Personalized Recommender System for Improving Urban Exploration
and Experience Documentation of International Students
Madjid Sadallah
a
and Marie Lefevre
b
Universite Claude Bernard Lyon 1, CNRS, INSA Lyon, LIRIS, UMR5205, 69622, Villeurbanne, France
Keywords:
Recommender Systems, International Students, Urban Exploration, Experience Sharing, Spatial Annotation.
Abstract:
International students face significant integration challenges in new urban environments. Documenting their
experiences is crucial for reflection and adaptation; however, linguistic and cultural barriers often hinder ef-
fective documentation. This study introduces a personalized recommender system designed to facilitate this
process, enhancing social engagement. The system provides targeted prompts that guide students towards
richer, more reflective annotations. Utilizing a mixed-methods approach—quantitative analysis of user inter-
actions and qualitative feedback—we evaluated its impact. Our analysis demonstrates that the recommender
system substantially enriches student documentation, fostering deeper connections with new surroundings,
enhancing textual and emotional expression, and promoting diverse and reflective perspectives. These find-
ings highlight the system’s potential to accelerate international student adaptation and offer insights for future
technologies aimed at improving their global integration and well-being.
1 INTRODUCTION
International students often encounter significant
challenges adapting to new cultural and urban envi-
ronments, which can adversely impact their academic
performance and well-being (Patel et al., 2024).
These challenges, encompassing cultural differences,
unfamiliar urban landscapes, and language barriers,
can exacerbate isolation and hinder their sense of be-
longing (Gutema et al., 2024). While social networks
and local support systems are critical for academic
success (Zhou et al., 2008), along with practical skills
such as navigating public transportation and accessing
essential services, traditional support mechanisms of-
ten fail to provide the continuous, personalized assis-
tance that international students require (Martirosyan
et al., 2019). Digital tools, conversely, offer scal-
able solutions by providing real-time, tailored support
through mobile applications and social media.
The MOBILES application (Lefevre et al., 2024)
is designed to enrich the social and spatial experiences
of international students by facilitating the documen-
tation and reflection of their urban interactions. While
the application offers a platform for recording experi-
ences, it falls short in guiding students toward mean-
a
https://orcid.org/0000-0001-9118-0235
b
https://orcid.org/0000-0002-2360-8727
ingful reflection. Without structured support, deep en-
gagement remains elusive, as effective documentation
is key to fostering thoughtful interactions, enhancing
language skills, and supporting integration. However,
linguistic barriers and unfamiliar cultural norms of-
ten hinder students from creating insightful records of
their experiences. This study is therefore designed to
answer two critical research questions: (1) How can
digital tools be optimized to assist international stu-
dents in producing insightful documentation of their
urban experiences? and (2) How do personalized rec-
ommendations affect the quality of annotations made
by students during urban exploration?
This article contributes by integrating a person-
alized recommender system into the MOBILES ap-
plication and empirically evaluating its effectiveness.
The system provides tailored prompts to encourage
reflective, detailed documentation, addressing lin-
guistic and contextual barriers and fostering deeper
connections to the local environment. Through analy-
sis of student interactions and feedback via question-
naires, this study provides valuable insights into the
role of recommender systems in improving documen-
tation practices and supporting student integration.
The remainder of this paper begins with a liter-
ature review, followed by a description of the MO-
BILES app and its recommendation model, the study
and results, and a discussion of findings.
Sadallah, M. and Lefevre, M.
Personalized Recommender System for Improving Urban Exploration and Experience Documentation of International Students.
DOI: 10.5220/0013202800003932
Paper published under CC license (CC BY-NC-ND 4.0)
In Proceedings of the 17th International Conference on Computer Supported Education (CSEDU 2025) - Volume 2, pages 923-930
ISBN: 978-989-758-746-7; ISSN: 2184-5026
Proceedings Copyright © 2025 by SCITEPRESS – Science and Technology Publications, Lda.
