End-user Development for Smart Spaces: A Comparison of Block and

Data-ﬂow Programming

Marcel Altendeitering

and Sonja Schimmler

Fraunhofer ISST, Dortmund, Germany

Fraunhofer FOKUS, Berlin, Germany

Keywords:

Block Programming, Data-ﬂow Programming, Visual Languages, Smart Spaces.

Abstract:

Block and data-ﬂow programming are well-known concepts for enabling end users to visually create their own

customized solutions. They both offer comprehensive visual interfaces and are becoming popular within the

smart spaces domain. However, there is currently no systematic, comparative evaluation of the two concepts

in the domain. In this user study, we implemented two prototypes for block and data-ﬂow programming and

compared their performance on typical usage scenarios in common smart spaces. We used a mixed methods

approach of quantitative and qualitative analysis to gain an in-depth understanding of the user experience.

Our results show that data-ﬂow programming is overall better received by users than block programming

and is considered being state-of-the art and visually more appealing. For block programming, our results

reveal that participants appreciate the playful character and the feedback provided. Our study contributes

to the improvement of block and data-ﬂow solutions in place and discusses aspects relevant to the future

advancement of end-user development in smart spaces.

1 INTRODUCTION AND

RELATED WORK

Smart spaces that support people in their daily lives

are more and more becoming a reality (Mennicken

et al., 2014). The availability of affordable hardware

fosters the introduction of smart devices in different

areas and tailored services are connecting these de-

vices in a useful manner (Jensen et al., 2018). How-

ever, the smart spaces that surround our modern lives

are inherently complex and make it difﬁcult to design

one-size-ﬁts-all solutions (Mennicken et al., 2014).

Not all services that the inhabitants of smart spaces

(e.g., smart homes, smart cities, etc.) desire are

known a priori. End users have the best knowledge

of their requirements, can build the functions they de-

sire, and invent novel services that were not intended

by system designers (Coutaz et al., 2010; Jensen et al.,

2018; Mennicken et al., 2014).

At the same time, users of smart spaces want to

be in control and have the ability to regulate all as-

pects of their lives with cyber-physical technologies

surrounding them (Davidoff et al., 2006). However,

existing solutions for conﬁguring smart spaces are of-

ten considered as being too complicated or not offer-

ing sufﬁcient functionalities (Reisinger et al., 2017;

Altendeitering and Schimmler, 2020).

To address these problems, low code visual pro-

gramming languages have become wide-spread meth-

ods for interacting with smart spaces as they enable

end-users without coding knowledge to create cus-

tomized solutions (Reisinger et al., 2017; Altendeit-

ering and Schimmler, 2020). To provide adequate us-

ability and accessibility, it is important to utilize a vi-

sual language that is suitable for a speciﬁc user group.

Otherwise, inexperienced users could face barriers

that hinder them from fully exploiting the possibili-

ties offered by smart spaces (Davidoff et al., 2006).

In literature and practice, form-ﬁlling (e.g.,

(IFTTT, 2021)), data-ﬂow (e.g., (Node-RED, 2021)),

and block programming (e.g., (Blockly, 2021;

Scratch, 2021)) are well-established visual program-

ming concepts. These approaches have in com-

mon that they try to reduce the semantic, syntac-

tic, and pragmatic barriers to programming inexpe-

rienced users are facing. They are different in their

expressiveness, visual representation and cognitive

style (Ur et al., 2014; Reisinger et al., 2017). Block

and data-ﬂow programming tools come with a more

comprehensive visual interface as compared to form-

ﬁlling solutions and are becoming increasingly pop-

ular within the smart spaces domain (Altendeitering

Altendeitering, M. and Schimmler, S.

End-user Development for Smart Spaces: A Comparison of Block and Data-ﬂow Programming.

DOI: 10.5220/0010983200003203

In Proceedings of the 11th International Conference on Smart Cities and Green ICT Systems (SMARTGREENS 2022), pages 15-22

ISBN: 978-989-758-572-2; ISSN: 2184-4968

and Schimmler, 2020; Reisinger et al., 2017).

Several studies examined the differences between

programming paradigms in different scenarios. Most

of them focus on the differences between graphical

and textual approaches (e.g., (Stein and Hanenberg,

2011; Weintrop and Wilensky, 2015)). The two stud-

ies that come closest to the idea of our paper are

(Reisinger et al., 2017) and (Mason and Dave, 2017).

