Professional Documents
Culture Documents
net/publication/335771582
CITATIONS READS
13 1,662
6 authors, including:
Some of the authors of this publication are also working on these related projects:
Successful transition from Secondary to Higher Education using Learning Analytic (STELA) View project
All content following this page was uploaded by Robin De Croon on 30 September 2019.
ABSTRACT 1 INTRODUCTION
The dynamics of the labor market and the tasks with which jobs Diferent technical solutions have been proposed to support better
are being composed are continuously evolving. Job mobility is not job seekers in fnding relevant jobs from an often very abundant
evident, and providing efective recommendations in this context overload of vacancies [29]. Recommendation techniques have been
has also been found to be particularly challenging. In this paper, we extensively used to flter out relevant vacancies for a job seeker [2].
present Labor Market Explorer, an interactive dashboard that en- Although the approaches have been shown to perform well based on
ables job seekers to explore the labor market in a personalized way both implicit and explicit preference indicators of job seekers [29],
based on their skills and competences. Through a user-centered the dynamics of the labor market also demand exploration support
design process involving job seekers and job mediators, we devel- to enable potential job mobility. Typical recommender systems use
oped this dashboard to enable job seekers to explore job recom- information about an individual to suggest relevant items, but often
mendations and their required competencies, as well as how these lack support for exploration and user control [15].
competencies map to their profle. Evaluation results indicate the Interactive data visualization [6, 7] is a second approach that has
dashboard empowers job seekers to explore, understand, and fnd been proposed to allow users to flter and get an overview of jobs.
relevant vacancies, mostly independent of their background and Although these interactive visualizations have been proposed, to
age. the best of our knowledge they have not been co-designed with
job seekers or job mediators, and there may be a gap between
CCS CONCEPTS actual needs of end-users and the needs that these tools address.
• Information systems → Recommender systems; Decision Moreover, user studies that assess the utility of visualizations with
support systems; Personalization; • Human-centered computing a heterogeneous user group, including users with low technical
→ Human computer interaction (HCI). skills and non-native speakers, have not yet been conducted.
In this paper, we present “Labor Market Explorer", an interactive
KEYWORDS dashboard that enables job seekers to explore a diverse set of job
recommendations. Through a user-centered design process involv-
Recommender systems, user control, personal characteristics, ex-
ing job seekers and job mediators, we developed this dashboard to
planations, actionable insights.
enable job seekers to explore current vacancies and gain actionable
ACM Reference Format: insights by engaging with both an overview visualization as well
Francisco Gutiérrez, Sven Charleer, Robin De Croon, Nyi Nyi Htun, Gerd
as diverse kinds of flters. Note that our design tries to complement
Goetschalckx, and Katrien Verbert. 2019. Explaining and Exploring Job Rec-
existing systems, such as job or career fnders, not to replace them.
ommendations: a User-driven Approach for Interacting with Knowledge-
based Job Recommender Systems. In Thirteenth ACM Conference on Recom- This raises key research questions whether enabling job seekers to
mender Systems (RecSys ’19), September 16–20, 2019, Copenhagen, Denmark. interact with visualization techniques empowers them to explore,
ACM, New York, NY, USA, 9 pages. https://doi.org/10.1145/3298689.3347001 understand, and fnd job recommendations (RQ1), and whether per-
sonal characteristics, such as age and background, impact the user
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
perception and user interaction with such an interface (RQ2). The
for proft or commercial advantage and that copies bear this notice and the full citation contributions of this paper are two-fold: frst, we present results
on the frst page. Copyrights for components of this work owned by others than ACM of an iterative design process involving both job seekers and job
must be honored. Abstracting with credit is permitted. To copy otherwise, or republish,
to post on servers or to redistribute to lists, requires prior specifc permission and/or a mediators with diverse backgrounds. The resulting dashboard en-
fee. Request permissions from permissions@acm.org. ables job seekers to interact with a rich set of job recommendations
RecSys ’19, September 16–20, 2019, Copenhagen, Denmark responding to their user profle, including competences and skills.
© 2019 Association for Computing Machinery.
