Providing Personalised Recommendations of Critical Incident Narratives

in a Cross-platform Mobile Application

Tim Wenzel

, Doris Fetscher

, Wolfgang Golubski

, Susanne Klein

and Rainer Wasinger

1 a

Faculty of Physical Engineering/Computer Sciences, University of Applied Sciences, Zwickau, Germany

Faculty of Applied Languages and Intercultural Communication, University of Applied Sciences, Zwickau, Germany

Keywords:

Personalisation, User Modelling, Mobile Applications, HCI, Prototyping, Critical Incident Narratives.

Abstract:

This work describes the design and implementation of a cross-platform mobile application that has been cre-

ated to provide users with personalised recommendations of Critical Incident (CI) narratives. CIs provide

brief descriptions of situations in which misunderstandings arise as a result of the cultural differences of the

interacting parties. They are useful for increasing intercultural awareness. This paper describes the design and

implementation of the mobile application called ‘Nils2Go’. The main focus is the personalisation strategy that

is employed in the recommendation of CIs, and the identiﬁcation and use of user demographic characteristics

that can be utilised to further personalise a retrieved list of CI narratives.

1 INTRODUCTION

Various studies show that the use of mobile applica-

tions often drops sharply within the ﬁrst few weeks

and months after installation. For example, (Yan and

Chen, 2011) show how the average usage drops by

about 50% over a period of three months. Although

also dependant on the type of application (e.g. com-

munication applications are usually used for signiﬁ-

cantly longer than games that are used exclusively for

entertainment (Li et al., 2020)), ‘user experience’ has

been shown to inﬂuence long-term application usage

(Kujala et al., 2011).

This paper focuses on the question “How can CI

recommendations in a mobile application be person-

alised?”. This is important because personalisation

has been shown to lead to an improvement in user ex-

perience in the past. To achieve this in the context of

a mobile application, a user model is created, and per-

sonalisation in the form of a recommender system is

carried out to present users with a list of relevant CIs.

This paper starts with a deﬁnition of Critical In-

cidents. Following this, Section 2 outlines the back-

ground concepts of personalisation, user modelling,

and recommender systems, and also provides an out-

line of the Network Intercultural Learning and Sensi-

tivity (NILS) website. In Section 3, the low-ﬁdelity

prototype and the implemented mobile application

https://orcid.org/0000-0003-1069-5814

called Nils2Go

are presented, as too the personal-

isation strategy and technical implementation of the

application. Finally, in Section 4, our conclusions and

directions for future work are provided.

1.1 Critical Incidents

Critical Incidents (CIs) are brief descriptions of situ-

ations in which a misunderstanding, problem, or con-

ﬂict arises as a result of the cultural differences of the

interacting parties (Apedaile and Schill, 2008). CIs

form an important tool for increasing our awareness

and understanding of human attitudes, expectations,

behaviours, and interactions. Some examples of CIs

are provided in Figures 1, 2B, and 3.

In addition to the narrative, different metadata

about the CI is also stored. For example, it can be rele-

vant where the information came from, who recorded

it, and who the actors are. A reﬂection (i.e. a retro-

spective view of the past situation) by the author from

today’s perspective, is also often included in a CI.

The mobile application described in Section 3 in-

corporates both the CI narrative and its associated

metadata into the presentation of CIs to the user. This

metadata is of decisive importance in the personali-

sation of the CIs that are returned to the user and is

described in more detail in Section 2.4.1.

Nils2Go App: Available at https://play.google.com and

https://apps.apple.com.

Wenzel, T., Fetscher, D., Golubski, W., Klein, S. and Wasinger, R.

Providing Personalised Recommendations of Critical Incident Narratives in a Cross-platform Mobile Application.

DOI: 10.5220/0011528400003323

In Proceedings of the 6th International Conference on Computer-Human Interaction Research and Applications (CHIRA 2022), pages 137-144

ISBN: 978-989-758-609-5; ISSN: 2184-3244

137

2 BACKGROUND

In this section, the concepts of personalisation, user

modelling, and recommender systems are outlined, as

too the NILS website. These concepts form the basis

on which this work is then built upon.

2.1 Personalisation

There are various approaches to increasing the user

experience in the area of mobile applications. One of

these approaches is the personalisation of the software

used. In this context, personalisation involves assign-

ing relevant characteristics to individual users. Based

on these characteristics, different users will have dif-

ferent experiences when using the software. The ex-

periences are adapted to the preferences and needs of

the user. The work in this paper focuses on how per-

sonalisation can be used in the presentation of Critical

Incident narratives.