923

2 LITERATURE REVIEW
2.1 International Student Integration
The integration of international students into host
cities involves cultural adaptation, social integra-
tion, and navigating unfamiliar urban environments
(Gutema et al., 2024). These challenges are cen-
tral to developing intercultural competence and per-
sonal growth (Deardorff, 2006), but cultural adapta-
tion often disrupts academic performance and well-
being, with students facing isolation and frustration
due to unfamiliar norms and academic expectations
(Zhou et al., 2008). This initial disorientation compli-
cates adjustment, especially in the early stages. The
COVID-19 pandemic further exacerbated these chal-
lenges by increasing isolation through online learning
and travel restrictions (Gutema et al., 2024). While
universities have introduced support programs like
language assistance and peer mentorship, these efforts
often fail to meet the diverse needs of students (Mar-
tirosyan et al., 2019), highlighting the need for more
targeted approaches.
Digital tools present promising solutions to sup-
port international students. For instance, mobile apps
offering city guides, language translation, and cul-
tural insights help students navigate their new envi-
ronments (Kukulska-Hulme, 2020). These apps im-
prove language skills, boost confidence in daily inter-
actions, and foster social connections, reducing iso-
lation (Loewen et al., 2020; Sun, 2023). However,
many existing tools focus solely on specific tasks,
such as orientation or language learning (Huynh and
Tran, 2023). There is a clear need for culturally re-
sponsive platforms, tailored to diverse backgrounds
and learning styles, to better address integration chal-
lenges and promote academic success.
2.2 Recommender Systems in
Education
Recommender Systems (RS) suggest items or content
by analyzing user behavior, preferences, and interac-
tions. Leveraging advanced algorithms, they deliver
personalized recommendations that enhance experi-
ences across various domains, including e-commerce,
entertainment, healthcare, and education. In the edu-
cational context, RS support personalized learning by
recommending tailored resources, courses, and path-
ways, leading to improved academic performance,
motivation, and outcomes (Silva et al., 2022). Key
approaches include collaborative filtering (drawing
on preferences of similar users), content-based filter-
ing (suggesting similar items), and hybrid methods.
(a) Homepage. (b) An annotation.
Figure 1: MOBILES main interface.
Knowledge-based systems rely on domain-specific
criteria, while context-aware systems incorporate fac-
tors like time, location, and activity to provide highly
relevant recommendations.
Beyond academics, RS contribute to social inte-
gration and adaptation in new environments. Person-
alized recommendations not only support academic
pursuits but also enhance the social aspects of stu-
dent life (Urdaneta-Ponte et al., 2021). For instance,
context-aware systems can suggest local events, study
spots, and social gatherings, facilitating integration
(Sassi et al., 2017). They further enrich experi-
ence documentation by encouraging reflective entries
that foster deeper cultural engagement (Gumbheer
et al., 2022). To aid international students in adapt-
ing to new environments, we leverage context-aware
systems to enhance urban experience documentation
through personalized prompts and location-based rec-
ommendations.
3 MOBILES RECOMMENDER
SYSTEM
3.1 MOBILES Application Overview
MOBILES is an application designed to facilitate ur-
ban exploration, experience documentation, and so-
cial interaction for international students in France.
It enables users to plan tours, document activities,
and share experiences, fostering cultural engagement.
The main interface (Figure 1) provides access to key
features, including location discovery and a map vi-
sualizing user-generated content. This map highlights
geographic points of interest and student annotations,
enhancing navigation and contextualizing urban expe-
riences. Routes are presented as a sequential narrative
of urban explorations, enriching the user’s journey.
CSEDU 2025 - 17th International Conference on Computer Supported Education
924

(a) Context information. (b) Icons selection.
(c) Tags selection. (d) Emoticon selection.
Figure 2: Key elements in the annotation creation process
(the interface text is translated into English).
Annotations are multi-faceted records capturing
the breadth of student experiences. The annotation
creation process (Figure 2) enables students to cap-
ture key contextual details, including the temporal and
spatial scope of their observations (Figure 2a). This
scope can be categorized as unique, occasional, or
regular, and can range from a specific point to city-
wide. Each annotation includes a text field for narra-
tive descriptions, with the option to add photos and
select from a range of categorized icons (Figure 2b).
