Towards Meaning in User Interface Prototypes
Gretchen Torres De Macedo
1,2 a
, Lucas Viana
1 b
and Bruno Gadelha
1 c
1
Institute of Computing, Federal University of Amazonas - UFAM, Manaus, Amazonas, Brazil
2
Amazonas State Prosecution Service - MPE/AM, Manaus, Amazonas, Brazil
Keywords:
User Interface Prototypes, Semantic Analysis, Heuristics.
Abstract:
User interface prototypes aid several activities during the development process lifecycle. However, there are
still many manual activities performed during this process. This research project investigates how semantic
meaning identification in UI prototypes can help carry out manually performed tasks. We started by analyzing
a set of UI prototypes obtained from prototyping activities with students. As a result, we created a database of
856 UI prototypes labeled semantically in 19 classes and two semantic levels: layout and functionalities. Each
class of UI prototypes was analyzed to identify its distinguishing characteristics, from which we obtained a
set of 19 heuristics specifications. This set of heuristics allows the development of solutions for automatic
analysis of UI prototypes, thus supporting software prototyping activities.
1 INTRODUCTION
User interface prototyping is a commonly used re-
source to represent software features through user in-
terface images during the software development pro-
cess (Bjarnason et al., 2021). Analysts can use UI
prototypes to communicate with customers, explore
solutions, and present and validate ideas (Dow et al.,
2011; Lauff et al., 2020). In projects that use the agile
development methodology, prototyping has shown to
be a critical requirements elicitation instrument that
can be associated with design thinking techniques
(Garcia et al., 2017; K
¨
apyaho and Kauppinen, 2015).
However, prototyping is a challenging task (Lee
et al., 2020). A designer must analyze the functionali-
ties to represent during software prototyping and then
choose design elements that characterize such func-
tionalities. It is necessary, at the same time, to follow
design rules and be creative and innovative to attract
and maintain user engagement (Chen et al., 2020a).
Therefore, there must be ways to assist the design of
graphical user interfaces (GUI).
Some works deal with this challenge by automat-
ically analyzing GUIs to find design problems. Ex-
amples include using computer vision techniques to
check whether a GUI implementation is according to
its design (Moran et al., 2018), or GUIComp (Lee
a
https://orcid.org/0000-0003-1428-4924
b
https://orcid.org/0000-0002-8154-0062
c
https://orcid.org/0000-0001-7007-5209
et al., 2020), a tool to assist GUI design by suggesting
similar GUIs, evaluating the arrangement of compo-
nents, and showing attention areas on the prototype.
However, GUI design checking approaches based on
images do not consider the meaning of the elements
and functionality represented in the prototype, which
limits the types of checks made during the GUI eval-
uation to the arrangement of UI components on the
screen.
Another way to aid the conceptualization of GUIs
is through GUI searching (Huang et al., 2019), where
designers can search a variety of examples of similar
user interface designs (Bernal-C
´
ardenas et al., 2019;
Chen et al., 2020b). However, GUI searching based
on images only searches for GUIs visually similar,
limiting the diversity of recovered examples.
Given the limitations of image-based UI design
assistance methods, new approaches leverage seman-
tic identification of interfaces to perform UI related
tasks. A basis work, Liu et al. (2018) use a screen-
shot dataset and screen metadata to generate seman-
tic annotations for mobile apps interfaces. Another
work uses interaction data and transition components
detection to predict the subsequent screens given a
pair of screenshots (Zecheng et al., 2020). Chen et
al. (2020a) use a combination of images and textual
information to predict screen categories and suggest
tags that describe UI designs.
Those examples show that semantic identifica-
tion enables more detailed UI verifications and could
De Macedo, G., Viana, L. and Gadelha, B.
Towards Meaning in User Interface Prototypes.
DOI: 10.5220/0011067100003179
In Proceedings of the 24th International Conference on Enterprise Information Systems (ICEIS 2022) - Volume 2, pages 235-246
ISBN: 978-989-758-569-2; ISSN: 2184-4992
Copyright
c
2022 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
235
therefore be used to provide better feedback during in-
terface design by indicating design errors. An exam-
ple is the analysis of recurrent features. Some features
are common to different applications, such as user
authentication, search, or registration forms. How-
ever, the knowledge regarding the descriptive char-
acteristics of these recurring features is tacit: prac-
titioners widely know it, but it is not formally spec-
ified. Therefore, an explicit description of each type
of functionality can help prototyping activities by pro-
viding a means for identifying the functionality repre-
sented in a prototype and indicating possible inconsis-
tencies between the prototype and the specification of
the functionality. Furthermore, identifying semantic
meaning in UI prototypes could improve the search
for alternative designs for a feature (Liu et al., 2018).
Semantic identification can also support the im-
plementation of functional prototypes. For example,
developers can use functional prototypes (Bjarnason
et al., 2021) to demonstrate functionalities more re-
alistically and evolve the prototypes to become the
first working version of the application, also known as
a Minimum Viable Product (MVP) (Lenarduzzi and
Taibi, 2016). However, this involves additional cod-
ing effort and, therefore, increased development cost.
Hence, identifying the functionality represented in a
prototype could also be used for suggesting code for
implementation, retrieved from local or remote code
repositories.
Considering these new applications of seman-
tic identification to assist user interface design, this
work seek to answer the following research ques-
tion: “How to perform semantic identification in
user interface prototypes?”. To answer this ques-
tion, we addressed the problem of semantic identi-
fication through a heuristic-based approach. In this
approach, we identified different categories of pro-
totypes and described their key characteristics using
heuristic specifications (Qui
˜
nones et al., 2018). Ini-
tially, to obtain a prototype base for analysis, we con-
ducted prototyping activities with undergraduate stu-
dents, resulting in 856 screens of 31 different projects
from varied domains. From the iterative, open cod-
ing analysis (Felderer and Travassos, 2020) of UI
prototypes, we labeled them in 19 categories, which
fit in two different levels of abstraction
1
. Then, for
each class, we wrote a descriptive heuristic specifi-
cation. The set of heuristics and the labeled proto-
types dataset can assist UI prototyping or integrate
automated UI prototype analysis solutions.