In the paper of (Reisinger et al., 2017), the authors

compared form-ﬁlling and data-ﬂow programming

concepts using exemplary usage scenarios. They

found that data-ﬂow excels form-ﬁlling through a bet-

ter comprehensiveness and usability. In the short pa-

per of (Mason and Dave, 2017), the authors inves-

tigated the differences between block and data-ﬂow

programming, similar to our study. However, only

preliminary results were presented and no in-depth in-

vestigation of the two concepts was provided.

With our study we want to contribute such an in-

depth analysis and help researchers and practitioners

to distinguish the concepts and determine which one

is preferable. We, therefore, conducted a compar-

ative evaluation and implemented a visual language

in two exemplary prototypes for block and data-ﬂow

programming that enable the creation of typical usage

scenarios in smart spaces. Speciﬁcally, we formulated

the following research question:

• Research Question: What are the differences in

user perception between block and data-ﬂow pro-

gramming within smart spaces? Can known dif-

ferences be replicated?

The remainder of this article is structured as fol-

lows. In Section 2, we start by describing our re-

search methodology including the prototypes we re-

alized and the experimental setting. In Section 3, we

present the results of our study. We conclude with a

discussion of the results in Section 4.

2 RESEARCH METHODOLOGY

To answer the proposed research questions, we con-

ducted a mixed methods user study on two proto-

types for the visual programming paradigms under

investigation. We decided to apply a mixed meth-

ods approach of (1) quantitative and (2) qualitative

analysis to increase the robustness of our results by

triangulating different methods for the same phe-

nomenon (Graziano and Raulin, 1993; Zhao et al.,

2019). Speciﬁcally, we followed the example of

(Reisinger et al., 2017) and conducted a laboratory

user study with twelve participants, a sufﬁcient num-

ber for ﬁnding most usability issues (Nielsen and

Landauer, 1993). Hereby, the participants were con-

fronted with three tasks based on typical usage sce-

narios in smart spaces and asked to solve these tasks

using both prototypes (Zhao et al., 2019).

2.1 Participants

Of the twelve participants ﬁve were female and seven

were male. Potential candidates were recruited from

the scope of a German smart spaces research project

and approached by email. The participants’ mean age

was 26.67 with a standard deviation (SD) of 3.61. The

group of participants had an average computer liter-

acy with a mean of 2.92 on a scale from 1 to 5. The

programming knowledge was slightly below average

with a mean of 2. Only one participant reported to

have high programming skills. All other participants

reported low to medium programming skills. Both,

the computer literacy and programming knowledge,

were self-reported by the participants.

Everyone was familiar with the concept of smart

spaces or smart homes. Two persons had heard of

Node-RED before, but did not use it on a regular ba-

sis. The other participants had not heard of either

Node-RED or Blockly before. Overall, our user group

is not representative for the general population, but

matches a typical smart spaces user group, which con-

sists of young adults with an interest in digital tech-

nologies and limited programming skills (Reisinger

et al., 2017; Altendeitering and Schimmler, 2020).

2.2 Experimental Setting

To evaluate the participants’ performance, we de-

ﬁned three common usage scenarios for programming

smart spaces. We based these usage scenarios on use

cases that were presented in related studies (Altendei-

tering and Schimmler, 2020; Reisinger et al., 2017;

Mason and Dave, 2017). We decided to use generic

usage scenarios instead of concrete task descriptions

to gain a better understanding of the ease of use of

the two prototypes for inexperienced users. The three

scenarios vary in their difﬁculty, which we speciﬁed

as the amount of block or data-ﬂow items that are re-

quired to realize the respective scenario (see Table 1).

For enabling the participants to realize the de-

scribed usage scenarios we created a visual language

and two prototypes for block and data-ﬂow program-

ming. For this, we extended the standard vocabu-

lary of established solutions with custom blocks (for

block programming) and nodes (for data-ﬂow pro-

gramming). Speciﬁcally, we created artifacts for

accessing the relevant data sources from the smart

spaces (i.e., Smart Parking, Smart Home, Smart Me-

SMARTGREENS 2022 - 11th International Conference on Smart Cities and Green ICT Systems

Table 1: Usage scenarios for smart spaces.