A sub-area of the personalisation of software is

represented by so-called recommendation systems.

These systems suggest speciﬁc information to the user

based on matches to his or her personal interests. In

general, personalisation deals with the discrepancy

between general-purpose applications, which achieve

the highest possible beneﬁt in a broad ﬁeld with the

least possible development effort, and specialised ap-

plications, which should fulﬁl an individual-speciﬁc

beneﬁt (Kuo, 2013).

2.1.1 Advantages and Disadvantages of

Personalisation

The development of personalised applications has a

number of advantages. One clear advantage, espe-

cially for software development companies, is driven

by ﬁnance. In (Gavril and Ionescu, 2017), the authors

demonstrate that personalised applications generate

16.5% to 24% higher revenues. Another advantage

that concerns the end user is the increase in user satis-

faction. In (Liang et al., 2006), the authors show that

personalisation, especially in the form of recommen-

dation systems, has a positive effect on user satisfac-

tion. The authors in (Tong et al., 2012) also demon-

strate the positive inﬂuence that personalisation has

on user satisfaction in software systems. Another

positive aspect is that personalised applications often

stand out from non-personalised applications (Arora

et al., 2008).

In addition to the advantages mentioned, there are

also various disadvantages. The disadvantages in-

clude the higher development costs and the increased

complexity of the development of an application, but

these are often offset by higher revenues (Blech-

schmidt et al., 2005). Another possible disadvantage,

especially for the end user, is data privacy, as personal

data is often collected and used for user modelling

(Blechschmidt et al., 2005). It is however often the

case that the positive aspects outweigh the negative

ones. In the Nils2Go application described in this pa-

per, user data that is collected never leaves the device.

2.2 User Modelling

Whereas personalisation is the process of providing

individually relevant and interesting information for

individual users of an application, user modelling

refers to the construction of a User Model (UM) that

incorporates user-speciﬁc data. This model is gener-

ated and managed by means of corresponding soft-

ware components. The collected and managed data

can refer to the demographic characteristics of the

user being modelled. Furthermore, contextual data of

the user can also be relevant. The resulting model re-

ﬂects assumptions of the system regarding the mod-

elled user and can often be retrieved in the form of

a user proﬁle. Depending on the application, such a

model can also be designed for reuse in other systems

(Kay, 1998).

Examples of user model data relevant to the

Nils2Go application that is described in this work in-

clude demographic data (e.g. age, gender, countries

of interest, spoken languages, and country of resi-

dence), general pre-deﬁned topics of interest that the

user can select from (e.g. education, food, family,

holidays, and travel), and speciﬁc interests provided

in the form of user-deﬁned keywords. This data is

shown in Figure 5D.

In (Vu and Proctor, 2011; Hothi and Hall, 1998),

the following types of User Model are outlined:

• Static UMs: This is the simplest type of UM.

Once the necessary data and information has been

provided, the UM remains unchanged for entirety

of the application’s remaining use. Changes in

user interests and preferences are not registered

or updated in the model.

• Dynamic UMs: Dynamic user models can ac-

commodate for changes in a user’s interests and

preferences. In this way, there is always an up-to-

date model of the user. Furthermore, a user’s in-

teractions with the application can also ﬂow into

the model. Depending on the use-case, it is also

possible to manually adjust the model.

• Stereotype-based UMs: Stereotype-based UMs

enable a much more anonymous way of generat-

ing and maintaining user models. Here, the inter-

ests and characteristics of the individual user are

CHIRA 2022 - 6th International Conference on Computer-Human Interaction Research and Applications

138

not part of the modelling. Rather, users are clas-

siﬁed into common stereotypes. General informa-

tion about these groups is then used to generate

preferences for the user. The basis for these as-

sumptions are primarily driven by statistics that

have been compiled for the respective groups of

people.

• Highly Adaptive UMs: Highly adaptive UMs

realise the counterpart to stereotype-based mod-

elling. Here, great amounts of information about

a speciﬁc user are collected and generated, such

that almost all anonymity of the user is lost. One

advantage of this UM is that user proﬁles are ex-

tremely speciﬁc and accurate. As a result, cor-

responding applications can be particularly well

adapted to the current user and the user receives

the best possible user experience.