These icons are color-coded into thematic categories:
Activity, Environment, Sensory, Social, and Affective.
Additional expressive features include tags (Figure
2c) and emoticons (Figure 2d). Annotations can be
shared publicly, restricted to specific groups, or kept
private, and can be edited at any time.
In addition to individual annotations, MOBILES
enables the creation of routes, which record the over-
all path of urban explorations. Routes can be gener-
ated via automatic GPS tracking, manual route map-
ping, or retrospective journey sequencing of individ-
ual annotations. The application allows linking anno-
tations to specific tours and repositioning on the map,
promoting adaptability. User interaction is also facil-
itated through comment and reaction features.
The application backend integrates the Kernel for
Trace-Based Systems (kTBS) (Settouti et al., 2009), a
RESTful service managing timestamped event traces.
Each user interaction is captured as a time-stamped
trace specifying the event type and associated data.
This trace data supports detailed behavioral analysis
and informs application improvement.
Additionally, the application features modules de-
signed to enhance user experience, including a Fa-
vorite Manager for organizing content and a Notifi-
cation Center for real-time updates, ensuring contin-
uous engagement.
3.2 Recommender System
To improve international students’ use of the applica-
tion, we have developed a personalized recommender
system (Sadallah and Lefevre, 2024) that enhances
their documentation practices. The system provides
context-aware prompts and suggestions to encourage
reflective engagement with their surroundings, while
also helping students overcome linguistic and contex-
tual challenges.
The application relies on interaction logs captured
through the integrated kTBS module, which records
timestamped user actions and geospatial data. The
recommender system processes these logs to generate
personalized recommendations using three key meth-
ods: (1) Collaborative filtering: by analyzing inter-
action patterns, the system recommends content fa-
vored by similar users, fostering a sense of shared
experience; (2) Content-based filtering: this method
matches content attributes with user profiles to pro-
vide suggestions aligned with individual interests;
and (3) Geographical context: leveraging location
data, the system delivers recommendations tailored
to the user’s current environment. The combination
of these techniques ensures that recommendations are
personalized, contextually relevant, and adaptive to
the user’s needs.
3.3 Recommendation Strategies
The recommendation engine employs five research-
backed strategies to provide personalized, contextu-
ally relevant, and comprehensive guidance to students
during their experience documentation.
Stimulating Activity. Active engagement in docu-
mentation is critical for capturing a broad range of
experiences. Higher engagement yields richer con-
tent for reflection and analysis. We define the Activity
Personalized Recommender System for Improving Urban Exploration and Experience Documentation of International Students
925

Engagement (ActEng) metric as:
ActEng =
n
∑
i=1
(C
i
+ M
i
+ T
i
) (1)
where C
i
denotes the number of annotations created,
M
i
the number of annotations modified, and T
i
the
number of tours edited. When ActEng falls below a
predetermined threshold, the system prompts users to
increase documentation activity.
Promoting Textual Narrative Richness. Detailed
textual descriptions are essential for promoting
deeper reflection. Personalized prompts enhance
contribution quality and encourage critical thinking
(Mueller and Richardson, 2022). Textual Mass (TM)
quantifies textual quality:
T M =
1
n
n
∑
i=1

LV
i
+ LD
i
2

(2)
where LV
i
is the Lexical Volume (total word count in
annotation i), and LD
i
is the Lexical Diversity (ratio
of unique to total words in annotation i). The system
prompts for more detailed text when a user’s T M falls
below the average.
Encouraging Sensitive Expression. Expressing
emotions through annotations enhances the documen-
tation process, adding authenticity and offering in-
sights into students’ experiences. Highlighting these
aspects is essential for creating detailed, meaningful
annotations, given the importance of emotional com-
munication (De Stefani and De Marco, 2019). The
Sensitive Rate (SensR) metric is defined as:
SensR = w · AAR + (1 − w) ·AIR (3)
where AAR is the Affective Annotation Rate (propor-
tion of annotations with affective icons), AIR is the
Affective Icon Rate (ratio of affective to total icons),
and w is a weight parameter (default: 0.7). The sys-
tem prompts for increased emotional expression when
SensR falls below a threshold.