ACM ISBN 978-1-4503-6243-6/19/09. . . $15.00
Design implications for job recommendation dashboards include
https://doi.org/10.1145/3298689.3347001
60
RecSys ’19, September 16–20, 2019, Copenhagen, Denmark Gutiérrez et al.
the importance of overview components that explain recommenda- complexity for job seekers. As suggested by previous research on
tions and required competencies, as well as a wide variety of flter interactive recommender systems, novel interactive visualizations
and sorting features to support user control and actionable insights. of job recommendations can be helpful for job seekers to support
Second, we present the results of an elaborate user study (n=66) better exploration, understanding and user control. In our work,
that assesses the perceived efectiveness and user experience of our we build on these visual approaches to recommender systems to
dashboard with a heterogeneous user group of job seekers. Results improve job recommendations.
indicate that the dashboard was perceived as an efective tool by dif-
ferent user groups, mostly independent of their age and background. 2.2 Tools to Support Job Recommendation
Diferent actionable insights are obtained through the use of our Novel approaches for job recommendations include improving the
dashboard, including a greater autonomy, a better understanding algorithmic side [1, 4, 14, 30] or the fltering for exploration [16].
of needed competencies and potential location-based job mobility. RésuMatcher [14] is a job search tool that uses a similarity index
The dashboard also triggers job seekers to update their user profle, based on the job seekers experience, academic, and technical quali-
which, in turn, will result in better job recommendations. fcations. CASPER [28] uses among others a collaborative fltering
technique to support fnding relevant jobs. Absolventen.at [18] uses
2 RELATED WORK hybrid user profling and makes use of both explicit and implicit
relevance feedback. Proactive [21] is a comprehensive job recom-
2.1 Interactive Recommender Systems mender system developed for diferent categories of target users:
In recent years, several interactive visualizations have been elabo- users with a broad range of preferences and interests and users
rated to support both explanations and user control over recommen- with clear career goals and narrow interests.
dations [15]. Besides, interaction with visualizations can strongly Bakri et al. [6] designed a system with interactive visualizations
infuence users’ understanding of complex data [34]. Since visual to allow users to flter and get an overview of all posted jobs that
representation of information can help reduce complex cognitive ef- meet their criteria. The system designed by Baneres and Conesa [7]
forts [20], the use of such visualizations has been argued to be very is another example related to knowledge and missing skills for a
crucial, particularly in high-risk domains such as health-care and job position. A Sankey diagram is used to visualize the connections
fnancing [17]. EpistAid [11] is a prominent example that supports between educational programs and needed skills. LinkedVis [9] is
the screening process of physicians by using an interactive visu- another interesting example in this domain and has been deployed
alization to understand the relevance feedback algorithm. Several to enable user control over job recommendations. In our work,
other visualizations focus on improving the accuracy of recom- we focus on job seekers and use a similar interactive visualization
mendations with both explanations and support for user control. approach to support job seekers in fnding jobs. Other interesting
PeerChooser [23] and SmallWorlds [13] support interaction with approaches use conversational recommendation techniques [3].
collaborative fltering recommender engines. As in the work of Bakri et al. [6] and Baneres and Conesa [7],
Many visualizations have also been developed to interact with hy- we focus on exploration of job information, knowledge and skills,
brid recommender systems. TasteWeights [8] is a system that allows as well as potential competence gaps. In contrast to these earlier
users to control the impact of diferent algorithms as well as vari- approaches, we adopt a user-centered design process, involving
ous input data sources on the recommendation results. SetFusion both job seekers and job mediators, to elaborate a design that meets
[24] is an interface that relationships between recommendations the needs of a heterogeneous user group, ranging from non-native
and allows users to fne-tune weights of a hybrid recommender speakers to highly technically skilled users, in all age ranges.
system. MoodPlay [5] is a hybrid music recommender system that
integrates diferent techniques to support explanation and control 3 TECHNICAL ARCHITECTURE
of afective data. The system allows the exploration of a music
We designed the application following a client-server architecture,
collection through latent emotional dimensions, thereby improving
see Figure 1. The server includes the following components that
acceptance and understanding of recommendations. Yucheng et
provide diverse services to the application: the MyCareer API con-
al. [19] also explored user control in recommendations, creating a
tains information about the profle of the job seeker in XML format.
visualization on top of the Spotify recommender system.