• Hybrid UMs: Hybrid UMs attempt to combine

the advantages of both static and dynamic UMs,

such that a balance is found between good perfor-

mance with low memory requirements and good

personalisation with detailed user models. Often,

as is the case in (Billsus and Pazzani, 2000), the

UM consists of sub-models that take both the cur-

rent point in time as well as a longer period of

time into account. This allows information about

a user to be collected in order to ﬁrst generate a

static user model, while also allowing for dynamic

changes in the interests over a longer period of

time.

As described later in Section 3.4 the mobile ap-

plication in this work incorporates a hybrid UM ap-

proach.

2.3 Recommender Systems

In (Melville and Sindhwani, 2010), a Recommender

System is deﬁned as a software system that provides

relevant and meaningful recommendations or sugges-

tions of different products or information items to a

number of users who may be interested in them. Rec-

ommender systems usually make use of either col-

laborative ﬁltering, contextual ﬁltering, or a combi-

nation of both. For the purpose of brevity, the reader

is directed to (Aggarwal, 2016) for further details on

recommender systems. The mobile application de-

scribed in this work incorporates a content-based ﬁl-

tering approach as is described in Section 3.4.

2.4 NILS: Network Intercultural

Learning and Sensitivity

Network Intercultural Learning and Sensitivity

(NILS) is a research project of the University of Ap-

plied Sciences Zwickau. The work focuses on a sci-

entiﬁcally innovative preparation of critical incident

narratives (Fetscher and Klein, 2020). In addition

to gaining new insights into intercultural communi-

cation, the aim of the project is to sensitise users by

helping them to question cultural stereotypes. It is as-

sumed that this goal can be achieved through the in-

teractive browsing of CIs (Fetscher and Klein, 2020).

The NILS website

has been primarily designed

for research, teaching, and learning purposes. It al-

lows researchers (as well as other registered users)

to enter CI narratives into the system. It also al-

lows users of the website to retrieve CIs and ﬁlter the

CIs based on keyword searches. There are currently

around 200 CIs available through the website.

Data entry and retrieval of CI narratives is accessi-

ble via the project’s website, with the work described

in this paper providing the missing link to further view

CIs in a personalised manner via a cross-platform mo-

bile application that has been developed for Android

and iOS smartphones and published on the platforms’

respective App store fronts. An example CI narrative

from the website is shown below in Figure 1, while

Figures 2B and 3 in Section 3 show further example

narratives available through the mobile application.

Figure 1: NILS website illustrating a search for relevant CIs

based on the keyword “Restaurant”.

NILS Website: https://nils.fh-zwickau.de/

Providing Personalised Recommendations of Critical Incident Narratives in a Cross-platform Mobile Application

139

2.4.1 Critical Incident Metadata

Each CI has up to 105 attributes associated with it.

These attributes can be grouped into the categories:

textStory, origin, medium, media, hotspots, contact

domains, communication domains, reﬂection, author,

actors, and location. Each of these categories has mul-

tiple attributes associated with it, e.g. a textStory in-

cludes the narrative (labelled as textStory.story in the

dataset) as well as attributes depicting whether it has

been transcribed, the language it is written in, and de-

tails about the author.

Some of the more important attributes that are col-

lected for each CI include an ID, the title and the ac-

tual story text, a timestamp of when the data was cre-

ated, the origin (i.e. whether the story is a personal

experience, an observation, retelling, or hearsay),

the language in which the content is recorded, the

kind of data (i.e. whether it is primary data or

secondary/edited data), the narrative perspective (i.e.

ﬁrst or third person), the location of the event, the ac-

tors to whom the narrative refers, and personal data

about the author. Authors also have the ability to re-

ﬂect on their experience by including a reﬂection at a

later time. Authors can also provide hotspots in the

form of free-text keywords to describe features of the

CI (e.g. humorous, historical).

In general, the CIs can be written in any language.

In addition to German, the current set of CIs con-

tain some narratives in languages such as English,

French, Spanish, and Russian. The actors in the set of

CIs come from various European and non-European

countries and belong to different age groups.

