or it may not even be necessary (e.g., if a scale is used only within an in- stitution and no interoperability is intended). Competence descriptions can refer to specific items of these scales in order to represent the proficiency level acquired/required. Algorithms could take rela- tionships among proficiency levels into account in or- der to find out how much training/learning is required to reach a determined employee/learner proficiency level. For example, if advanced English skills are re- quired, training an employee who already acquired in- termediate English skills will cost less time and money than training another employee who has only beginner English skills.
4.3 Context
(Webster, ) defines context as “the interrelated condi- tions in which something exists or occurs”, which in- cludes “the circumstances and conditions which sur- round it” (Wikipedia, ). Regarding to competences, context may refer to different concepts. It might be the specific occupation in which a competence is acquired (e.g., driving as an ambulance driver or a pizza deliv- ery employee), a set of topics within a domain (e.g., telecommunications or tourism, or theoretical vs. ap- plied physics) or even the personal settings related to the learner (e.g., competences are different if acquired in a group-based learning setting than individually). All these (and possibly more) are contexts which may be part of a competence. What actually makes up suffi- cient context descriptions can not be defined in general, but depends on the scope and purpose of the compe- tence descriptions to which they are attached. As with the skill definitions and proficiency levels, context def- initions may be reused. Modeling contexts may be a complex task, as it may coincide with modeling the whole domain knowledge
of an institution. Ontologies can capture such knowl- edge (Lau and Sure, 2002) and use arbitrary complex structures, from simple sets or tree structures to di- rected acyclic graphs. Up to date, our investigations of existing relationships between context elements (re- garding its use within competences) do not show the need for providing a graph representation or multiple inheritance. For this reason, we decided to first restrict the modeling of context to trees (see model 5 depicted in figure 2). This has multiple benefits:
- it reduces the computation complexity of compe- tences
- it is easier to understand by users
- it avoids the need for cycle-detection mechanisms while modeling is done
- it simplifies the algorithms for competence match- ing.
We are still investigating the advantages and drawbacks of this decision and we do not discard an extension of the model in case we find some scenarios for which such a structure would be beneficiary. Allowing for more advanced algorithms could also be a reason for choosing a more expressive context model. Further- more, the relationship among context concepts may also be used by algorithms analyzing competence gaps. For example, assume that a context models all occupa- tions of an airline company within an airport. In case it is needed to train a new pilot for passenger flights, it would be preferred to train some of the pilots of cargo planes instead of a person from the check-in counter. This information could be extracted from e.g. distances between the occupation “pilot” and the rest of occupa- tions in the tree/graph.
4.4 Competence
Competences are described as reusable domain knowl- edge. Any model representing competences describes what a competence is and how it is composed of sub- competences. These competences are general descrip- tions, independent of specific learners or job descrip- tions. For example, being a good taxi driver or an ex- pert Oracle database administrator are concepts with fixed meaning (domain knowledge), independent of which person possesses such competences. This is im- portant to be noticed, because competences are to be referenced from certifications or job descriptions, in or- der to stimulate their reuse. For instance, a company may define required, relevant or desirable competences for their business, which are included in job offers or
(^5) The set of attributes in the context structure is the mini- mum one allowing reference and reuse. This model may of course be extended with more data specific for the areas in which it is used
Proficiency Level -Universal Scale Mapping [0..1]
-Label
Aggregate Competence -Sequenced : boolean = false
Alternative Competence -minNumber-maxNumber :: Integer =Integer 1
SimpleCompetence Composite Competence
Metadata -RCD Schema Version -Additional Metadata
-RCD Schema
Proficiency Scale
Global Identifier -Catalogue -Entry
Statement
-Token
-Name -Text
-Id
Competence -Name
Definition -Model Source
RCD -Description -Title
Context -Label
Ordered list
-RCD Ref (^) -alternatives 2..*
-parts 2..*
-Context Ref
-levels 1..*
-Statements 1..*
-identifier 1
-subClassOf 0.. -Prof Level Ref
0..