Scenario Elements

[Flow/Block]

Difﬁculty Description Custom Artifacts

Smart Parking (a) 6/7 easy

After work, I want to spend as lit-

tle time as possible looking for a

parking lot. Notify me on the way

home if there are free parking lots

available.

Energy consumption

sensor for TV, fridge,

and oven; Warning and

Information engine

Smart Home (b) 7/9 medium

I want the air quality in my apart-

ment to be good. Open a window

when the air quality inside is poor

and good outside.

Air Quality sensor inside

and outside; Open Win-

dow function; method

for determining whether

Air Quality is good

Smart Meter (c) 13/13 hard

The power consumption of my

three appliances (TV, fridge,

oven) is very high. Warn me if

I consume more than 10 kWh

and inform me which device

consumes the most.

Energy consumption

sensor for TV, fridge,

and oven; Warning and

Information engine

ter) and organized these in categories that are named

after the respective smart space. We implemented a

demo back-end to provide the data sources with ex-

emplary sensor streams. Additionally, we included a

separate category that contains useful artifacts (e.g.,

for determining whether the air quality is sufﬁcient)

to simplify the usage of the two prototypes. Overall,

we included twelve new custom artifacts to enable the

creation of the usage scenarios (see Table 1).

To create a realistic experience, we used existing

solutions (e.g., (Xiaomi, 2021; Homee, 2021; Lox-

one, 2021)) as an example for the styling and naming

of our custom artifacts. All other required artifacts

remained in the standard categorization of the block

and data-ﬂow programming solutions we used. Both

prototypes offered an open canvas and allowed partic-

ipants to freely combine blocks or nodes in their own

way. There was no pre-deﬁned structure and all tasks

offered multiple ways to achieve the desired outcome.

We implemented the block-programming proto-

type on Google Blockly (Blockly, 2021). We decided

to extend Blockly as it is a well-established and eas-

ily extensible block programming language. Figure 1

shows an implementation of the smart home task.

For the data-ﬂow programming prototype we used

Node-RED (Node-RED, 2021) as a basis. We chose

to use Node-RED as it has a broad community that of-

fers many extensions and custom developments and is

well documented. In addition to creating new nodes,

Node-RED offers the possibility to create sub-ﬂows,

which are made up of existing Node-RED blocks. We

used this possibility instead of creating new nodes

wherever possible. Figure 2 displays a realization of

the smart home task using the Node-RED based data-

ﬂow prototype.

2.3 Measures & Data Analysis

Following the described mixed methods approach,

our data analysis has two separate parts: quantitative

and qualitative analysis.

2.3.1 Quantitative

With our quantitative analysis we followed the exam-

ple of (Reisinger et al., 2017), who conducted a sim-

ilar study. It consists of six measures based on the

usability factors (UFs) proposed by (Nielsen, 1994)

(see Table 2).

For the Task Completion Time (measure 1) we

manually tracked the time the participants spent on

the prototype trying to solve the tasks at hand. To

determine the Task Completion Rate (measure 2), we

analyzed the participants’ ﬁnal programs after com-

pletion. We deﬁned this measures as the number

of elements used in the optimal solution minus the

number of corrective steps needed. For measures 3

and 4 the participants rated their Perceived Success

and the Perceived Difﬁculty of the respective pro-

totype on a 7-item Likert scale. For analysis, we

used a MANOVA with these measures as dependent

variables and the respective prototype as independent

variable (Graziano and Raulin, 1993).

We investigated measure 5 with a standardized

User Experience Questionnaire (UEQ) as introduced

by (Laugwitz et al., 2008). For an analysis of the UEQ

End-user Development for Smart Spaces: A Comparison of Block and Data-ﬂow Programming

Figure 1: Block programming prototype showing the smart home task (b).

Figure 2: Data-ﬂow programming prototype showing the smart home task (b).

Table 2: Six measures for performance and usability analy-

sis (based on (Nielsen, 1994)).

# Measure Usability Factor

(UF)

1 Task Completion Time Efﬁciency

2 Task Completion Rate Errors

3 Perceived Success Satisfaction, Learn-

ability

4 Perceived Difﬁculty Satisfaction, Learn-

ability

5 UEQ Satisfaction, Learn-

ability

6 Self-Assessment Memorability,

Learnability

we used descriptive statistical measures. These con-

tain the calculation of mean, standard deviation (SD),

and effect sizes (Graziano and Raulin, 1993). Effect

sizes were calculated with Cohen’s d (Cohen, 1992).