3 Nils2Go MOBILE

APPLICATION

3.1 Purpose and Functionality

The mobile application described in this paper has

two main purposes. Firstly, it should allow for the

clear presentation of CI narratives included in the

NILS dataset. Both the actual CI story as well as

relevant metadata (see Section 2.4.1) should be made

available to the user. Secondly, the mobile application

should offer the possibility to personalise all existing

CI narratives within the given framework. In this way,

relevant elements classiﬁed as interesting for the user

are selected from the total amount of data (though still

allowing the user to access other CIs via a simple tog-

gle button as shown in Figure 5A).

In contrast to the website, the mobile application

has been kept deliberately simple and lightweight. It

does not provide the ability to add new CIs, and nei-

ther does it allow users to create an account from the

mobile device (which would be required to add new

CIs). Instead, it has been designed to present CIs

quickly and efﬁciently, and in a manner that is per-

sonalised to a given user’s proﬁle.

3.2 Low-ﬁdelity Prototype

The mobile application was developed using a

User-Centred Design (UCD) approach (Norman,

2013). This process was combined with a semester-

long course at the University of Applied Sciences

Zwickau

that focused on the creation of mobile ap-

plication prototypes for the search and retrieval of

CIs. In particular, the course focused on HCI prin-

ciples in which 13 groups of students designed low-

ﬁdelity prototypes in Balsamiq

and iteratively tested

their designs, ﬁrst in a formative and later in a sum-

mative manner with 3-5 participants each time. The

evaluation technique used for this process was the

think-aloud technique (Shneiderman, 2016). These

low-ﬁdelity prototypes were used as the starting point

for the design of the current mobile application, which

then additionally incorporated the concept of person-

alisation as shown in Figure 2, with the ability to se-

lect thematic interests (A) and to browse a list of per-

sonalised CIs (B).

Figure 2: Low-ﬁdelity prototype of the mobile application

showing a user’s thematic interests (A) and a personalised

list of CIs (B).

WHZ, URL: https://www.fh-zwickau.de/

Balsamiq, URL: https://balsamiq.com/wireframes/

CHIRA 2022 - 6th International Conference on Computer-Human Interaction Research and Applications

140

3.3 High-ﬁdelity Mobile Application

The mobile application has two main functions. The

ﬁrst is to provide access to the list of CIs. These CIs

are packaged locally with the mobile application on

compilation. This means that there is no reliance on

the Internet or the NILS website to access the CIs.

The set of CIs included with the mobile application

are however only a subset of the total available to reg-

istered users of the NILS website. Updates to the in-

cluded list of CIs is accomplished via application up-

dates to the respective App store fronts. Figures 3A,

3B, and 3C show the overview list in which CIs are ar-

ranged vertically in an abbreviated form. This means

that for each CI, the title and the ﬁrst three lines of the

actual story are visible. Furthermore, the creation date

for each narrative, as well as an information-button

that shows the related metadata is provided. To see

the complete story, the user simply selects the ‘show

more’ button. Feedback is also provided (when in the

personalised mode) to indicate if the match is due to

‘Interest Theme’, ‘Keyword’, or ‘Demographic char-

acteristic’ (Figures 3A, 3B, and 3C).

In addition to the CI narrative, the user also has

access to the metadata associated with each CI (Fig-

ure 3D), which the user can swipe through via a right-

to-left swipe gesture (in this case to see details on the

CI’s Hotspots - Figure 3D top-left) and via a left-to-

right swipe gesture (to see details on the CI’s Author

- Figure 3D top-right). In this view, all of the CI’s ad-

ditional metadata is shown. This information is pre-

sented separately to the main story so that it does not

impact on the readability of the CI narratives in the

overview page.

Also visible in Figure 4 is the ability to view

favourited CIs and to sort CIs, both alphabetically,

and via the manner in which CIs were selected as be-

ing relevant, i.e. based on the ‘Field of interest’, ‘De-

mographic Features’, or ‘Keywords’.

The second function is related to the personalisa-

tion of the retrieved list of CIs. The information con-

tained in the user proﬁle is decisive for the personal-

isation of the application and contains important data

collected about the user.

Figure 5B shows a multi-choice set of topics that

is provided to the user during their ﬁrst use of the ap-

plication. This allows the application to determine the

user’s thematic interests for the purpose of personal-

isation. The user also has the ability to skip this pro-

cess and conﬁgure their proﬁle at a later time via the

application’s side-menu. Also visible in Figure 5A

is the ‘Personalised View’ toggle-switch that allows a

user to see the CIs that are most relevant to them.