This profle consists of information regarding their desired jobs,
Some tools focus on exploration of recommendations using a
education, and detailed competencies and skill related information
multi-perspective approach [10, 33], where diferent recommender
that is entered by job seekers in the MyCareer system of the Public
engines that provide diverse machine-produced recommendations
Employment Service VDAB. The ELISE1 component provides a ser-
and data sources are used to increase the potential of fnding rele-
vice to match the profle of the user with open vacancies registered
vant items. It has also been shown that interactive recommender
in the platform that was developed by VDAB, using a knowledge-
systems have the potential to increase the diversity of content
based recommendation technique to suggest relevant vacancies to
[31, 32]. In this sense, diversity-enhanced visual interfaces might
job seekers [29].
signifcantly reduce the exploration eforts, serving as explanatory
To overcome the typical flter bubble problem of such a recom-
mechanisms for refning the user profle and assisting users to
mender system, we retrieve a much larger set of vacancies from
perceive the diversity of the recommended items [9, 32].
ELISE than the default top-N approach based on competences. This
Increasing resources of digitized labor market information and
set is then used to enable job seekers to narrow down relevant
job postings enable better job recommendations for job seekers [29].
However, such increasing volumes of data induce also additional 1 https://www.wcc-group.com/software/elise-software-platform
61
Explaining and Exploring Job Recommendations RecSys ’19, September 16–20, 2019, Copenhagen, Denmark
62
RecSys ’19, September 16–20, 2019, Copenhagen, Denmark Gutiérrez et al.
63
Explaining and Exploring Job Recommendations RecSys ’19, September 16–20, 2019, Copenhagen, Denmark
sessions, the central focus of the map was also retained. Additional profle of the job seeker, see Figure 5c. The map represents available
features requested were added, including adding the name of the jobs with a blue dot. A grey dot is used to show unavailable jobs
city next to the distance, functionality to save searches, and various (fltered out). Moreover, on the top of the map, the type of contract
other flter and sorting options. and diploma tab components contain check-boxes to flter out and
A frst formative user evaluation was conducted with a group of control the information fow of recommendations. Job seekers can
seven job mediator (7F) and six job seekers (3F). The think aloud also save the state of the application, including the confguration
protocol was used to capture feedback from participants. The screen state of the diferent flters and favorites, see Figure 5b, for later
was recorded as well as their voice when interacting with the in- consultation.
terface. Afterwards, general feedback was discussed. Overall, the
interface was well received: participants interacted frequently with 4.4.2 Vacancies table component. The table shows the list of all
the location components and the table with competencies and lan- the job recommendations, together with visualization components
guages. They indicated to like the map component and considered that explain (DG2) the recommended jobs, see Figure 5e. Instead
distance as an important steering wheel. One of the key challenges of showing a typical matching score, the table explains this score
that participants identifed was the comparison of jobs and their with diferent dots: blue dots indicate that the job seeker masters
competencies: the list with competencies was deemed too long. The a required competence, an empty blue dot represents a required
suggestion was made to group competencies in clusters. In addition, competence that is not mastered, see Figure 5f. In addition, the
several participants struggled with the overview of competencies: table shows the distance, location, and name of the vacancy. To-
although the visual dots enabled a visual comparison, a tooltip was gether, this information explains why the job seeker receives the
used in this prototype to show which competence a specifc dot recommended items and how well it matches with his/her profle.
represents. The approach turned out to be difcult to use: several Moreover, the competencies overview can be expanded by clicking
participants indicated that the text of the diferent competencies on the “view details" button, further explaining in detail the re-
needs to be visible on the screen to enable meaningful comparison. quired competencies in column headers (Figure 5, right). Inspired
The second formative user evaluation was conducted with 11 job by the UpSet visualization technique [22], the histogram (Figure 5h)
seekers and fve job mediators and used a new version of the proto- shows a blue bar at the top of each column indicating the number of
type that addressed several issues of the frst prototype. Column times that the competence is required, providing actionable insights
headers were added to the table overview with the names of the (DG3) into the demand of particular competences.