0..
Figure 2: Competence Model
projects descriptions. The exact meaning of these com- petences is provided by a company-wide competence model. Using this approach, the explanation of a com- petence needs not to be explicitly included every time it is used^6. These explanations may cover a broad range of aspects, such as:
- how a competence may be achieved, for example by acquiring some sub-competences;
- to which level each competence should be acquired. As an example, scientific research in a University may require only basic knowledge of mathematics while at NASA, expert knowledge is needed;
- whether sub-competences must be all achieved or simply a subset of them. For instance, it is typi- cal in curricula that in order to get a degree, some topics are mandatory and some other are optional, from which a subset has to be chosen (e.g., pass k optional courses out of n offers);
- if the sub-competences must be acquired in a spe- cific order. Some companies may require that an applicant acquired a competence on personal task organization before becoming a good team leader. Otherwise, they may assume that the performance related to the competence of being a good team leader is reduced. In order to model all these elements we created an object model derived from the Composite design pattern (Riehle, 1997) (see figure 2). In our model, a competence can be either simple, an aggregation of children, or a selection from children alternatives. Competence models a competence, with references to a skill (RCD id), a proficiency level and a context. It can be a SimpleCompentence (an atomic description) or a CompositeCompetence. The latter can be either be an
(^6) As with the use of ontologies, whose classes can be sim-
ply referenced without the need of copying the whole ontol- ogy every time they are used
AggregateCompetence or AlternativeCompetence. An AggregateCompetence can be used to de- fine a competence which consists of several sub- competences, all of them required. The sub- competences can be either an ordered set (mean- ing that the sub-competences must have been ac- quired in such an order) or unordered (default). An AlternativeCompetence can be used to construct a set of alternative sub-competences. It is possible to specify a minimum and a maximum number of alter- natives that must be acquired (e.g., minimum k out of n ). “Exactly” k sub-competences might be specified by setting both minimum and maximum to the same num- ber. By default minimum is set to 1 so at least one subcompetence of the set is required.
Such a model allows to represent atomic compe- tences, (un)ordered aggregation (all sub-competences must be acquired), alternative composition (a subset of sub-competences must be acquired) and any combina- tion of all of them.
It is important to notice that if a competence is com- posed from several sub-competences, the proficiency level referenced in each subcompetence represents the minimum level required. For example, if it is required to have intermediate English skills in the context of sci- ence in order to be a good researcher, then anyone with advanced skills fulfills such a requirement. The sub- sumption relationship modeled within the proficiency levels is used for this purpose, and the proficiency level on the competence itself needs not to include all possi- ble subsumers.
Our model is open to the addition of new relation- ships, among them an equivalence relationship. This is especially interesting if competence repositories of two communities are joined and mappings between over- lapping competences have to be modeled.
ProfileElement
-Supporting Information
-Issuer Organisation
-Competence Ref
-Expiration Date
-Description -Date Issued
-Name -Type
AggregateProfileElement -Sequenced : boolean = false
AlternativeProfileElement -minNumber :-maxNumber : Integer =Integer 1
SimpleProfileElement CompositeProfileElement
Global Identifier -Catalogue-Entry Value -Scale Reference StringValue
-maxValue-minValue
-Value
NumericValue
-maxValue-minValue
Acquired -Value Profile
RequiredProfile Profile
Composite profile elements arefor Required allowed onlyProfiles This association may hold a tag/weight
-alternatives 2..*
-parts 2..*
-Global Identifier0.. -contains *
-grade 0..
Figure 3: Competence Profile
on economics. Those represent the content (scope) re- quired to acquire the competence, independently of the grade received by learners.