The user’s proﬁle page is accessed via the side-

menu and is shown in Figure 5D. This provides a rep-

resentation of the user model, which is statically gen-

erated at the beginning and - if necessary - adapted

over the application’s useful life (see Section 2.2).

The sections on the user proﬁle page are divided into

demographic characteristics and interests. The demo-

graphic characteristics list personal information about

the user including their age, gender, countries that

they afﬁliate with, spoken languages, and their cur-

rent location of residence. The ability to enter this

data is mainly provided to the user via pre-deﬁned

drop-down lists that limit the amount of free-form text

that the application needs to interpret. The section on

interests provides predeﬁned topics that the user se-

lected on ﬁrst use of the application, as well as user-

deﬁned keywords that can be provided in free-form.

At the bottom of the screen is also a button that al-

lows the user to delete all of the information that has

been gathered on him or her. If the user selects this

button, their intention to delete their proﬁle must be

conﬁrmed and the user model is subsequently and ir-

revocably removed from the application.

3.4 Personalisation in the Application

Section 2.2 outlined static, dynamic, stereotype-

based, highly customisable, and hybrid approaches to

user modelling. For the Nils2Go mobile application

described in this paper, a hybrid UM approach was

used. The user model is generated during the ﬁrst use

of the application, similar to a static UM approach.

The user can then ‘manually’ adjust their interests and

demographic characteristics, which makes it possible

to update the UM without complex dynamic adjust-

ment functionalities. Users can voluntarily further

customise their UM with demographic data like age,

gender, and spoken languages.

Users can also decide not to provide such demo-

graphic data, but the list of returned CIs will then

be much less precise as the personalisation strategy

will not incorporate CI metadata based on the actors

and authors with similar demographic attributes to the

user. If the user does initially decline to provide de-

mographic data, they can still decide to add it later via

the application’s side-menu.

The recommender system used in this application

uses a content-based ﬁltering approach. In particu-

lar, recommendations are implemented based on rules

speciﬁed in the software. These rules use the Lev-

enshtein string distance algorithm (Navarro, 2001) to

determine similarities between the user’s proﬁle data

(Table 1, left) and the CIs (Table 1, right).

Using the data in Table 1 row 1 to illustrate, one

such rule ﬁnds similarities between a user’s thematic

Providing Personalised Recommendations of Critical Incident Narratives in a Cross-platform Mobile Application

141

Figure 3: The mobile application for Android, showing the user’s personalised list of CIs (A, B, C), and some of the CI

metadata (D).

Figure 4: Methods of sorting CIs, including both alphabetic

(A-Z) and personalised options.

topics of interest and the CI text story and reﬂection

text. Similarly, if a user provides details on the lan-

guage(s) that they speak, this will be used to help de-

termine relevant CIs based on the language spoken by

the actors, the author, and the language that the CI has

been written in (row 5 in the table).

From a technical perspective, the recommendation

of relevant CIs is based on the Levenshtein string dis-

tance algorithm. This is well suited to applications

in which the objective is to ﬁnd matches for short

strings (e.g. our pre-deﬁned themes or user-deﬁned

keywords) in longer texts (i.e. our dataset of Criti-

cal Incidents). In this mobile application, Levenshtein

distance thresholds are used to adjust the relevance of

the returned CIs, i.e. if words are less than or equal to

the Levenshtein distance, the CI is more likely to be

deemed relevant to the user. From a code perspective,

Table 1: Recommender system data. Demographic data is

marked with an asterisk (*).

User Proﬁle Critical Incidents (CI)

+ Metadata

Thematic Interests Text story, Reﬂection

Individual Keywords Title, Text story,

Reﬂection

Age* Age (Actors, Author)

Gender* Gender (Actors, Author)

Languages* Language (Actors,

Author), Language (CI)

Countries* Nationality (Actors,

Author), Country (CI)

Place / Region* Place (CI), Region (CI)

this is represented by the following variable:

const MAX_LEVENSHTEIN_DISTANCE_TO_MATCH = 3;

In addition to the Levenshtein distance threshold

(which can also be adjusted by the user as shown in

Figure 5C), CIs are also determined to be relevant

based on the user’s demographic data (when avail-

able). As outlined in (Beel et al., 2013), the use of de-

mographic data can have a signiﬁcant impact on the

success of the recommender system. In our mobile

application, it is a combination of several matching

demographic characteristics that is required for a CI

to be marked as being relevant based on the user’s de-

mographic data (Figure 5C).