This paper is organized as follows: In Section 2,
we present works that identify semantics in UI proto-
1
Research data available at https://doi.org/10.6084/m9
.figshare.17087717
types or design images for different purposes; in Sec-
tion 3, we describe the process of developing the de-
sign heuristics; in section 4, we present the heuristic
specification of the classes identified during the anal-
ysis, with some examples; in Section 5, we present the
evaluation of our set of heuristics using Expert Judg-
ment; in Section 6, we discuss the results obtained and
compare them with related works; finally, in Section
7, we present the conclusions of this work and future
work.
2 BACKGROUND
This section presents key concepts that form the ba-
sis of our research project: User Interface Prototypes
and Heuristics. Also, we present related works that
use different approaches to address the problem of UI
analysis.
2.1 User Interface Prototypes
User Interface Prototyping is a practice used in the
early stages of software development. Through pro-
totypes, UI designers can demonstrate the software
functionality by a visual representation of the soft-
ware user interface (Deininger et al., 2017). Besides,
designers can identify design problems and explore
different software solutions for a problem.
The systematic mapping of Garcia et al. (2017)
shows that prototypes are the most used artifact dur-
ing different software development stages. They fa-
cilitate communication with customers in agile set-
tings, serving later as a reference for development and
a guide for usability tests. K
¨
apyaho and Kauppinen
(2015) also cite prototyping as a powerful artifact for
requirements elicitation in agile settings, where there
is usually little documentation since they are easier to
create and update than written documentation.
Software interface prototyping activities can be
aided by semantic identification. In this work, we ap-
proached the semantic identification problem by de-
veloping design heuristics to characterize prototype
categories.
2.2 Heuristics
Heuristics are rules derived from experience used
in the implementation of decision processes to find
solutions to specific problems (Calle-Escobar et al.,
2016). Unlike deterministic methods where all vari-
ables are involved in finding a perfect solution, heuris-
tics ignore part of the information to find a good
enough solution for solving a problem (Gigerenzer,
ICEIS 2022 - 24th International Conference on Enterprise Information Systems
236
2008). That way, heuristics are used to reach a satis-
factory solution with less computational effort (Shah
and Oppenheimer, 2008).
Several works use heuristics to support products’
design. This type of heuristics is known as Design
Heuristics. For example, Figueiredo et al. (2012)
present design heuristics to assess concern-sensitive
designs, i.e., aspects of a software product that can af-
fect the maintenance of software modules. Yilmaz et
al. (2016) present a compilation of 77 design heuris-
tics commonly used in product design.
In this work, we adopted a heuristic-based ap-
proach to develop design heuristics that can assist de-
signers during UI prototypes development. Heuris-
tics could focus on a UI category’s most relevant and
deterministic characteristics to check if it meets the
specification.
2.3 Related Works
Several approaches have been used to support UI de-
sign tasks. In the field of user interface analysis,
Moran et al. (2018) presents an approach to iden-
tify errors in mobile GUI implementations by com-
paring it to its design. Packevi
ˇ
cius et al. (2018) lever-
age application code and screenshots to implement a
source code analysis method based on sets of defects
and rules to identify text presentation and semantic
defects. Finally, GUIComp (Lee et al., 2020) is a GUI
prototyping assistance tool that performs GUI check-
ing and provides feedback by recommending similar
user interfaces, evaluating visual complexity metrics,
and showing points in the prototype that grab more
user attention.
Other works also use the UI searching and rec-
ommendation approach to facilitate the conceptual-
ization of user interfaces. Guigle (Bernal-C
´
ardenas
et al., 2019) is a text-based UI search that leverages
screen texts, UI components, app name, and colors to
retrieve similar screens from Google Play apps. Swire
(Huang et al., 2019) implements a sketch-based GUI
retrieval from a sketch database built with UI exam-
ples from Rico dataset (Deka et al., 2017). Chen et al.
(2020b) also perform GUI search and retrieval using
a wireframe-based approach.
Several recent works adopt the semantic identi-
fication approach in user interface prototypes (Liu
et al., 2018; Leiva et al., 2020; Zang et al., 2021; Chen
et al., 2020a). In the work of Liu et al. (2018) , the
authors describe a semantic annotation approach of
elements present in images of mobile application in-
terfaces, specifically textual buttons, and icons, in ad-
dition to some structural elements, such as menus and
footers. This annotation can be used to leverage se-
mantic based operations, such as GUI search or flow
search.
ActionBert (Zecheng et al., 2020) is a methodol-
ogy for identifying semantic meaning based on ap-
plications’ structural information and interactive user
actions’ tracking. From a pair of consecutive screens,
it can identify the linking component and predict
the subsequent screen. The Enrico dataset (Leiva
et al., 2020) is a labeled database of 1460 application
screenshots, taken from the Rico (Deka et al., 2017)
dataset, sorted into 20 topics. The authors also present
several possible applications of Enrico to help design-
ers’ work during user interface design. Chen et al.
(2020a) use semantic identification to discover miss-
ing tags used to describe images on design sharing
sites. The presented solution combines application in-
terface design images and metadata to suggest com-
plementary tags and a visualization method to high-
light which parts of images gave rise to the tag hint.
The work of Li et al. (2021) presents a self-supervised
technique that combines text information, design and
layout patterns, application metadata, and interaction
data to generate embeddings that can be used in var-
ious applications, from assistive technologies to de-
sign support of UI.