6 Example
We assume the existence of repositories with in- formation about skills, proficiency levels, context and competences as depicted in figure 4. In this work we do not deal with the problem of ontology hetero- geneity and we therefore assume that there either ex- ist appropriate standards for this information or there are available mappings between different ontologies (see e.g. (Rahm and Bernstein, 2001; de Bruijn and Polleres, 2004)). In addition, how these models are in- stantiated is also out of the scope of this paper. We assume the existence of appropriate tools to hide the model from end users (e.g., competence management profile or CV creation). Typically, a recruiter in a HR department would write a job offer 9 like
Wanted: J2EE consultant
- Completed Master’s Degree (any faculty)
- Expert Knowledge in Java J2EE, Servlets, JSP
- Very good English and/or French
Among other drawbacks, such an advertisement does not indicate what is mandatory or optional and, more importantly, it is not machine-understandable. Per- forming a manual matching (as widely performed now from the recruiters), the recruiter will have a hard time matching applications against this offer. An alternative would be to use the model proposed
(^9) Excerpt extracted from a newspaper
here, to encode the job advertisement (see figure 4). The model not only enforces a well-structured profil- ing, it also saves the information in a machine-readable and machine-understandable way. The recruiter can as well reuse information created from previous job advertisements (e.g., reuse the definition of Java Ex- pert for his company, as well as use the well-accepted definition of Master). This ’indexable’ representation also has significant advantages compared to the man- ual approach for the applicants: the applicants can now quickly seek on the advertisements, filter out advertise- ments for which their profile does not satisfy the re- quirements. In an even more advanced scenario, the profile representation can enable some ranking of the advertisements for which the applicant satisfies the re- quirements and some of the optional competences. Fi- nally, the cycle is concluded when the applications come back to the recruiter. The recruiter can use a (semi)automatic matching engine to filter the non- satisfactory applicants according to their profiles, and rate the suitable applicants. For example, an applicant profile as depicted in figure 4 would be a perfect match for such an offer. More complex techniques could be used for partial matches and rankings/ratings, as they have been hinted along this paper or in (Colucci et al., 2003). However, elaborating on the matching tech- niques themselves is out of the scope of this work.
7 Conclusions and Further Work
This paper addresses the problem of competence representation and exchange. Current specifications focus on the modeling of competencies (not com- petences) and they miss important information that should be included, such as proficiency level and con- text. We provide a machine-processable representation
Intermediate IT Spanish Vocabulary : SimpleCompetence Context Ref =Prof Level Ref = IT Intermediate RCD Ref = Spanish Vocabulary Intermediate IT English Vocabulary : SimpleCompetence Context Ref =Prof Level Ref = IT Intermediate RCD Ref = English Vocabulary
Intermediate IT Spanish : Composite Competence
: SimpleProfileElement Type = Master’s DegreeCompetence Ref = Expert Computer Scientist Issuer Organisation = Hannover UniversityDate Issued = 08/04/
Advanced Spanish : SimpleCompetence ContextProf Level Ref = Ref = Root Advanced RCD Ref = Spanish
: SimpleProfileElement CompetenceIssuer Organisation Ref = Intermediate IT Spanish = Cervantes
Intermediate IT : Composite Competence English
: SimpleProfileElement Competence Ref = Intermediate IT EnglishIssuer Organisation = TOEFL
: SimpleProfileElement Type = CertificationCompetence Ref = Java Expert Issuer Organisation = Sun MicrosystemsDate Issued = 02/03/
Advanced English : SimpleCompetence ContextProf Level Ref = Ref = Root Advanced RCD Ref = English
Java Expert : Composite Competence Context RefProf Level Ref = Expert = Prog. Languages
Expert Servlet : SimpleCompetence Context RefProf Level Ref= Prog. = Expert Languages RCD Ref = Servlets
Expert J2EE : SimpleCompetence ContextProf Level Ref Ref= Prog. = Expert Languages RCD Ref = J2EE
Expert JSP : SimpleCompetence Context RefProf Level Ref = Expert = Prog. Languages RCD Ref = JSP
Intermediate : Proficiency Level : SimpleProfileElement Competence Ref = Java Expert
Beginner : Proficiency Level
Advanced : Proficiency Level
: SimpleProfileElement Type = CertificationCompetence Ref = English Issuer Organisation = TOEFLDate Issued = 10/10/
: AlternativeProfileElement
English Vocabulary : RCD
Spanish Vocabulary : RCD : AggregateProfileElement
Expert : Proficiency Level
Prog. Languages : Context Label = Prog. Languages
: Alternative Competence
Fluent : Proficiency Level
: Aggregate Competence
: Aggregate Competence (^) : Aggregate Competence
: SimpleProfileElement ... : Context^ Type =^ Master’s Degree Label = ...