3.5 Technical Implementation

In this section, we outline the framework and tech-

nologies that were used to build the cross-platform

mobile application.

CHIRA 2022 - 6th International Conference on Computer-Human Interaction Research and Applications

142

Figure 5: The mobile application for Apple iOS, showing the side-menu (A), the user’s thematic interests (B), personalisation

settings (C), and the user proﬁle (D).

For the technical implementation of the mobile

application, the open source framework Flutter

was

used in conjunction with the programming language

Dart

. Flutter is a portable UI toolkit developed

by Google for the development of natively-compiled

applications for mobile platforms (iOS, Android) as

well as desktop (MacOS, Windows, Linux) and web

(Web Apps) (Flutter, 2022). Most importantly, this

means that the framework offers the ability to develop

applications once and then make them available na-

tively on different platforms.

Cross-platform development technologies like

Flutter and Dart allow developers to reuse the same

code across multiple platforms. They also provide ac-

cess to native features of the smartphone, including

local storage, which is relevant for this mobile appli-

cation.

The Nils2Go application is based on Flutter v3,

which was released in May 2022. According to a

2022 survey by Statista of over 31,743 software devel-

opers (Vailshery, 2022), the use of the Flutter frame-

work by software developers increased to 42% be-

tween 2019 and 2021 (an increase of 12%). Flutter

is thus now already far ahead of competing frame-

works such as React Native (used by 38% of software

developers in the survey), Cordova (used by 16%),

Ionic (used by 16%) and Xamarin (used by 11%).

Flutter, URL: https://ﬂutter.dev/

Dart, URL: https://dart.dev/

In comparison to web-based cross-development plat-

forms like React Native, Flutter uses its own high-

performance rendering engine rather than web tech-

nology. The Flutter engine is written in C/C++ and

applications written in the Dart programming lan-

guage are compiled into native code Ahead-of-Time

(AoT) for both iOS and Android

. This focus on

performance means that Flutter applications can run

faster than the competing cross-platform technolo-

gies.

A comparison of Flutter applications written in

Dart compared to native Android applications written

in Kotlin and iOS applications written in Swift shows

that there is still a performance penalty when using

Flutter (Olsson, 2020), but the advantage of having

a cross-platform application outweighs the disadvan-

tages for the purposes of this work. In (Olsson, 2020),

it is further stated that from their survey of 39 people

from the IT industry, 74% of end users could not de-

tect the difference in look and feel between a Flutter

and a native mobile application, which provides even

more incentive to use Flutter/Dart.

4 CONCLUSIONS

This work described the design and implementation

of the Nils2Go cross-platform mobile application that

Flutter FAQ, URL: https://ﬂutter.dev/docs/resources/faq

Providing Personalised Recommendations of Critical Incident Narratives in a Cross-platform Mobile Application

143

has been created to provide users with personalised

recommendations of CI narratives. The focus was on

the personalisation strategy which was demonstrated

using a hybrid user model and a content-based recom-

mender system that incorporated the Levenshtein dis-

tance as well as demographic user data (when avail-

able). Future work will now be to conduct a user-

study with the mobile application to see if the em-

ployed personalisation strategy does in fact lead to an

increase in user experience for its users as has been

shown by other past research.

ACKNOWLEDGEMENTS

A special thank you goes to the PTI06970 HCI stu-

dents (Alec Sichenender, Jonas Langner, Justin Nt-

zold, Konrad Platz) and PTI06640 Mobile Applica-

tions students (Kenny Musterer, Felix Schober, Do-

minik Strer, Tommy-Lee Bannert) at the Westschsis-

che Hochschule Zwickau, who worked on earlier pro-

totypes of this application. Special thanks also go to

the NILS development team (Tobias Haubold, Wolf-

gang Grs, Peter Huster, and Dariya Kapinus) and to

our graphic designer (David Mantilla), who provided

the colour schemes, graphics, and icons.

REFERENCES

Aggarwal, C. C. (2016). Recommender Systems: The Text-

book. Springer.

Apedaile, S. and Schill, L. (2008). Critical Incidents

for Intercultural Communication. Technical report,

NorQuest College Intercultural Education Programs.