Unlike these related works, we use a method
based on the thorough analysis of different types of
prototypes to extract their main features. The result-
ing heuristics describe the different types of screens
through their layout or functional characteristics. Our
approach differs from others by going beyond the
recognition of prototypes’ categories, once the check-
list specified in each heuristic allows us to verify
whether the design meets the requirements described
in the heuristic specification.
3 METHOD
In this work, we present a set of design heuristics to
identify semantics in user interface prototypes. To de-
velop the design heuristics, we first identified cate-
gories of prototypes from a base of student-developed
UI prototypes and related work. Then, we identi-
fied the defining characteristics of each prototype cat-
egory, which make up the heuristics specification.
To achieve this set of prototype categories and
their predominant characteristics, we adapted the
methodology proposed by Quinones et al. (2018) to
the context of design heuristics specification. This
methodology is composed of 8 steps, which can be
performed sequentially or iteratively, in cycles involv-
ing one or more steps. Figure 1 shows how we per-
formed the methodology steps. In the Exploratory
Towards Meaning in User Interface Prototypes
237
and Experimental stages, we gathered elements for
the Descriptive stage. Also, in the Correlational stage,
we changed the prototypes categories obtained during
the Experimental stage. In this section, we describe
each of them and how we adapted it to the context of
this work.
Step 1, the Exploratory stage, initiates by defin-
ing a specific application domain to develop usabil-
ity and UX heuristics. In our case, we aim to assist
UI design activities, so we developed characterization
heuristics for UI prototypes from different domains
and platforms. Therefore, our research domain con-
sists of user interface prototype analysis in the con-
text of user interface design. In this stage, we also
must collect three types of data: information about the
domain, usability/ UX attributes under evaluation by
the set of new heuristics, and existing heuristics, prin-
ciples, guidelines, and patterns. First, as presented in
Section 2, we performed a literature review, where we
found works related to GUI design assistance, seman-
tic identification in UI prototypes, and GUI search. In
those works, the UI design and prototypes classifica-
tion is approached at different semantic levels: at the
component level, by identification of UI components
and interactive elements like buttons and icons (Liu
et al., 2018); at the structural level, where the layout
of its components characterizes the semantic of UI de-
signs (Leiva et al., 2020); and at the functional level,
describing more complex functionalities (Chen et al.,
2020a).
In our context, we collected the following types of
information from related works:
Components identification heuristics, text buttons
concepts, and icons classes from Liu et al. (2018);
UI designs topics from Leiva et al. (2020);
UI semantic categories (platform, color, app func-
tionality, screen functionality, screen layout) from
Chen et al. (2020a);
User interface design patterns from UI Patterns
website (Toxboe, 2022).
In Step 2, the Experimental stage, we performed
an experiment to create a user interface prototypes
dataset. We needed this dataset to analyze and ob-
tain a categorization of UI prototypes. Thus, we per-
formed prototyping activities with undergraduate stu-
dents. These activities resulted in 31 projects and
856 UI prototypes for mobile, web, and desktop plat-
forms.
Then, we analyzed the prototypes to obtain dif-
ferent categories of prototypes. Before starting the
prototypes’ analysis, one of the researchers removed
duplicated prototypes and descriptive UI elements.
Then, this researcher started labeling UI prototypes
using an open coding approach, where the labels of UI
prototypes were created and modified iteratively dur-
ing the analysis. The researcher replaced some pro-
totypes labels with more representative ones in this
process. Next, the researcher reviewed the labeling
performed in the first step, unifying distinct labels as-
signed to prototypes with the same semantic meaning
and splitting labels for prototypes with different se-
mantic meanings. This researcher also grouped very
similar labels with few elements. After the review
stage, another experienced researcher reviewed the
categorization. The disagreement points were dis-
cussed until they reached a consensus.
The labeling process resulted in 23 prototypes cat-
egories, where 47 UI prototypes received more than
one label. Table 1 presents the prototype categories
obtained during the Labeling process. It is worth
noticing that only the quantities of single-label pro-
totypes are listed.
In step 3, the Descriptive stage, we must select
and prioritize the topics collected in Steps 1 and 2.
Therefore, we analyzed and selected heuristics and
UI categories from related works. First, from Liu et
al. (2018), we selected all concept buttons and syn-
onyms definitions, as we will use them in our heuris-
tics definition. However, their heuristics were not se-
lected because they detect UI components and icons,
and our work goal is to identify semantics for the en-
tire UI prototype. Next, from the UI Patterns website
(Toxboe, 2022), Leiva et al. (2020), and Chen et al.
(2020a), we selected only the categories that matched
the ones we found during the prototypes labeling pro-
cess. From Chen et al. (2020a) we also selected the
app categories to classify our prototype projects. We
also adopted their screen functionality and layout UI
categories to group our heuristics. Table 2 presents
the information selected from each related work.
In step 4, the Correlational stage, the researchers
must match features and existing heuristics collected
in previous steps. We have not selected the pre-
existing heuristics in step 3, so we mapped the pro-
totype categories resulting from the labeling process
with the categorization lists from related works. The
Table 3 presents this mapping. Also, we grouped our
categories into two groups: Layout and Functionali-
ties. The Layout categories refer to generic software
features, such as registration forms and lists of ob-
jects. The Functionalities categories are related to
more complex features, such as authentication forms
and chat. The Layout categories identification starts
with the HEU1 prefix and the Functionalities cate-
gories, with the HEU2 prefix.
In Step 5, the Selection stage, we should deter-
mine which heuristics to keep, adapt, create and elim-
ICEIS 2022 - 24th International Conference on Enterprise Information Systems
238
Figure 1: Methodology to develop usability and user experience heuristics (Source: Qui
˜
nones et al. (2018)).
Table 1: Prototypes categogies obtained during the Experimental stage.
