Inconspicuous Guides: A Qualitative Study to Understand the Degree of

Intrusiveness in Visualization Onboarding

Margit Pohl

1 a

, Benjamin Potzmann

1 b

, Christina Stoiber

2 c

and Wolfgang Aigner

2 d

TU Wien, Austria

St. P

olten University of Applied Sciences, Austria

{ﬁrst author, second author}@tuwien.ac.at, {third author, fourth author}@fhstp.ac.at

Keywords:

Visualization Onboarding, Learning, Accepted Annoyance, Visualization Literacy, Intrusiveness.

Abstract:

Current visualizations are becoming more and more complex. Visualization onboarding is a possibility to

assist users in understanding such visualizations. Nevertheless, previous research indicates that users tend to

ignore this kind of assistance, even if it is evident that they need help to interact with the system efﬁciently.

This investigation aims to collect prototypes of visualization onboarding systems taking into account the de-

gree of intrusiveness through a user inquiry using sketching and a questionnaire. We conducted a study with

65 participants. We asked them to what extent intrusion on the part of the system (e.g., compulsory tutorials,

automatic pop-up messages) would be acceptable. In addition, we asked them to create solutions to this prob-

lem. Most found less intrusive visualization onboarding forms (e.g., tooltips, optional and concise tutorials)

helpful. Compulsory visualization onboarding forms were considered less helpful. Participants also suggested

chatbots or ML algorithms as convenient solutions for visualization onboarding.

1 INTRODUCTION

The increasing size and complexity of modern

datasets often surpass the capabilities of traditional

visualizations like bar charts, line graphs, and pie

charts to effectively represent the underlying data.

As data becomes more complex, these methods fre-

quently fail to capture its full depth. While advanced

visualization techniques can address this, many users

struggle to interpret them, which leads to misinter-

pretations (B

orner et al., 2016; Galesic and Garcia-

Retamero, 2010). This can result in frustration and

reluctance to use these otherwise powerful tools.

Despite the availability of visualization onboard-

ing systems (Stoiber et al., 2022d; Stoiber, 2023;

Dhanoa et al., 2022) designed to guide users in inter-

preting and navigating these complex visualizations,

users tend to ignore them (Pohl et al., 2023). In-

stead, they often rely on a “trial-and-error” approach,

attempting to ﬁgure out the visual encoding and the

system’s functionality on their own (Mahmud et al.,

2020; Andrade et al., 2009). This behavior increases

https://orcid.org/0000-0001-7880-8702

https://orcid.org/0000-0002-5286-2433

https://orcid.org/0000-0002-1764-1467

https://orcid.org/0000-0001-5762-1869

the risk of users drawing incorrect conclusions from

the data, as they may not fully understand the vi-

sual mappings or the capabilities of the system (Pohl

et al., 2023; B

orner et al., 2016; Galesic and Garcia-

Retamero, 2010). Moreover, unlike traditional soft-

ware, where errors typically trigger clear feedback

or error messages, visualization systems often do not

provide such indicators when users misinterpret in-

formation. This makes it difﬁcult for users to recog-

nize when they have made a mistake (Lee et al., 2016;

Rezaie et al., 2024).

Thus, it is crucial to understand how to motivate

users to engage with visualization onboarding sys-

tems. The challenge lies in ﬁnding the “sweet spot”

between effectively onboarding users (nudging) and

avoiding annoyance. Research in the area of help

systems has shown that this is a difﬁcult balance to

strike (Mahmud et al., 2020; Andrade et al., 2009). If

the assistance is too unobtrusive, users may not no-

tice it; if, however, it is too intrusive, users may ﬁnd

it annoying and deactivate it completely. The case of

Microsoft’s Clippy is a prime example of how a well-

intentioned but poorly executed help system can lead

to widespread user frustration and a general aversion

to such systems (Chundury et al., 2023; Yalc¸ın, 2016).

To contribute to the challenge of ﬁnding the right

degree of intrusiveness in visualization onboarding

706

Pohl, M., Potzmann, B., Stoiber, C. and Aigner, W.

Inconspicuous Guides: A Qualitative Study to Understand the Degree of Intrusiveness in Visualization Onboarding.

DOI: 10.5220/0013135900003912

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 20th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2025) - Volume 1: GRAPP, HUCAPP

and IVAPP, pages 706-717

ISBN: 978-989-758-728-3; ISSN: 2184-4321

systems, we conducted a user inquiry using sketch-

ing (Roberts et al., 2015) with 65 Computer Science

master students at a university. Their feedback was

instrumental in generating innovative design ideas for

visualization onboarding solutions in complex visual-

ization systems that consider the degree of intrusive-

ness. Furthermore, we developed a questionnaire to

collect their opinions and attitudes toward visualiza-

tion onboarding.