competencies and several other features such as flters to narrow Moreover, each of the competencies columns also includes a Fa-
down the scope of vacancies based on distance were added. Overall, vorites star to control (DG1) the recommendation output by giving
the interface was again well received. The overview was considered priority to particular competences, see Figure 5i. The favorites star
to be a great improvement compared to the earlier version. Job button moves the competence to the beginning of the table as a
seekers also indicated that some of their competencies were miss- means to help job-seekers gain insight into job recommendations
ing and were “triggered" to update their profle. Feature requests that require a particular set of competences of interest.
included a “favorites" option that would move competencies to the
front of the table to enable easier comparison of vacancies with 5 FINAL EVALUATION
specifc competencies of interest. In addition, the concept of “hot
items" was requested (stars in previous designs) that shows insight 5.1 Participants
into the current demand of specifc competencies. The fnal evaluation was conducted with 66 job seekers (age 33.9
± 9.5, 18F), recruited from eight diferent training programs. Ta-
4.4 Design goals and fnal prototype ble 2 shows an overview of the diferent groups and their training
Based on the feedback gathered from the previous iterations, we programs. A heterogeneous composition was an important objec-
defned the following design goals for our fnal prototype: tive and a purposeful sampling strategy [25] was used to create a
(DG1) Exploration/control: job seekers should be able to control job heterogeneous group of participants. The aim is to create a group
recommendations and flter out the information fow coming of participants with a wide diversity. Factors that are taken into
from the recommender engine. account were age, gender, and background. Participation was vol-
(DG2) Explanations: recommendations and matching scores should untary and not compensated.
be explained, and details should be provided on-demand. To arrange the data analysis, we organized the participants ac-
(DG3) Actionable insights: the interface should provide actionable cording to their background and age. Four groups were created that
insights to help job-seekers fnd new or more job recommen- cluster participants according to their background: the technical
dations from diferent perspectives. group consists of job seekers participating in training programs for
learning the programming languages Java and C#. The construction
Following our design goals, we have defned a set of user inter-
group consists of participants in training programs for welding,
face components. We describe the components according to our
residential installation, and maintenance electricians. The sales
goals in the following paragraphs.
group consists of participants in training for sales and commercial
4.4.1 Map and filtering components. We designed a map and vari- assistance. The fourth group non-native speakers are participants
ous fltering components to explore and control (DG1) a broad set of of a vocational training for non-native speakers. We also grouped
job recommendations. The map shows the location of the vacancies participants by age. The groups were the following: 15-24, 25-34,
recommended to the job seeker that respond to the competence 35-44, 45+ years old and participants that did not say.
64
RecSys ’19, September 16–20, 2019, Copenhagen, Denmark Gutiérrez et al.
Figure 5: Two screen shots of the Labor Market Explorer. Left: Overview of competencies. a) Navigation tabs, b) Save button,
c) Map of Vacancies, d) Distance-to-job slider, e) Vacancies Table (collapsed), f) competencies Overview. Right: Detail of com-
petencies. g) Vacancies Table (expanded), h) Competencies Histogram, i) Favorite Competence.
Table 2: Distribution of the participants over the eight train- job recommendations. As language is not trivial for this group, also
ing programs. working with an interface in a diferent language as well as fnding
suitable jobs is evidently much more difcult. Perceived usefulness
Training program # Px Age Group (Q2) was a bit lower for the construction group. The wide range of
Commercial assistants 9 (8f) P1-P9 35.7 ± 11.4 Sales very specifc competencies in their domain was one of the reasons
Sales 2 (1f) P10-P11 33.5 ± 12.5 Sales they highlighted for this lower score.