For the Self-Assessment (measure 6) we asked the

participants to draw two of their solutions, one for

each prototype, from memory and asked how conﬁ-

dent they felt about the correctness of their drawing

on a 7-item Likert scale. A high recollection rate and

self-assessment is an important factor for visual pro-

gramming in smart spaces as it supports the coopera-

tive work with friends and families (Reisinger et al.,

2017; Davidoff et al., 2006).

2.3.2 Qualitative

The qualitative part of our data analysis adds an in-

depth analysis to the quantitative measures. Fur this

purpose, we added open questions on the positive and

negative experiences made while using the prototypes

to the questionnaires. This allowed us to gain an in-

depth understanding of the phenomenons the partic-

ipants experienced while using the prototypes (Zhao

et al., 2019). To systematically analyze these ques-

tions, we used content analysis as a common approach

for analyzing unstructured data like written text. By

repeatedly chunking different sentences into fewer re-

lated categories and sub-categories, it delivers a dense

description of the phenomenon under investigation

(Elo and Kyng

as, 2008; Vaismoradi et al., 2013). Fol-

lowing (Vaismoradi et al., 2013), content analysis has

a low level of interpretation and is useful for gaining

insights in usability issues experienced by users of a

product or software.

For the content analysis two researchers indepen-

dently coded the answers for positive and negative

aspects. Afterwards, they iterative grouped related

codes that describe the same phenomenon into ab-

stract themes. Potential differences in the derived

codes and themes were clariﬁed in subsequent discus-

sions between the two researchers.

2.4 Procedure

The participants had a one-hour time frame to com-

plete all tasks and questionnaires. They worked on

the tasks alone with one of the authors in the same

room for tracking the time needed and handing out the

tasks. Unnecessary or distracting items were removed

from the room to allow for a neutral setting (Bordens

and Abbott, 2002). We initiated the user study with a

short explanation of smart spaces and visual program-

ming to ensure a common understanding. Afterwards,

the participants started working on the three tasks in

SMARTGREENS 2022 - 11th International Conference on Smart Cities and Green ICT Systems

ascending difﬁculty. Hereby, six people started with

one prototype, while the remaining six started with

the other one.

Once they ﬁnished their tasks, the present author

noted quantitative measures, the participants rated the

perceived success and difﬁculty, and ﬁlled out the

UEQ. Moreover, they answered open questions on

both prototypes for qualitative analysis. After this

ﬁrst iteration the participants repeated the same proce-

dure for the other prototype. Once they ﬁnished their

tasks and questionnaires, we started the measurement

of the self-assessment. At last, we requested the par-

ticipants to ﬁll out a short demographic questionnaire

to enable the differentiation of the responses by gen-

der, age, and other factors. Overall, they worked on

scenarios a, b, and c twice, answered two question-

naires, and drew two solutions on paper.

3 RESULTS

With regard to the study’s procedure we found that,

overall, it was a good ﬁt and helped us to be in accor-

dance with our research goal. We were able to gain

insights in the usability of the two prototypes and an

answer to our research question. However, in future

studies we would allow the users more time in using

the applications to investigate the learning effect and

select a more diverse user group. In the following

two sections we present our quantitative and qualita-

tive results in detail.

3.1 Quantitative Results

We summarized the results of measures 1-4 in Ta-

ble 3. An analysis of the Task Completion Time

shows that users needed almost a minute longer with

Blockly (7.52) as compared to Node-RED (6.67) to

realize all scenarios. The most important difference

between both prototypes is the Task Completion Rate,

which shows that users were signiﬁcantly more suc-

cessful with the data-ﬂow prototype. For Perceived

Success, Perceived Difﬁculty, and Self-Assessment we

did not observe signiﬁcant differences between both

programming paradigms.

The results of measure 5 (UEQ) are shown in Fig-

ure 3. The UEQ included questions in six dimensions.

Only in the dimension Perspicuity (p = .69, |d| = .09)

the block approach (0.64) excels data-ﬂow (0.42). In

all other dimensions, data-ﬂow scored higher or sim-

ilar. In case of Attractiveness (p = .03, |d| = .5) and

Novelty (p = .01, |d| = .8) with a signiﬁcant differ-

ence and medium effect size.