# Label Quantity Description
01 Account Registration 57 Creating a user account to access an application features
02 Authentication 40 Credential information form for accessing application resources
03 Chat 24 Functionality for exchanging instant messages between users of the
application
04 Checkout 17 Information on payment, shipping, and billing options in e-commerce
applications
05 Content Filter 7 A set of fields that filter the displayed information
06 Date Picker 8 Query based on dates presented as a calendar
07 File Upload 6 A file upload form
08 Grid 20 A set of instances of an entity from the application domain displayed
in a grid format
09 Home 11 Application home page presenting the most relevant information in a
wide variety of formats.
10 List 123 Set of instances from a domain entity
11 Master-Detail 28 Displaying attributes of an instance of a domain entity followed by a
list of instances of another entity related to the first
12 Menu 82 Application feature options
13 Modal 21 A pop-up window asking for a user action.
14 Notification 48 System status information
15 Product Page 20 Detailed information on a product
16 Registration 80 Form for registering an instance of an entity in the application domain
17 Search 50 Search engine
18 Shopping Cart 12 List of products or services and their quantities, to be purchased in an
e-commerce application
19 Terms of Use 10 Application terms of use
20 Update 34 A form that presents the attribute values of an entity instance so that
they can be modified
21 User profile 32 User account information
22 Videoconference 20 Videoconference functionality
23 View 59 Displays attribute values of an object from an entity in the application
domain
Table 2: Information selected from different sources during step 3, Descriptive stage.
Source Collected information Selected information
Liu et al. (2018) 28 UX concept buttons and synonyms (no, lo-
gin, back, go, ok, all, next, add, create, more,
retry, skip, set, delete, Facebook, view, book,
continue, list, save, search, finish, agree, show,
share, buy, update, edit); Heuristics for: ex-
tracting text buttons, extracting icons, code-
based heuristics to identify components
All concept buttons and synonyms were se-
lected. None of the heuristics were selected.
Leiva et al. (2020) Enrico dataset topics (bare, dialer, camera,
chat, editor, form, gallery, list, login, maps,
media player, menu, modal, news, other, pro-
file, search, settings, terms, tutorial
Chat, form, list, login, menu, modal, profile,
search, terms.
Chen et al.
(2020a)
Semantic UI categories (platform, color, app
functionality, screen functionality, screen lay-
out)
All app categories; screen functionality cate-
gories: login, signup, checkout, search, profile;
layout categories: form, list, grid.
UI Patterns web-
site (Toxboe,
2022)
User interface design patterns Modal, notifications, table filter, product page,
shopping cart, chat, account registration.
Towards Meaning in User Interface Prototypes
239
Table 3: Match between categories from the labeling process and related works.
ID Category Enrico (Leiva
et al., 2020)
Chen et al.
(2020a)
UI Design Patterns
(Toxboe, 2022)
Layout categories
HEU1-01 Content Filter Table Filter
HEU1-02 File Upload
HEU1-03 Grid Grid
HEU1-04 List List List
HEU1-05 Master-Detail
HEU1-06 Menu Menu
HEU1-07 Registration Form Form
HEU1-08 Update Form Form
HEU1-09 View
Functionalities categories
HEU2-01 Account Registration Signup Account Registration
HEU2-02 Authentication Login Login
HEU2-03 Chat Chat Chat
HEU2-04 Checkout Checkout
HEU2-05 Home
HEU2-06 Product Page Profile Product Page
HEU2-07 Search Search Search
HEU2-08 Shopping Cart Shopping Cart
HEU2-09 User profile Profile Profile
HEU2-10 Videoconference
inate. We analyzed the prototype categories in this
work to identify their characteristics that represent
prototype design requirements. This analysis veri-
fied that four categories do not represent software fea-
tures: Date Picker, Terms of Use, Notification, and
Modal. Date Picker is a differentiated data entry com-
ponent for dates selection used in various situations.
Terms of Use have no specific design requirements.
Finally, the Notification and Modal categories repre-
sent situations in which texts are shown to the user
and may require user action. These categories can be
used in a wide variety of situations and also do not
represent software features. Therefore, we did not se-
lect these four categories for the specification stage.
Step 6 is the Specification stage, where the heuris-
tics are formally specified. In this work, we specified
heuristics for the 19 remaining categories from the la-
beling process. We adapted the heuristics definition
standard template proposed by Qui
˜
nones et al. (2018)
to the context of this work. As we are not dealing
with usability heuristics, we removed the fields Ap-
plication Feature, Benefits, Usability Attributes, and
UX Factors.
Also, we added the fields Synonyms and Addi-
tional Information. The field Additional Information
in the heuristics specification lists additional sets of
information a heuristic needs in its checklist items.
For example, some heuristics, such as Registration,
Update, and List, are related to the application do-
main entities. Besides, the heuristics use terminolog-
ical and visual characteristics, mapped as terms and
icons sets, to describe the prototype elements. For in-
stance, a set of terms used to label buttons on authen-
tication forms is (login, connect, sign in). We present
the heuristics specification in Section 4.
In Step 7, the Validation stage, we performed a
validation of the proposed design heuristics through
an expert judgment experiment (Qui
˜
nones et al.,
2018). We proposed an app development scenario
where the participants designed UI prototypes for the
software requirements. In this study, we assessed the
utility, ease of use, and improvement recommenda-
tions for the heuristics. Section 5 presents the experi-
ment results in detail.
In step 8, the Refinement stage, we performed re-
finements in the set of heuristics using the validation
stage results. We merged the heuristics Content Filter
and Search. We also merged the Grid and List heuris-
tics because they represent groups of objects with dif-
ferent arrangements. We also added the carousel as
a form of presenting objects in this heuristic. The ex-
pert analysis also suggested the inclusion of heuristics
to design prototypes with accessibility requirements
and evaluation forms.
4 UI PROTOTYPE DESIGN
HEURISTICS
This section presents characterization heuristics for
the categories defined during the labeling process.