Employee training non- 10 (7f) P12-P21 38.4 ± 7.9 Non-native The construction and sales groups indicated more positively to
native speakers fnd more jobs than expected (Q7). The median score dropped to
ICT, C# training 12 P22-P33 30.9 ± 6.9 Technical “neutral" for both the technical and non-native speaker categories.
Maintenance electrician 4 P34-P37 28 ± 0 Construction In general, we observed in these two user groups that the number of
Residential installer 11 P38-P48 30.5 ± 9.6 Construction vacancies that was provided was smaller than in the other groups.
ICT, Java training 9 P49-P57 36.8 ± 11.7 Technical In general, we do not see many diferences across the diferent
Welding 9 P58-P66 31 ± 4.2 Construction age categories (Figure 6c). Overall, all questions were answered pos-
itively in the diferent age group. The explorer was slightly better
perceived by older participants (45+) with respect to overall satis-
5.2 Experimental setup and data collection faction (Q5) and use intention (Q4). Confdence (Q6) was slightly
A mixed methods methodology was used. Users were asked to freely higher for younger participants.
interact with the Labor Market Explorer and use the interface to Additional feedback was captured with two open questions. The
fnd relevant recommendations based on their own competence frst question inquired whether participants would recommend the
profle. Afterwards, they were asked to fll out a questionnaire explorer to friends or family, and why. Seventy-fve percent of the
that measured diferent aspects using the ResQue framework [27]. participants would recommend the explorer. Eighteen participants
Additional feedback was captured with two open questions. Logging explicitly mention the competence-based explanations (DG2) as a
was used to capture interactions with the various elements of the key enabler for fnding job recommendations. P17 indicates that “It
dashboard. A one hour slot was used for each group of participants. helps me to see which competencies I need for a certain job. It is an
easy way to fnd a job.” P48 indicates that “nowadays there are so
5.3 User feedback many requirements that people do not know if they have the needed
competencies.” P21 comments on similar needs: “we can see our own
Figure 6 presents the results of the ResQue questions. Overall, par-
competencies and see whether they are sufcient or not.” P1 mentions
ticipants perceived the Labor Market Explorer as an efective tool,
“the diverse sub competencies that are visible and can be selected.”
as all questions had a median score of 4 (“agree"). This positive trend
The diverse set of flters seems to be another important aspect
was maintained across the diferent backgrounds and age groups,
(DG1), as it was mentioned by twelve other participants, and partic-
see Figure 6b. Technical users were a bit more negative with respect
ularly the option to flter on combinations (P24, P55). P4 indicates
to overall satisfaction (Q5). They gave very detailed and concrete
that “you can fnd ftting job ofers thanks to the broad flters". Partici-
suggestions of how particular components could be further im-
pants also indicate that the overview and flter combination enables
proved, including re-sizability of the table. Perceived accuracy (Q1)
to “easily explore the job market" (P39) and enables “analysis" (P30).
was a bit lower for non-native speakers: sixty percent of non-native
Such actionable insights (DG3) were also refected on by P60: “with
speakers were neutral as to whether the explorer helps fnding good
65
Explaining and Exploring Job Recommendations RecSys ’19, September 16–20, 2019, Copenhagen, Denmark
66
RecSys ’19, September 16–20, 2019, Copenhagen, Denmark Gutiérrez et al.
6 DISCUSSION the location where jobs are ofered, as well as a table component
that provides explanations of the job recommendations, including
6.1 Answering the research questions
needed competencies of jobs that match the profle of the job seeker.
In this section, we discuss the results of the questionnaire and link The iterative design process identifed several key features for in-
them to the interaction patterns to answer our research questions. teracting with such an overview, including the option to favorite
User empowerment (RQ1). Responses to the questions indi- competencies that are of interest. In addition, a visual represen-
cate that job seekers value the use of interactive visualization tech- tation of labor market related information such as the number of
niques to fnd relevant job vacancies. A key objective of the inter- times a specifc competence is demanded was deemed interesting
face is supporting greater autonomy. The labor market is a rapidly by both job seekers and job mediators, and was conceptualized
evolving and dynamics may be difcult to understand for diferent as a miniature bar chart in the dashboard, inspired by the UpSet
end-users. In general, the approach is perceived as efective to ex- visualization technique [22].