Finally, for the Self-Assessment (measure 6) we

Figure 3: Results for the UEQ (measure 5). Scores are bal-

anced (-3 to +3) and grouped by GUI with 75% conﬁdence

interval.

observed that Node-RED scored (2.22) considerably

higher than Blockly (0.78), having a big gap between

the means (p = .09, |d| = .9). Scores were balanced

from -3 to 3. Although not statistically signiﬁcant

the result shows an advantage for Node-RED over

Blockly that should be further investigated.

3.2 Qualitative Results

To visualize the results of our qualitative analysis,

we organized the themes, which we derived from the

open questions, as two word clouds. We chose to use

word clouds for visualization as they offer the reader

an easy and intuitive way for identifying the essen-

tial feedback for both programming paradigms. They,

furthermore, allow us to compare the balance between

positive and negative remarks for the data-ﬂow and

block programming prototypes. Green themes have

a positive sentiment, while red represents a negative

theme. The size of the labels relates to the number of

occurrences of the respective theme.

The strongest positive theme for data-ﬂow pro-

gramming (see Figure 4) was simple representation,

which was mentioned nine times (+9). People liked

the concept of the ﬂow model (+4) in general and that

the connections were clear. It was easy to understand

that the data ”ﬂows like a river” and is transformed

by nodes in sequence. On the contrary, people criti-

cized that the interaction with the nodes was not dy-

namic. There was no feedback (-6) if the connection

of two nodes made any sense or if they are incompat-

ible. This left error handling and debugging entirely

up to the user.

Block-programming’s (see Figure 5) strongest

positive themes included the intuitive puzzle me-

chanic (+3) together with connection feedback (+3)

as elements would ”jump away” when incompati-

ble. Additionally, people ﬁnd the simple representa-

tion (+2) and the interactive limitation (+2) appealing.

End-user Development for Smart Spaces: A Comparison of Block and Data-ﬂow Programming

Table 3: Descriptive measures and MANOVA using task observations as dependent variables and the respective prototype as

independent variable.

Mean (± sd) MANOVA

Blockly Node-RED

Task Completion Time 7.52 6.67 F=1.65

in minutes (±3.12) (±4.40) p=.22

Task Completion Rate 83 94 F=6.72

as ratio % (±14) (±9) p=.02

Perceived Success 5.52 5.33 F=0.86

7-item Likert Scale (±1.78) (±1.90) p=.37

Perceived Difﬁculty 4.59 4.78 F=1.67

7-item Likert Scale (±1.65) (±1.55) p=.21

Figure 4: Data-ﬂow: 8 positive (21 total) and 6 negative (13

total) themes.

However, the puzzle mechanic also comes with some

downsides including a high complexity, difﬁcult inter-

action, and unappealing representation (all -3). Over-

all, data-ﬂow has a larger share of positive themes (21

positive / 13 negative) as compared to block program-

ming (16 positive / 14 negative).

4 DISCUSSION, LIMITATIONS

AND FUTURE WORK

By comparing the two prototypes for block and

data-ﬂow programming, we can summarize that the

data-ﬂow (Node-RED) excels the block prototype

(Blockly) in several dimensions. Most importantly,

we observed that the users were signiﬁcantly more

successful in completing the tasks with the Node-

RED prototype. They also felt more conﬁdent in us-

ing the data-ﬂow application and preferred the user

interface. In this sense, our results are tallied with

Figure 5: Block: 9 positive (16 total) and 6 negative (14

total) themes.

previous studies (Reisinger et al., 2017; Leitner et al.,

2013; Mason and Dave, 2017) and indicate that the

data-ﬂow visual programming paradigm might be

preferable in smart spaces. However, since most of

our measures were not statistically signiﬁcant further

studies are necessary to deepen the analysis of the two

concepts.

With regard to the data-ﬂow prototype, users

found the solution visually more appealing as com-

pared to the block prototype and considered it as be-

ing state-of-the-art. They also felt more secure in us-

ing the prototype and liked the freedom offered by

the ﬂow-based interaction model. These results are

interesting as people also noted that they sometimes

feel overwhelmed by the possibilities and criticized

the lack of feedback. Intelligent hints or limitations

in node connections could help make the tool more

useful and improve on Hick’s law (Rosati, 2013).