Due to space limitation, we present three heuristics
from the Layout and three from the Functional cat-
egories. We chose the more representative heuristics
ICEIS 2022 - 24th International Conference on Enterprise Information Systems
240
from each category, i.e., with the higher number of in-
stances. The remaining heuristics are available in the
technical report
2
.
4.1 Layout Heuristics
Layout heuristics comprise UI prototypes whose
predominant characteristic is to present general
application features. Figure 2 shows a List prototype
example. In this section, we detailed three layout
heuristics: List, Menu, and Registration.
Figure 2: Level Layout example – List prototype.
ID: HEU1-04
Name: List
Synonyms: Vertical list, horizontal list
Definition: List of objects from a domain entity
Explanation: A list presents a set of objects from an
entity in the application domain.
Examples: A user views a movies list on a streaming
service application.
Additional information:
List of domain entities and attributes.
Checklist:
1. The prototype presents a list of objects from an
entity in the application domain.
2. Each item in the list presents values of at-
tributes from the object.
Related heuristics: Search, Videoconference, Master-
Detail, Grid
2
Technical report available at https://doi.org/10.6084/
m9.figshare.17087717
ID: HEU1-06
Name: Menu
Synonyms: Toolbar, drop-down menu, top bar,
bottom bar
Definition: Application features options.
Explanation: A menu displays the actions a user can
perform on a screen.
Examples: The user browses departments of an
online clothes store.
Additional information:
1. Application features.
2. Application features labels.
3. Application features icons.
Checklist:
1. The prototype presents a set of elements in a
side list, top bar, bottom bar, or buttons.
2. Elements labels and icons represent the
functionalities offered by the application.
Related heuristics: none
ID: HEU1-07
Name: Registration
Synonyms: Record
Definition: Form for registering an object of an entity
in the application domain.
Explanation: A registration form illustrates user
input information that is stored in the application’s
database.
Examples: A news website employee publishes
information about the municipal elections.
Additional information:
1. List of domain entities and attributes.
2. RegBtnLblSet set of labels to describe
registration buttons. Examples: create, send, save,
request inclusion, add, forward, finalize, confirm,
register, generate.
3. RegIconSet – set of icons to represent registra-
tion buttons.
Checklist:
1. The prototype presents a data entry form with
one or more fields whose values are not filled in.
2. The form fields represent attributes of an object
in the application domain.
3. The form presents a button whose label
belongs to a set of labels used to describe registration
buttons (RegBtnLblSet) or whose icon belongs to
a set of icons used to represent registration buttons
Towards Meaning in User Interface Prototypes
241
(RegIconSet).
Related heuristics: Account Registration, File Upload
4.2 Functional Heuristics
Functional heuristics comprise UI prototypes that
represent high level functionalities that are frequent
in many applications. In this section, we present
the heuristics specification for Account Registration,
Authentication and Search prototypes categories.
Figure 3 shows an Account Registration prototype
example.
ID: HEU2-01
Name: Account Registration
Synonyms: Sign up, sign-up, signup, user registration
Definition: Creates a user account to access the
application’s features.
Explanation: A user from the application creates an
account to access personalized or restricted features.
Examples: A user registers for a restaurant website to
order a pizza.
Additional information:
1. List of user entities and attributes from the
application domain.
2. UserIdLblSet a set of labels to describe user
identification fields. Examples: id, login, email,
identification, name, user, username, registration
number, phone number.
3. PwdLblSet a set of labels used to describe
password fields. Examples: password, keyword.
4. RetPwdLblSet – a set of labels used to describe
retype-password fields. Examples: confirm pass-
word, retype your password.
5. RegBtnLblSet a set of labels to describe
registration buttons. Examples: register, confirm,
confirm registration, finalize, continue, create ac-
count, finalize registration, complete, register.
6. RegIconSet a set of icons to represent
registration buttons.
Checklist:
1. The prototype presents a data entry form with
one or more fields whose values are not filled in.
2. The form fields represent attributes of a user
entity in the application domain.
3. The form has a user identification, a password
and a retype-password fields.
4.The label of the user identification field belongs
to a set of labels used to describe user identification
fields (UserIdLblSet).
5. The label of the password field belongs to
a set of labels used to describe password fields
(PwdLblSet).
6. The label of the retype-password field belongs
to a set of labels used to describe retype-password
fields (RetPwdLblSet).
7. The form has a registration button whose label
belongs to a set of labels used to describe registration
buttons (RegBtnLblSet), or whose icon belongs to
a set of icons used to represent registration buttons
(RegIconSet).
Related heuristics: Registration
ID: HEU2-02
Name: Authentication
Synonyms: Login, log in
Definition: Credential information form for accessing
application resources.
Explanation: A user types in identification and a
password to access restricted application features.
Examples: A user access a social media application.
Additional information:
1. UserIdLblSet a set of labels to describe user
identification fields. Examples: id, login, email,
identification, name, user, username, registration
number, phone number.
2. PwdLblSet a set of labels used to describe
password fields. Examples: password, keyword.
3. AuthBtnLblSet a set of labels to describe
authentication buttons. Examples: login, log in, con-
nect, enter, log in with google, log in with facebook,
log in with gmail account, continue.
4. AuthIconSet a set of icons to represent
authentication buttons.
Checklist:
1. The prototype presents a data entry form with
a user identification field and a password field.
2. The label of the user identification field belongs
to a set of labels used to describe user identification
fields (UserIdLblSet).
3. The label of the password field belongs to
a set of labels used to describe password fields
(PwdLblSet).
4. The form presents a button whose label be-
longs to a set of labels used to describe authentication
ICEIS 2022 - 24th International Conference on Enterprise Information Systems
242
buttons (AuthBtnLblSet) or whose icon belongs to a
set of icons used to represent authentication buttons
(AuthIconSet).
5. Optionally, the prototype presents: a Forgot
Password link; a Login with another login system
option.