plore job recommendations (DG1). The interaction patterns also The competence-based explanations as well as the diverse set of
indicate that participants engaged well with the interface, with a flters, together with location-based search and overview features,
mean number of clicks above 78 for all user groups. Most partici- were highlighted as important. Interaction patterns indicate that
pants indicated that they will use the explorer again. Also important all users engaged well with the interface, although some users did
is that they feel confdent using the tool. not use all features to interact with the table overview. The reasons
Also explanations (DG2) seem to contribute well to better sup- may be two-fold: for some users, the table is less crowded and still
port of user empowerment. Many of the responses hint to better manageable without rearranging the view or narrowing down the
understanding of the job recommendations by being able to see scope. For other users, and particularly the non-native speaking
which competencies are required. A diverse set of actionable in- user group, the interface may need further simplifcation. Another
sights (DG3) were also mentioned by participants. Participants option is to use the dashboard in collaboration with job mediators
indicate that the overview of competencies enables them to explore to fully exploit the diferent features.
the labor market and understand whether they have the needed
competencies. They also gain insight into regions where most jobs 6.3 Limitations
are ofered, supporting potential location-based job mobility. Other
There are some limitations to this work that need to be articulated.
actionable insights include the observation that their user profle
First, the fnal prototype used for evaluation was a fully working
is not up to date. Such insights are key as well, as they trigger job
prototype, but the link to the actual vacancies did not yet work. The
seekers to update their profle and, in turn, receive better recom-
limitation may have impacted the perceived utility of the dashboard.
mendations.
Second, the dashboard was evaluated with job seekers of diverse
Personal characteristics (RQ2). In general, we can observe
trainings at VDAB. While we reached a very diverse user group in
only few diferences with respect to the age and gender of diferent
terms of technical skills, language and age, a wider audience would
participants. The explorer was slightly better perceived by older
be useful to assess the utility of the dashboard. Third, the interface
participants (45+), who all indicated that the explorer is a good tool
was evaluated in the very specifc domain of job recommendations.
and that they would use the explorer again. We do see diferences in
To generalize the fndings, it would be interesting to replicate the
some of the interaction patterns and subjective responses in relation
design process in a diferent domain.
to the background of participants. Participants in the technical
group engaged more with all the diferent features of the dashboard,
6.4 Future work
including the options to flter and sort by competencies and to
favorite specifc competencies. Although the perceived efectiveness Future work will focus on a “simulation mode" of the dashboard,
scores in general are not very diferent, this group makes the most where users will be able to enter information from scratch, includ-
efective use of the diferent features of the interface. ing the location and activities/competencies they would like to
The interaction patterns as well as subjective data of non-native do instead of starting from the user profle. The approach will be
speakers, sales and construction groups indicate that they engage researched to further investigate job mobility scenarios. We expect
more with the map component. The sales group also interacts well that such a simulation mode will trigger not only job seekers but
with the table overview. The construction group mostly focuses on citizens in general to explore further “what-if" scenarios, such as
sorting by distance and job title. The non-native speakers engaged changing locations and adding skills that they can acquire. A follow-
also more with the map, but we see a very wide spread in the number up study will assess whether such an approach provides further
of clicks in this group. Some participants may have struggled too insights into potential job mobility, and the impact on autonomous
much with the language. The table overview was perceived as very exploration of the labor market.
useful by all user groups, but the interaction may need further
simplifcation for some users. ACKNOWLEDGMENTS
Part of this research has been funded by the European Social Fund
6.2 Design implications (ESF Vlaanderen) - under grant agreement no. 7099 (Arbeidsmark-
tverkenner). We thank all the job seekers and job mediators who
The user-centered design process involving both job seekers and job
participated in the user studies for their interesting feedback and
mediators identifed the importance of diferent parameters used
input.
by job recommendation and search services. Based on this rank-
ing, a dashboard was designed with a map component that shows
67
Explaining and Exploring Job Recommendations RecSys ’19, September 16–20, 2019, Copenhagen, Denmark
68