Increasing Graphic Usage. Incorporating visual
elements enhances annotation quality, improving clar-
ity and engagement (Guo et al., 2020). The Graphic
Expression (GrEx) metric is defined as:
GrEx =
N
graphics
N
total
(4)
where N
graphics
is the number of annotations with
graphics, and N
total
is the total number of annota-
tions. Users receive prompts to incorporate more vi-
suals when GrEx is low.
Diversifying Icon Use. Diverse icon use enriches
visual storytelling, making documentation more ac-
cessible and engaging while overcoming language
barriers (Santos, 2020). The Icon Diversity (IcD) met-
ric is:
IcD = α · ANI + β ·ANIT (5)
where ANI is the average number of icons per anno-
tation and ANIT is the average number of icon types
per annotation. The system also analyzes icon type
distribution using:
IcD
type
=
Icons
type
Total
icons
(6)
where Icons
type
is the number of icons of a specific
type, and Total
icons
is the total number of icons in
an annotation. Recommendations are generated based
on the least diverse icon types when IcD falls below a
specific threshold.
3.4 System Architecture
The MOBILES recommender system is an indepen-
dent module within the application’s architecture, de-
signed to generate and deliver personalized recom-
mendations. It automatically captures real-time user
interactions through a tracking module, which logs
the data on a kTBS server for behavioral insights. The
application database stores essential information such
as user profiles, annotations, trajectories, and prefer-
ences, crucial for personalizing recommendations.
The system uses a multi-phase data pipeline to
collect, process, and analyze user data, ensuring adap-
tive, data-driven recommendations. Data from both
the kTBS server (which logs events) and the applica-
tion database are continuously gathered. The data is
then preprocessed to remove duplicates, handle miss-
ing values, and standardize formats. Behavioral met-
rics are calculated to generate actionable insights for
personalized recommendations.
Recommendation generation involves deriving
these insights and formulating personalized sugges-
tions based on strategic metrics. Each recommen-
dation consists of a Prompt, which encourages ac-
tions based on past behavior, and a Suggestion, which
provides contextual elements, such as examples from
peers, to support these actions.
To optimize engagement, recommendations are
tailored to the user’s history, with delivery timing
adjusted for maximum effectiveness. These recom-
mendations are delivered via push notifications and
are accessible in the notification center. A feedback
loop tracks user interactions, logging responses on the
kTBS server for continuous system refinement.
This modular design ensures that the system is
scalable and maintainable, enabling seamless updates
CSEDU 2025 - 17th International Conference on Computer Supported Education
926

to individual components without disrupting service.
It also ensures continuous access to up-to-date data,
enhancing the overall user experience.
4 STUDY
4.1 Study Design and Participants
This study investigated the impact of personalized
recommendations on international students’ docu-
mentation practices within the MOBILES application
for urban exploration. Specifically, it focused on how
recommendations influenced annotation quality and
engagement, contributing to social integration in an
urban context. A mixed-methods approach was em-
ployed to assess the recommender system’s effective-
ness and understand participants’ experiences.
International students from universities in Lyon,
France, were recruited via on-campus postings and
university mailing lists. A diverse group of 31 stu-
dents participated, comprising 15 males and 16 fe-
males, with a median age of 24 (range: 19-43). The
cohort included 9 first-year and 22 mid-year students
(14 undergraduates, 8 postgraduates). Geographi-
cally, 18 participants were from Africa, 4 from the
Americas, 4 from Asia, and 2 from Europe.
The study, conducted from March 25 to May 2,
2024, comprised three phases. In the initial phase,
participants provided informed consent, were intro-
duced to the research project, and installed the ap-
plication. During the second phase, participants in-
dependently explored Lyon, documenting their expe-
riences using the app. Regular personalized recom-
mendations were delivered via in-app notifications.
The application tracked user interactions with con-
sent, providing detailed behavioral data in a real-
world context. In the final phase, participant feedback
was collected using an online questionnaire and semi-
structured focus groups.