Related heuristics: none
ID: HEU2-07
Name: Search
Synonyms:
Find
Definition: Search engine
Explanation: A Search prototype illustrates how users
enter keywords to find information in a software
application.
Examples: A user looks for a book by its name in an
online bookstore.
Additional information:
1. List of user entities and attributes from the
application domain.
2. SrchLblSet a set of labels to describe search
term fields. Examples: search by name, type the ID,
search product, what are you looking for.
3. SrchBtnLblSet a set of labels to describe
search buttons. Examples: search, find.
4. SrchIconSet – a set of icons to represent search
buttons.
Checklist:
1. The prototype presents a form with a text field
related to a domain entity to inform the search terms.
2. The text field label belongs to a set of labels
used to describe search term fields (SrchLblSet).
3. The prototype presents a Button whose label
belongs to a set of labels used to describe search
buttons (SrchBtnLblSet) or whose icon belongs
to a set of icons used to represent search buttons
(SrchIconSet).
4. Optionally, the prototype presents a list of ob-
jects from a domain entity illustrating the search re-
sults.
Related heuristics: List, Content Filter
5 HEURISTICS VALIDATION
This section presents the validation experiment re-
sults of our design heuristics set. We validated the
set of heuristics through an experiment that analyzed
the perception of UI design experts. First, we selected
Figure 3: Example of an Account Registration prototype.
two HCI experts and two software development prac-
titioners for the experiment. Next, the experts evalu-
ated the heuristics through a survey and a focus group.
The following sections present the details of experi-
ment preparation and analysis of the results.
5.1 Preparation
To validate the heuristics, we conducted a Validation
through an Expert Judgment experiment (Qui
˜
nones
et al., 2018). As it was the first experiment to evaluate
the heuristics, the study aimed to assess the general
perception of specialists concerning the dimensions
proposed by Quinones et al. (2018) , which consist
of Utility, Clarity, Ease of Use, and Necessity of an
Additional checklist.
The study consisted of UI prototyping activities
for a fictitious e-commerce website. We asked the ex-
perts to design UI prototypes for three freely-chosen
requirements from those presented in the problem
specification.
The researchers recommended that the experts use
the set of heuristics during prototyping. To facilitate
using the heuristics set, we created a website with
the heuristics specifications accompanied by exam-
ples from the prototype base developed by the stu-
dents.
At the end of the prototypes’ design, the experts
filled out a questionnaire to evaluate the heuristics.
Table 4 presents the questions of this questionnaire.
After completing the questionnaire, we conducted a
focus group with the experts, where they evaluated the
heuristics and the prototyping experience using them.
Towards Meaning in User Interface Prototypes
243
Table 4: Survey on the experts’ judgment.
# Question
1 Were any of the heuristics useful in developing the prototype? Please check the ones that were used
below.
2 Have you made any modifications to your prototypes after consulting the heuristics? If yes, please
detail these changes.
3 Would you like to suggest modifications to existing heuristics? What modifications?
4 Would you like to suggest new heuristics? Which ones?
5 Would you like to suggest removing heuristics that you think are unnecessary? Which ones?
5.2 Results
We performed the analysis of the experiment results
according to the aspects of Utility, Clarity, Ease of
Use, and Necessity of an Additional Checklist, pre-
sented by Qui
˜
nones et al. (2018) .
Regarding the Utility dimension, all participants
claimed to have used heuristics to develop or improve
their prototypes. Specifically, the evaluators men-
tioned using the List, Search, View, Grid, Registra-
tion, Product Page, Chat, Shopping Cart, and Con-
tent Filter heuristics. One of the practitioners (P1)
reported that during the design of a Registration pro-
totype (HEU1-07), which he often implements, he
included elements that were not in the checklist and
were, in fact, unnecessary. Likewise, he missed sev-
eral elements when designing a shopping cart proto-
type, which he often does not implement. He noted
the missing items only after using the Shopping Cart
heuristic (HEU2-08). The other practitioner (P2)
mentioned that heuristics could be used alone or com-
bined to form more complex prototypes.
Regarding Clarity’s dimension, the experts re-
ported some problems. The two HCI specialists
and one of the practitioners indicated that they were
confused about the differences between the Search
(HEU2-07) and Content Filter (HEU1-01) heuristics
and between the List and Grid heuristics. Regard-
ing the Search and Content Filter heuristics, the ex-
perts reported that the specifications are complemen-
tary and could be unified. Regarding the List and Grid
heuristics, they reported that they are different ways
of presenting a group of objects, suggesting the in-
clusion of a carousel display. One of the HCI spe-
cialists did not understand the meaning of the Master-
Detail heuristic (HEU1-05). However, the practition-
ers reported it as a heuristic of great relevance in the
projects they develop. Also, one of the evaluators did
not understand the relationship between the View and
User Profile heuristics, suggesting that the most ap-
propriate relationship would be with the Registration
heuristic.
As for the Ease of Use dimension, the evaluators
reported some issues. For example, one of the HCI
experts found it difficult to memorize the name of the
heuristics. He also used the View heuristic (HEU1-
09) in the product page prototype design because
he did not find the corresponding heuristic (HEU2-
06). Despite this, the evaluators also reported ease
in applying the heuristics to the modeled prototype
and praised the navigation between heuristics through
links in the Related Heuristics section.
Regarding the Necessity of an Additional Check-
list dimension, the evaluators suggested the addition
of two new design heuristics: for the design of rating
and accessibility UIs. The suggestion of these new
categories showed that functionality heuristics need to
cover more diverse aspects of prototype design. The
specification of new heuristics demands new analysis
activities and will be the subject of future work.
In addition, the discussion on the problems re-
ported by the evaluators resulted in the following
changes in the heuristics set:
Addition of an indication to the accepted pass-
word format to the Account Registration heuristic
checklist (HEU2-01);
Unification of Content Filter (HEU1-01) and
Search (HEU2-07) heuristics;
Unification of List (HEU1-04) and Grid (HEU1-
03) heuristics. Addition of a Carousel as an option
for displaying an object list.