Looking at the block prototype, the participants

liked the heightened perspicuity they received from

the puzzle mechanic, which offers a more playful

character. The puzzle mechanic also provides users

with immediate feedback on their programs, which

SMARTGREENS 2022 - 11th International Conference on Smart Cities and Green ICT Systems

was well received. On the other hand, the interac-

tion model and the visual representation were seen as

complex and unappealing. A more modern user inter-

face and full-grown IDE could help overcome these

issues and increase the usability of the prototype (In-

ayama and Hosobe, 2018).

For both tools we observed downsides in their in-

teraction model, which led participants to perceive the

tools as complex and rather difﬁcult. Future develop-

ments should combine the strengths of both concepts,

especially a simple and modern user interface with

feedback mechanisms. Additionally, new formats of

human-computer interaction, such as conversational

AI (e.g., chatbots or voice interaction) could sup-

port inexperienced users by indicating potential pit-

falls (Jung et al., 2019; Inayama and Hosobe, 2018).

Although our study uncovered some interesting

details it is subject to several limitations. A study with

a larger and more-diverse group of participants could

shed additional light on the subject. Furthermore, our

study could be biased by the choice of the underly-

ing technology (i.e., Blockly and Node-RED) and the

interaction model of the particular tools. Other so-

lutions might offer different forms of interaction and

user feedback and achieve different results. It would,

moreover, be interesting to include the learning effect

as an additional metric in further studies and investi-

gate how well participants remember the user inter-

face after some time.

ACKNOWLEDGMENT

This research was partly supported by the German

Federal Ministry for Economic Affairs and Climate

Action (BMWK) and by the German Federal Ministry

of Education and Research (BMBF) under grant no.

16DII128 (“Deutsches Internet-Institut”).

We also thank Sakander Zirai for his support in de-

veloping the prototypes and conducting the research.

REFERENCES

Altendeitering, M. and Schimmler, S. (2020). Data-ﬂow

programming for smart homes and other smart spaces.

In 2020 IEEE Symposium on Visual Languages and

Human-Centric Computing (VL/HCC), pages 1–4.

IEEE.

Blockly (2021). A javascript library for

building visual programming editors.

https://developers.google.com/blockly. Last ac-

cessed on 19.11.2021.

Bordens, K. S. and Abbott, B. B. (2002). Research design

and methods: A process approach. McGraw-Hill.

Cohen, J. (1992). Statistical power analysis. Current direc-

tions in psychological science, 1(3):98–101.

Coutaz, J., Fontaine, E., Mandran, N., and Demeure, A.

(2010). Disqo: A user needs analysis method for

smart home. In Proceedings of the 6th Nordic Con-

ference on Human-Computer Interaction: Extending

Boundaries, pages 615–618.

Davidoff, S., Lee, M. K., Yiu, C., Zimmerman, J., and

Dey, A. K. (2006). Principles of smart home control.

In International conference on ubiquitous computing,

pages 19–34. Springer.

Elo, S. and Kyng

as, H. (2008). The qualitative content anal-

ysis process. Journal of advanced nursing, 62(1):107–

115.

Graziano, A. M. and Raulin, M. L. (1993). Research meth-

ods: A process of inquiry. HarperCollins College Pub-

lishers.

Homee (2021). node-red-contrib-homee.

https://ﬂows.nodered.org/node/node-red-contrib-

homee. Last accessed on 19.11.2021.

IFTTT (2021). Ifttt: A world that works for you.

https://ifttt.com/. Last accessed on 28.11.2021.

Inayama, Y. and Hosobe, H. (2018). Toward an efﬁcient

user interface for block-based visual programming.

In 2018 IEEE Symposium on Visual Languages and

Human-Centric Computing (VL/HCC), pages 293–

294. IEEE.

Jensen, R. H., Strengers, Y., Kjeldskov, J., Nicholls, L., and

Skov, M. B. (2018). Designing the desirable smart

home: A study of household experiences and energy

consumption impacts. In Proceedings of the 2018 CHI

Conference on Human Factors in Computing Systems,

pages 1–14.

Jung, H., Kim, H. J., So, S., Kim, J., and Oh, C. (2019).