4.2 Procedure and Instruments
This study used a mixed-methods approach to evalu-
ate the recommender system, analyzing the evolution
of recommendation metrics alongside a focus group
discussion and a structured questionnaire.
4.2.1 Impact of Recommendation Strategies
To evaluate the recommendation strategies, the evolu-
tion of key metrics was tracked over time, including
Activity Engagement (ActEng), Textual Mass (TM),
Sensitive Rate (SensR), Graphic Expression (GrEx),
and Icon Diversity (IcD). This longitudinal tracking
was used to identify improvements or regressions at-
tributable to the system’s suggestions. For example,
an increase in TM would suggest that textual prompts
were effective in encouraging richer and more diverse
contributions, and changes in SensR would reflect the
effectiveness of emotional prompts in enhancing the
emotional richness of annotations.
4.2.2 Focus Group Session
At the end of the study, a focus group session was
held to discuss participant experiences with the ap-
plication, specifically focusing on feedback regarding
the recommender system. This provided insights into
the relevance and usefulness of the personalized rec-
ommendations.
4.2.3 Structured Questionnaire
Participants completed an online questionnaire,
which included a section on the recommender system
with two parts. The first featured closed-ended ques-
tions, where participants rated their agreement with
statements on a 5-point Likert scale (1 = strongly dis-
agree, 5 = strongly agree):
1. Relevance to personal interests: The recommen-
dations were relevant to my personal interests and
preferences.
2. Variety of recommendations: The recommenda-
tions offered a diverse range of places and events
to explore.
3. Facilitation of discovery: The recommendations
facilitated my discovery of new and relevant
places or events.
4. Quality of annotations: I am satisfied with the
quality of the annotations suggested for docu-
menting my experiences.
5. Clarity of recommendation logic: I clearly under-
stand the reasoning behind the recommendations,
based on my history and preferences.
6. Encouragement to engage: The recommendations
motivated me to engage more with the application
and use it regularly.
7. Enhancement of experience: Overall, the recom-
mendation module significantly enhances my ex-
perience.
The second part included open-ended questions to
capture deeper insights, providing context for the rat-
ings and addressing areas not captured by the closed-
ended questions. These questions focused on:
Personalized Recommender System for Improving Urban Exploration and Experience Documentation of International Students
927

1. Positive Impact: Recommendations that posi-
tively influenced experiences and the reasons for
their impact.
2. Personalization: Suggestions for better tailoring
the recommendations to individual needs.
3. Overall Impact: Assessments of the recommen-
dations’ influence on app usage and exploration.
4.3 Results
The study results focus on two areas: the impact of
the recommendation strategy and participants’ direct
feedback. Combining quantitative metrics with quali-
tative insights provides a comprehensive evaluation of
the recommender system’s effectiveness in enhancing
urban documentation.
4.3.1 Impact of Recommendation Strategies
The system’s impact on user documentation was eval-
uated using metric evolution analysis and focus group
insights. Metrics for recommendation strategies were
analyzed weekly during the active usage phase. To
assess the strategies’ impact over time, each metric
was examined. Table 1 shows average evolution rates,
means, medians, and the number of recommendations
sent and used. Figure 3 illustrates the distribution of
these evolution rates by metric.
Stimulating Activity. Personalized recommenda-
tions significantly impacted documentation activity.
The Activity Engagement (ActEng) metric, tracking
annotation and tour creation/editing, showed a weekly
average increase of 10.11% (median: 7.50%) follow-
ing 24 recommendations. This indicates increased
user activity due to relevant prompts. Participants
noted that tailored recommendations served as both
reminders and encouragement, particularly during
initial adoption. One participant stated, “The notifi-
cations were not just reminders, but also encourage-
ments that kept me engaged, especially in the begin-
ning,” highlighting the role of personalization in driv-
ing early engagement.
Promoting Rich Textual Narratives. The Textual
Mass (TM) metric, measuring annotation textual rich-
ness, showed a median weekly increase of 5.86% and
an average rise of 21% after 20 personalized recom-
mendations. This growth indicates more elaborate
and diverse user entries. Participants noted that per-
sonalized prompts encouraged reflection, citing ex-
ample entries as inspiration. One participant stated,
“The annotation suggestions that accompanied the
recommendations introduced me to well-written en-
Figure 3: Impact distribution by recommendation type.