Addition of a relationship between User Profile
(HEU2-09) and Registration (HEU1-07) heuris-
tics.
The heuristics presented in this work specify what
distinguishes a prototype of a particular category.
However, the evaluators confused their purpose in
several situations. They understood that the heuris-
tics indicated how to design prototypes concerning
the arrangement of components on the screen. For
instance, one of the HCI experts (P3) reported that
a Content Filter (HEU1-01) example used bright red
and green colors, suggesting it as a design recommen-
dation. The other HCI expert (P4) reported that he de-
cided to design a Registration prototype (HEU1-07)
in multiple pages, requiring too many user clicks, by
looking at examples of this heuristic.
When analyzing the experiment data, we verified
that the availability of design examples from the pro-
ICEIS 2022 - 24th International Conference on Enterprise Information Systems
244
totype base obtained in Step 2 induced the partici-
pants to understand that the prototypes should have
a similar appearance to the examples. Another factor
we identified was that the Grid (HEU1-03) and List
(HEU1-04) heuristics contained indications of how
the objects should be arranged in their specification,
in disagreement with the other heuristics and the cat-
alog’s purpose.
6 DISCUSSIONS
In this work, we presented a set of design heuristics
for UI prototypes. These heuristics and their framing
at different semantic levels can be used to assist in
user interface design tasks, such as searching for in-
terfaces with similar functionality or verifying design
adequacy to predefined patterns in the heuristics.
Compared to image-based UI designs classifica-
tion methods, the heuristic-based approach has the
advantage of being based on the semantic character-
istics of the UI design. Therefore, it is not sensi-
tive to the placement of UI components or interfered
with by overlayed menus and notifications. Despite
that, approaches based on images, image metadata,
and user interaction data (Zecheng et al., 2020) can
be combined with the heuristic-based approach dur-
ing the classification process, as indicated by Li et al.
(2021), or complemented after classification, where
the heuristics would work as a checklist to verify the
adequacy of the UI design.
Regarding the aspects that are semantically ana-
lyzed, the work by Liu et al. (2018) classifies struc-
tural components, text buttons, and icons, located at
a semantic level prior to those presented in this ar-
ticle, which classifies interfaces as a whole and not
isolated elements. The Enrico dataset (Leiva et al.,
2020), in turn, filtered and classified a subset of the
Rico dataset, finding 20 categories, although it does
not consider semantic levels. Their work lists several
applications of the labeled dataset, some of which ap-
ply to the dataset we present in this work: UI descrip-
tion based on the identified label, UI classification,
listing the most likely labels for a UI, and optimized
search of rank-based GUIs.
Like our work, the work of Chen et al. (2020a)
also identified tags categories at different semantic
levels: platform, color, app functionality, screen func-
tionality, and screen layout. In their work, each UI
can receive up to one tag from each of the identified
categories. In our approach, most UI prototypes fit
into only one category or class, except for prototypes
labeled with more than one class.
Some of the classes identified in this work have
few instances, limiting the relevant features found in
the examples to define the respective heuristics and
making it difficult to develop a UI prototype classi-
fier in the following steps. To mitigate this problem,
we can extend our dataset by incorporating labeled
databases as the Enrico dataset (Leiva et al., 2020).
7 CONCLUSION
This research presented an approach to classifying in-
terface prototypes based on heuristics. First, we pre-
sented the classes of UI prototypes identified during
the labeling process and the respective set of identi-
fication heuristics. We also identified three semantic
levels in the prototypes, which supported the creation
of heuristics.
The heuristics we present in this article are not a
closed set, as we created them on the basis of our set
of UI prototypes. In the future, the dataset can be
extended by incorporating prototypes of new proto-
typing activities with students, resulting in new cate-
gories and heuristics.
The next stage of our work consists of using the
set of heuristics to develop a UI prototype classifica-
tion and analysis tool. Our goal is to support the in-
terface design process, offering deeper analysis than
those performed in similar works (Lee et al., 2020),
by incorporating semantic information into the analy-
sis process.
ACKNOWLEDGEMENTS
We are grateful to the Amazonas State Prosecution
Service MPE/AM for the incentive to this research
provided under the Program for the Functional Im-
provement of Employees.
This research, carried out within the scope of
the Samsung-UFAM Project for Education and Re-
search (SUPER), according to Article 48 of De-
cree 6.008/2006(SUFRAMA), was funded by Sam-
sung Electronics of Amazonia Ltda., under the terms
of Federal Law 8.387/1991, through agreement
001/2020, signed with Federal University of Ama-
zonas and FAEPI, Brazil. This research was also
supported by the Brazilian funding agency FAPEAM
through process number 062.00150/2020, the Coor-
dination for the Improvement of Higher Education
Personnel-Brazil (CAPES) financial code 001 and,
the S
˜
ao Paulo Research Foundation (FAPESP) under
Grant 2020/05191-2.
Towards Meaning in User Interface Prototypes
245
REFERENCES
Bernal-C
´
ardenas, C., Moran, K., Tufano, M., Liu, Z., Nan,
L., Shi, Z., and Poshyvanyk, D. (2019). Guigle: A gui
search engine for android apps. In 2019 IEEE/ACM
41st ICSE-Companion, pages 71–74. IEEE.
Bjarnason, E., Lang, F., and Mj
¨
oberg, A. (2021). A model
of software prototyping based on a systematic map.
In Proceedings of the 15th ACM/IEEE International
Symposium on Empirical Software Engineering and
Measurement (ESEM), pages 1–11.
Calle-Escobar, M., Mej
´
ıa-Guti
´
errez, R., Nadeau, J.-P., and
Pailhes, J. (2016). Heuristics-based design process.
International Journal on Interactive Design and Man-
ufacturing (IJIDeM), 10(4):369–386.