Turtletalk: an educational programming game for

children with voice user interface. In Extended Ab-

stracts of the 2019 CHI Conference on Human Factors

in Computing Systems, pages 1–6.

Laugwitz, B., Held, T., and Schrepp, M. (2008). Construc-

tion and evaluation of a user experience questionnaire.

In Symposium of the Austrian HCI and usability engi-

neering group, pages 63–76. Springer.

Leitner, G., Fercher, A. J., and Lassen, C. (2013). End users

programming smart homes–a case study on scenario

programming. In International Workshop on Human-

Computer Interaction and Knowledge Discovery in

Complex, Unstructured, Big Data, pages 217–236.

Springer.

Loxone (2021). Loxone home automation.

https://www.loxone.com/dede/. Last accessed

on 19.11.2021.

Mason, D. and Dave, K. (2017). Block-based versus ﬂow-

based programming for naive programmers. In 2017

IEEE Blocks and Beyond Workshop (B&B), pages 25–

28. IEEE.

Mennicken, S., Vermeulen, J., and Huang, E. M. (2014).

From today’s augmented houses to tomorrow’s smart

homes: New directions for home automation research.

In Proceedings of the 2014 ACM International Joint

Conference on Pervasive and Ubiquitous Computing,

End-user Development for Smart Spaces: A Comparison of Block and Data-ﬂow Programming

UbiComp ’14, page 105–115, New York, NY, USA.

Association for Computing Machinery.

Nielsen, J. (1994). Usability engineering. Morgan Kauf-

mann.

Nielsen, J. and Landauer, T. K. (1993). A mathematical

model of the ﬁnding of usability problems. In Pro-

ceedings of the INTERACT’93 and CHI’93 conference

on Human factors in computing systems, pages 206–

213.

Node-RED (2021). Node-red: Low-code programming for

event-driven applications. https://nodered.org/. Last

accessed on 19.11.2021.

Reisinger, M. R., Schrammel, J., and Fr

ohlich, P. (2017).

Visual languages for smart spaces: End-user program-

ming between data-ﬂow and form-ﬁlling. In 2017

IEEE Symposium on Visual Languages and Human-

Centric Computing (VL/HCC), pages 165–169. IEEE.

Rosati, L. (2013). How to design interfaces for choice:

Hick-hyman law and classiﬁcation for information ar-

chitecture. In Slavic, A.; Salah, A.; Davies, C. Clas-

siﬁcation and visualization: interfaces to knowledge:

proceedings of the International UDC Seminar. The

Hague, The Netherlands, pages 125–138.

Scratch (2021). Scratch. https://scratch.mit.edu/. Last ac-

cessed on 19.11.2021.

Stein, D. and Hanenberg, S. (2011). Comparison of a vi-

sual and a textual notation to express data constraints

in aspect-oriented join point selections: A controlled

experiment. In 2011 IEEE 19th International Con-

ference on Program Comprehension, pages 141–150.

IEEE.

Ur, B., McManus, E., Pak Yong Ho, M., and Littman, M. L.

(2014). Practical trigger-action programming in the

smart home. In Jones, M., Palanque, P., Schmidt, A.,

and Grossman, T., editors, Proceedings of the SIGCHI

Conference on Human Factors in Computing Systems,

pages 803–812, New York, NY. ACM.

Vaismoradi, M., Turunen, H., and Bondas, T. (2013). Con-

tent analysis and thematic analysis: Implications for

conducting a qualitative descriptive study. Nursing &

health sciences, 15(3):398–405.

Weintrop, D. and Wilensky, U. (2015). To block or not

to block, that is the question: students’ perceptions

of blocks-based programming. In Proceedings of the

14th international conference on interaction design

and children, pages 199–208.

Xiaomi (2021). node-red-contrib-xiaomi-smart-devices.

https://ﬂows.nodered.org/node/node-red-contrib-

xiaomi-smart-devices. Last accessed on 19.11.2021.

Zhao, L., Loucopoulos, P., Kavakli, E., and Letsholo, K. J.

(2019). User studies on end-user service composi-

tion: a literature review and a design framework. ACM

Transactions on the Web (TWEB), 13(3):1–46.

SMARTGREENS 2022 - 11th International Conference on Smart Cities and Green ICT Systems