Table 1: Impact statistics by recommendation type.
Recommendation metrics Median Mean #
Activity (ActEng) 7.5% 10.1% 24
Textual Mass (TM) 5.9% 21% 20
Sensitive Express. (SensR) 5% -5.9% 21
Graphical Express. (GrEx) 55% 55% 5
Activity Icons IcD
activity
8.9% 8.9% 5
Env. Icons (IcD
env
) 5% 30% 7
Senses Icons (IcD
senses
) 16.5% 24.3% 9
Social Icons (IcD
social
) 10.8% 15.3% 6
Affective Icons (IcD
a f f ect
) 5% 19.6% 8
tries that served as inspiration, even though I don’t
feel I have the language skills to write at that level,”
highlighting the system’s role in inspiring richer an-
notations despite varying language confidence.
Encouraging Sensitive Expression. The Sensitive
Rate (SensR) metric, measuring emotional expression
through icons, tags, and emoticons, showed a median
weekly increase of 5% and a mean decrease of -5.96%
after 21 recommendations. This indicates varied re-
sponses to emotional prompts. Some participants val-
ued the emotional prompts, stating, “The icons helped
me share emotions I often find hard to put into words.”
Others found them irrelevant or prescriptive, com-
menting, “The prompts felt artificial and pushed me
to express myself in ways that don’t match how I nat-
urally describe my experiences.” This highlights the
need for more tailored emotional support.
Increasing Graphic Usage. The Graphic Expres-
sion (GrEx) metric, assessing visual inclusion in an-
notations, showed a 55% average weekly increase
after five personalized recommendations. Despite
the strong increase, only five recommendations were
sent as participants already frequently used images.
Prompts served as valuable reminders, with one par-
ticipant noting, “Including photos made my entries
come to life, and the prompts reminded me to capture
moments I would have missed.” This suggests the sys-
tem effectively encouraged mindful visual documen-
tation despite the already high baseline.
CSEDU 2025 - 17th International Conference on Computer Supported Education
928

Diversifying Icon Use. Personalized recommenda-
tions significantly diversified icon usage in annota-
tions. Activity icons increased by 8.85%, while en-
vironmental icons showed a median increase of 5%
(mean: 30%). Sensory icons exhibited a median
increase of 16.54% (mean: 24.27%), social icons
increased by 10.83% (median), and affective icons
showed a median increase of 5% (mean: 19.95%).
This underscores a positive influence on icon diversity
and annotation richness. Participants noted that these
recommendations helped them explore the variety of
icons, with one participant stating, “These recommen-
dations improved my annotations and allowed me to
clearly tell my experiences.” This demonstrates how
recommendations encouraged enriched documenta-
tion by promoting broader icon usage.
4.3.2 Participants’ Feedback
The closed-ended questionnaire, with responses from
25 participants, assessed key aspects of the recom-
mender system using a Likert scale (1 = strongly dis-
agree to 5 = strongly agree). Table 2 summarizes sat-
isfaction regarding recommendation relevance, vari-
ety, applicability, and perceived clarity.
Table 2: Summary of closed-ended questionnaire results
(from 25 participants).