Chen, C., Feng, S., Liu, Z., Xing, Z., and Zhao, S. (2020a).
From lost to found: Discover missing ui design se-
mantics through recovering missing tags. Proc. ACM
Hum.-Comput. Interact., 4(CSCW2).
Chen, J., Chen, C., Xing, Z., Xia, X., Zhu, L., Grundy,
J., and Wang, J. (2020b). Wireframe-based ui design
search through image autoencoder. ACM Transactions
on Software Engineering and Methodology (TOSEM),
29(3):1–31.
Deininger, M., Daly, S. R., Sienko, K. H., and Lee, J. C.
(2017). Novice designers’ use of prototypes in engi-
neering design. Design Studies, 51:25–65.
Deka, B., Huang, Z., Franzen, C., Hibschman, J., Afergan,
D., Li, Y., Nichols, J., and Kumar, R. (2017). Rico:
A mobile app dataset for building data-driven design
applications. In Proceedings of the 30th Annual ACM
Symposium on User Interface Software and Technol-
ogy, pages 845–854.
Dow, S., Fortuna, J., Schwartz, D., Altringer, B., Schwartz,
D., and Klemmer, S. (2011). Prototyping dynamics:
sharing multiple designs improves exploration, group
rapport, and results. In Proceedings of the SIGCHI
conference on human factors in computing systems,
pages 2807–2816.
Felderer, M. and Travassos, G. H. (2020). Contempo-
rary Empirical Methods in Software Engineering.
Springer.
Figueiredo, E., Sant’Anna, C., Garcia, A., and Lucena, C.
(2012). Applying and evaluating concern-sensitive
design heuristics. Journal of Systems and Software,
85(2):227–243.
Garcia, A., Silva da Silva, T., and Selbach Silveira, M.
(2017). Artifacts for agile user-centered design: a sys-
tematic mapping. In Proceedings of the 50th Hawaii
International Conference on System Sciences.
Gigerenzer, G. (2008). Why heuristics work. Perspectives
on psychological science, 3(1):20–29.
Huang, F., Canny, J. F., and Nichols, J. (2019). Swire:
Sketch-based user interface retrieval. In Proceedings
of the 2019 CHI Conference on Human Factors in
Computing Systems, pages 1–10.
K
¨
apyaho, M. and Kauppinen, M. (2015). Agile require-
ments engineering with prototyping: A case study.
In 2015 IEEE 23rd International Requirements Engi-
neering Conference (RE), pages 334–343.
Lauff, C. A., Knight, D., Kotys-Schwartz, D., and
Rentschler, M. E. (2020). The role of prototypes in
communication between stakeholders. Design Stud-
ies, 66:1–34.
Lee, C., Kim, S., Han, D., Yang, H., Park, Y.-W., Kwon,
B. C., and Ko, S. (2020). Guicomp: A gui design
assistant with real-time, multi-faceted feedback. In
Proceedings of the 2020 CHI Conference on Human
Factors in Computing Systems, pages 1–13.
Leiva, L. A., Hota, A., and Oulasvirta, A. (2020). Enrico:
A dataset for topic modeling of mobile ui designs. In
22nd International Conference on Human-Computer
Interaction with Mobile Devices and Services (Mo-
bileHCI’20 Extended Abstracts).
Lenarduzzi, V. and Taibi, D. (2016). Mvp explained: A sys-
tematic mapping study on the definitions of minimal
viable product. In 2016 42th Euromicro Conference
on Software Engineering and Advanced Applications
(SEAA), pages 112–119. IEEE.
Li, T. J.-J., Popowski, L., Mitchell, T., and Myers, B. A.
(2021). Screen2Vec: Semantic Embedding of GUI
Screens and GUI Components. Association for Com-
puting Machinery, New York, NY, USA.
Liu, T. F., Craft, M., Situ, J., Yumer, E., Mech, R., and Ku-
mar, R. (2018). Learning design semantics for mobile
apps. In The 31st Annual ACM Symposium on User
Interface Software and Technology, pages 569–579.
ACM.
Moran, K., Li, B., Bernal-C
´
ardenas, C., Jelf, D., and Poshy-
vanyk, D. (2018). Automated reporting of gui de-
sign violations for mobile apps. In Proceedings of the
40th International Conference on Software Engineer-
ing, pages 165–175.
Packevi
ˇ
cius,
ˇ
S., Barisas, D., U
ˇ
saniov, A., Guogis, E., and
Barei
ˇ
sa, E. (2018). Text semantics and layout de-
fects detection in android apps using dynamic execu-
tion and screenshot analysis. In International Con-
ference on Information and Software Technologies,
pages 279–292. Springer.
Qui
˜
nones, D., Rusu, C., and Rusu, V. (2018). A method-
ology to develop usability/user experience heuristics.
Computer Standards & Interfaces, 59:109–129.
Shah, A. K. and Oppenheimer, D. M. (2008). Heuristics
made easy: an effort-reduction framework. Psycho-
logical bulletin, 134(2):207.
Toxboe, A. (2022). User interface design pattern gallery.
http://ui-patterns.com/.
Yilmaz, S., Daly, S. R., Seifert, C. M., and Gonzalez, R.
(2016). Evidence-based design heuristics for idea gen-
eration. Design Studies, 46:95–124.
Zang, X., Xu, Y., and Chen, J. (2021). Multimodal icon
annotation for mobile applications. In Proceedings of
the 23rd International Conference on Mobile Human-
Computer Interaction, pages 1–11.
Zecheng, H., Sunkara, S., Zang, X., Xu, Y., Liu, L.,
Wichers, N., Schubiner, G., Lee, R., and Chen, J.
(2020). Actionbert: Leveraging user actions for
semantic understanding of user interfaces. arXiv
preprint arXiv:2012.12350.
ICEIS 2022 - 24th International Conference on Enterprise Information Systems
246