IT-... : Context Label = ...
Servlets : RCD
Root : Context Label = Root
Spanish : RCD
English : RCD IT : Context Label = IT J2EE : RCD
: Required Profile
: Acquired Profile
JSP : RCD
PROFICIENCY LEVELS
CONTEXT
RCDs
Figure 4: Competence Profile and Personal Profile Example
of competences, relationships among them and com- petence profiles. Such a model has been specially de- signed for reusability and allows advanced algorithms for competence and profile matching. We are currently working on the development of ap- plications in order to help end users to provide such competences and profiles. We will use our model within two different areas. On the one hand, we plan to develop advanced algorithms for competence match- ing and gap analysis in the business context as part of the EU PROLIX project. On the other hand, we plan to apply to the creation of competence development programmes and advanced assessment and position- ing services within the EU TENCompetence project. Furthermore, we are in contact with representatives of the IEEE LTSC WG20 on Competency Definitions and the HR-XML Consortium in order to contribute to the improvements of their specifications according to the ideas presented in this paper.
REFERENCES
Cheetam, G. and Chivers, G. (2005). Professions, Compe- tence and Informal Learning. Edgard Elgar Publishing Limited. Colucci, S., Noia, T. D., Sciascio, E. D., Donini, F. M., Mongiello, M., and Mottola, M. (2003). A formal ap- proach to ontology-based semantic match of skills de- scriptions. J. UCS , 9(12):1437–1454. de Bruijn, J. and Polleres, A. (2004). Towards an ontology mapping specification language for the semantic web. DERI Technical Report.
HR-XML (2004). HR-XML Measurable Competencies. http://www.hr-xml.org. IEEE RCD (2005). IEEE 1484.20.1/draft - draft stan- dard for Reusable Competency Definitions (RCD). http://ieeeltsc.org/wg20Comp/Public/IEEE_ 1484.20.1.D3.pdf. IMS RDCEO (2002). IMS Reusable Definition of Com- petency or Educational Objective (RDCEO). http: //www.imsglobal.org/competencies/. Lau, T. and Sure, Y. (2002). Introducing ontology-based skills management at a large insurance company. In Modellierung 2002, Modellierung in der Praxis - Mod- ellierung f¨ur die Praxis , pages 123–134, Tutzing, Deutschland. Lefebvre, B., Gauthier, G., Tadi´e, S., Duc, T. H., and Achaba, H. (2005). Competence ontology for domain knowl- edge dissemination and retrieval. Applied Artificial In- telligence , 19(9-10):845–859. Rahm, E. and Bernstein, P. A. (2001). A survey of approaches to automatic schema matching. VLDB J. , 10(4):334–
Riehle, D. (1997). Composite design patterns. In ACM SIG- PLAN Conference on Object-Oriented Programming Systems, Languages & Applications , pages 218–228. SRCM (2006). Simple Reusable Competency Map proposal (SRCM). http://www.ostyn.com/resources.htm. Sure, Y., Maedche, A., and Staab, S. (2000). Leveraging cor- porate skill knowledge - from proper to ontoproper. In Mahling, D. and Reimer, ., editors, 3rd International Conference on Practical Aspects of Knowledge Man- agement , Basel, Switzerland. Webster. Webster online dictionary. http://www.webster. com/. Wikipedia. Wikipedia encyclopedia. http://en. wikipedia.org/.