Question Median, Mean(SD)
Relevance to Interests 3.5 – 3.50 (1.21)
Recommendations Variety 4 – 3.66 (0.89)
Facilitation of Discovery 4 – 3.69 (0.92)
Relevance of Suggestions 4 – 3.84 (1.04)
Clarity of Recommendation 3 – 3.38 (1.02)
Encouragement to Engage 4 – 3.61 (1.10)
Enhancement of Experience 4 – 3.73 (1.96)
Participants rated recommendation relevance as
moderate (median/mean: 3.5), with varied satisfac-
tion, reflecting mixed perceptions of the system’s
ability to enhance documentation. Satisfaction with
recommendation variety was higher (median: 4,
mean: 3.66), though some lower ratings indicated a
need for greater diversity. Discovery facilitation was
also valued (median: 4, mean: 3.69), but lower rat-
ings indicated some students may find recommenda-
tions less useful for already familiar locations. The
relevance of suggestions also received positive, al-
beit varied, ratings (median: 4, mean: 3.84). Un-
derstanding of the recommendation logic was rated
moderately (median: 3, mean: 3.38), suggesting a
need for greater transparency. Encouragement to en-
gage with the app was moderate as well (median: 3,
mean: 3.61), highlighting a need for more personal-
ized prompts. Finally, the user experience was per-
ceived as enhanced (median: 3, mean: 3.73), but
neutral responses among several students suggest the
need for further refinements.
Qualitative data from the open-ended questions
revealed that participants valued the suggested an-
notations and their personalization. However, some
noted the need for more transparent recommenda-
tion logic and improved tailoring to individual prefer-
ences. Overall, these findings indicate that while the
recommender system is effective in various aspects,
enhancing personalization and addressing user expe-
rience variability could significantly improve overall
satisfaction and engagement.
5 DISCUSSION & CONCLUSION
This study explored the impact of personalized rec-
ommendations on international students’ urban expe-
rience documentation. Our findings reveal their po-
tential to deepen annotations and increase engage-
ment, though their effectiveness varies depending on
individual users and contexts, highlighting the com-
plexity of experiential documentation.
Personalized prompts often led to more de-
tailed and thoughtful reflections, but responses var-
ied widely. This variation underscores the need for a
recommender system that can continuously adapt to
different emotional styles and offer nuanced support.
For some, these prompts sparked deeper emotional
expression, while others found them less relevant,
reinforcing the importance of personalized, context-
sensitive recommendations.
The study also confirmed the value of diverse ex-
pressive tools. Participants appreciated the variety of
icons, and visual prompts were effective in encourag-
ing the capture of key moments, emphasizing the role
of multimodal features in meaningful documentation.
While user feedback indicated general satisfac-
tion with the relevance and diversity of recommen-
dations, it also highlighted the challenge of catering
to individual preferences. For greater user trust and
engagement, more transparency in the recommenda-
tion logic is needed. Additionally, the system’s ability
to help users explore unfamiliar areas was highly ap-
preciated, reinforcing its potential to foster a sense of
belonging and support integration.
Despite these insights, the study is subject to cer-
tain limitations. The relatively small sample size and
heterogeneity of responses pertaining to emotional
expression limit the generalizability of the findings.
Furthermore, the study did not assess the long-term
impact of these recommendations, thereby identify-
ing a critical area for future research.
Future research should expand beyond annotation
Personalized Recommender System for Improving Urban Exploration and Experience Documentation of International Students
929

quality to include urban discovery, social integration,
and language acquisition. Adaptive learning algo-
rithms to further enhance personalization should also
be explored. Finally, examining long-term effects on
social integration and well-being is paramount.
This study underscores the significant role of rec-
ommender systems in helping international students
document, reflect on, and engage with their experi-
ences in host cities. By addressing their specific chal-
lenges, the system fosters self-reflection, enriches lo-
cal engagement, and supports integration through tai-
lored suggestions. These systems not only enhance
students’ ability to capture their experiences but also
contribute to their personal growth and sense of be-
longing. Ultimately, this research demonstrates the
potential of adaptive, personalized technologies to
create more inclusive and engaging urban experiences
for diverse global communities.
ACKNOWLEDGEMENTS
This work is supported by the MOBILES project
funded by the Agence Nationale de la Recherche
(ANR), France (grant ANR-20-CE38-0009).
REFERENCES
De Stefani, E. and De Marco, D. (2019). Language, gesture,
and emotional communication: An embodied view of
social interaction. Frontiers in Psychology, 10:2063.
Deardorff, D. K. (2006). Identification and assessment of
intercultural competence as a student outcome of in-
ternationalization. Journal of studies in international
education, 10(3):241–266.
Gumbheer, C. P., Khedo, K. K., and Bungaleea, A.