Arora, N., Dreze, X., Ghose, A., Hess, J. D., Iyengar, R.,

Jing, B., Joshi, Y., Kumar, V., Lurie, N., Neslin, S.,

et al. (2008). Putting one-to-one marketing to work:

Personalization, customization, and choice. Market-

ing Letters, 19(3):305–321.

Beel, J., Langer, S., N

urnberger, A., and Genzmehr, M.

(2013). The Impact of Demographics (Age and Gen-

der) and Other User-Characteristics on Evaluating

Recommender Systems. In International Conference

on Theory and Practice of Digital Libraries, pages

396–400. Springer.

Billsus, D. and Pazzani, M. J. (2000). User Modeling for

Adaptive News Access. User Modeling and User-

Adapted Interaction, 10(2):147–180.

Blechschmidt, T., Wieland, T., Kuhmunch, C., and

Mehrmann, L. (2005). Personalization of End User

Software on Mobile Devices. In Second IEEE Interna-

tional Workshop on Mobile Commerce and Services,

pages 130–137. IEEE.

Fetscher, D. and Klein, S. (2020). Wissensplattform fr Criti-

cal Incidents und interkulturelle Erfahrungen [Knowl-

edge Platform for Critical Incidents and Intercul-

tural Experiences]. In Enjeux et dﬁs du numrique

pour l’enseignement universitaire, pages 79–92. Peter

Lang, Berlin.

Flutter (2022). Flutter: Design beautiful apps.

Gavril, L. and Ionescu, S. (2017). Financial Advantages

of Software Personalization. Journal of Information

Systems & Operations Management, 11(1):207–217.

Hothi, J. and Hall, W. (1998). An Evaluation of Adapted

Hypermedia Techniques using Static User Modelling.

In Proceedings of the Second Workshop on Adaptive

Hypertext and Hypermedia, pages 45–50.

Kay, J. (1998). A Scrutable User Modelling Shell for User-

Adapted Interaction. PhD thesis, Basser Department

of Computer Science, Faculty of Science, University

of Sydney.

Kujala, S., Roto, V., Vnnen, K., Karapanos, E., and Sin-

nel, A. (2011). UX Curve: A method for evaluating

long-term user experience. Interacting with Comput-

ers, 23:473–483.

Kuo, T. C. (2013). Mass customization and personaliza-

tion software development: a case study eco-design

product service system. Journal of Intelligent Manu-

facturing, 24(5):1019–1031.

Li, T., Zhang, M., Cao, H., Li, Y., Tarkoma, S., and Hui, P.

(2020). ”What Apps Did You Use?”: Understanding

the Long-term Evolution of Mobile App Usage. In

Proceedings of the Web Conference 2020, pages 66–

76.

Liang, T.-P., Lai, H.-J., and Ku, Y.-C. (2006). Personalized

Content Recommendation and User Satisfaction: The-

oretical Synthesis and Empirical Findings. Journal of

Management Information Systems, 23(3):45–70.

Melville, P. and Sindhwani, V. (2010). Recommender Sys-

tems. Encyclopedia of Machine Learning, 1:829–838.

Navarro, G. (2001). A Guided Tour to Approximate String

Matching. ACM Computing Surveys, 33(1):3188.

Norman, D. (2013). The Design of Everyday Things. Basic

Books.

Olsson, M. (2020). A Comparison of Performance

and Looks between Flutter and Native Applications.

Bachelor’s thesis, Faculty of Computing, Blekinge In-

stitute of Technology, Sweden.

Shneiderman, B. (2016). Designing the User Interface:

Strategies for Effective Human-Computer Interaction.

Pearson.

Tong, C., Wong, S. K.-S., and Lui, K. P.-H. (2012). The

Inﬂuences of Service Personalization, Customer Sat-

isfaction and Switching Costs on E-Loyalty. Int. Jour-

nal of Economics and Finance, 4(3):105–114.

Vailshery, L. (2022). Cross-platform mobile frameworks

used by software developers worldwide in 2019 to

2021. Technical report, Statista.

Vu, K.-P. L. and Proctor, R. W. (2011). Handbook of Human

Factors in Web design. CRC Press.

Yan, B. and Chen, G. (2011). AppJoy: personalized mobile

application discovery. In Proceedings of the 9th Inter-

national Conference on Mobile Systems, Applications,

and Services, pages 113–126.

CHIRA 2022 - 6th International Conference on Computer-Human Interaction Research and Applications

144