(2022). Personalized and adaptive context-aware
mobile learning: review, challenges and future di-
rections. Education and Information Technologies,
27(6):7491–7517.
Guo, D., McTigue, E. M., Matthews, S. D., and Zimmer,
W. (2020). The impact of visual displays on learning
across the disciplines: A systematic review. Educa-
tional Psychology Review, 32(3):627–656.
Gutema, D. M., Pant, S., and Nikou, S. (2024). Exploring
key themes and trends in international student mobil-
ity research—a systematic literature review. Journal
of Applied Research in Higher Education, 16(3):843–
861.
Huynh, T. and Tran, L. T. (2023). Digitalisation of the
teaching and learning for international students in
higher education: A systematic review. Journal
of Comparative and International Higher Education,
15(3):63–80.
Kukulska-Hulme, A. (2020). Mobile and personal learning
for newcomers to a city. Electronic Journal of Foreign
Language Teaching, 17(1):93–103.
Lefevre, M., Benedetto, R., Sadallah, M., Chevalier, D.,
Cunty, C., Grassin, J.-F., Joliveau, T., Mathian, H.,
and Scheffler, C. (2024). MOBILES: A Mobile Ap-
plication for Documenting the Urban and Social Ex-
periences of International Students. Software, Version
1.2.41.
Loewen, S., Isbell, D. R., and Sporn, Z. (2020). The effec-
tiveness of app-based language instruction for devel-
oping receptive linguistic knowledge and oral commu-
nicative ability. Foreign Language Annals, 53(2):209–
233.
Martirosyan, N. M., Bustamante, R., and Saxon, D. P.
(2019). Academic and social support services for in-
ternational students: Current practices. Journal of In-
ternational Students, 9(1):172–191.
Mueller, C. M. and Richardson, P. (2022). Talking about
oneself: Effects of personalized prompts on l2 learn-
ers’ speech. Journal of Second Language Studies,
5(2):344–363.
Patel, N., Calhoun, D. W., and Tolman, S. (2024). Under-
standing the role of cultural competence in peer men-
torship programs for international students: A student
development theory perspective. Georgia Journal of
College Student Affairs, 40(1):62–80.
Sadallah, M. and Lefevre, M. (2024). MOBILES RecSys:
A Personalized Recommendation System for Enhanc-
ing the Urban and Social Experiences of International
Students. Software, Version 1.0.
Santos, R. (2020). The visual storytelling as a way to create
knowledge and empathy between generations in aca-
demic institutions. In The Barcelona Conference on
Arts Media & Culture 2020, Official Conference Pro-
ceedings IAFOR Barcelona, pages 1–14.
Sassi, I. B., Mellouli, S., and Yahia, S. B. (2017). Context-
aware recommender systems in mobile environment:
On the road of future research. Information Systems,
72:27–61.
Settouti, L. S., Pri
´
e, Y., Cram, D., Champin, P.-A., and
Mille, A. (2009). A trace-based framework for sup-
porting digital object memories. In Workshops Pro-
ceedings of the 5th International Conference on Intel-
ligent Environments, pages 39–44. IOS Press.
Silva, F. L., Slodkowski, B. K., Ara
´
ujo da Silva, K. K.,
and Cazella, S. C. (2022). A systematic literature re-
view on educational recommender systems for teach-
ing and learning: research trends, limitations and op-
portunities. Education and Information Technologies,
28:3289–3328.
Sun, L. (2023). Social media usage and students’ so-
cial anxiety, loneliness and well-being: does digi-
tal mindfulness-based intervention effectively work?
BMC psychology, 11(1):362.
Urdaneta-Ponte, M. C., Mendez-Zorrilla, A., and
Oleagordia-Ruiz, I. (2021). Recommendation
systems for education: Systematic review. Electron-
ics, 10(14):1611.
Zhou, Y., Jindal-Snape, D., Topping, K., and Todman, J.
(2008). Theoretical models of culture shock and adap-
tation in international students in higher education.
Studies in Higher Education, 33(1):63–75.
CSEDU 2025 - 17th International Conference on Computer Supported Education
930