Our investigation focuses on three key research

questions:

RQ1. What kind of attitudes do participants have

regarding the intrusiveness of visualization on-

boarding?

RQ2. What suggestions do participants have con-

cerning the intrusiveness of visualization on-

boarding?

RQ3. What kind of onboarding prototypes do they

construct to avoid intrusiveness? How do they

combine different features of onboarding?

By addressing these questions, this paper con-

tributes to the ongoing discourse on designing effec-

tive visualization onboarding systems and their role in

enhancing user interaction with complex data visual-

izations. In detail, the main contributions of the pa-

per are the following: (1) Discussion of related work

in the ﬁeld of help systems, visualization onboard-

ing, and guidance as related ﬁelds (in Section 2); (2)

Presentation of the results of an investigation with 65

participants aiming to ﬁnd a feasible visualization on-

boarding concept that is accepted by users in terms of

intrusiveness (Section 3). (3) Additionally, we present

examples of the participants’ resulting sketches (see

Section 3.4.3).

2 RELATED WORK

Within the landscape of Human-Computer Interaction

(HCI), research has intensively examined the chal-

lenges users face while seeking help with complex,

feature-rich software (Kiani et al., 2019). Visualiza-

tion onboarding (Stoiber, 2023; Dhanoa et al., 2022)

and guidance (P

erez-Messina et al., 2022; Ceneda

et al., 2017, 2019a; Sperrle et al., 2021) are integrated

as part of user assistance (Stoiber et al., 2022a) in

visual analytics (VA), helping users during different

phases of analysis by providing structured introduc-

tions and support for interpreting data (Stoiber et al.,

2022d, 2023a, 2021). Guidance provides support for

users as they solve analytical tasks (P

erez-Messina

et al., 2022). In the context of data visualization and

VA, a recommendation system, also known as a rec-

ommender system, refers to any system that proposes

content to a user based on a calculated correlation

between the content and the user’s interests (Pazzani

and Billsus, 2007; Zhou et al., 2023). Recommender

systems for VA can suggest relevant visualizations,

data ﬁlters, or analysis methods to users (Resnick

and Varian, 1997). Some recommender systems al-

ready use AI-based approaches (e.g., some forms of

content-based ﬁltering). The widespread adoption of

ChatGPT indicates a possible additional area of re-

search. The ﬁrst attempts for software tasks (Khurana

et al., 2024) and programming-related tasks (Xu et al.,

2022) have been made. To the best of our knowledge,

extensive research in the context of visualization on-

boarding, guidance, or help systems has not yet been

conducted in this area.

The synergy between these areas lies in their col-

lective goal to empower users to effectively engage

with complex systems, particularly in VA. In the fol-

lowing, we discuss related work in the ﬁeld of (1) help

systems, (2) visualization onboarding, and (3) guid-

ance in VA.

2.1 Help Systems

The design of help systems has been discussed at

some length in the scientiﬁc literature. Grayling

(2002), for example, provides an overview of ap-

proaches on how to design help systems. Accord-

ing to him, research indicates that users tend to ig-

nore help systems and prefer trial and error. He sug-

gests possible solutions to this problem, e.g., tooltips

or embedded help panels. In general, he states that

help should be context-speciﬁc, useful, obvious to

invoke, non-intrusive, and easily available. Vouligny

and Robert (2005) describe an approach for help sys-

tems based on situated action theory. They also point

out that users are often sceptical of help systems.

Technical documentations often provide descriptions

of a system’s functionalities, but they are not tailored

to assist users in their tasks and are, therefore, rarely

used. The authors also discuss the issue of intrusive

help systems. They argue that these systems are often

annoying because they rely on simplistic assumptions

about the users’ intentions. Silveira et al. (2001) pro-

pose a semiotic engineering approach. The emphasis

of this approach is on identifying the intents of the

users and the design of appropriate help messages.

The authors developed a taxonomy of users’ utter-

ances expressing speciﬁc needs for help (e.g., Where

is . . . ?, What’s this? What happened?). Our research

indicates that similar to research on help systems,

users enjoy ﬁnding out about the features of the sys-

Inconspicuous Guides: A Qualitative Study to Understand the Degree of Intrusiveness in Visualization Onboarding

707

tems themselves and prefer unobtrusive and context-

sensitive forms of assistance.

2.2 Visualization Onboarding

Visualization onboarding methods (Stoiber et al.,

2022d; Stoiber, 2023; Stoiber et al., 2023a; Dhanoa

et al., 2022) aim to support end users in compre-

hending data visualizations and taking full advan-

tage of the tools at hand. There are different types

of help-seeking behaviors: help initiated by the user

(pull) vs. the system (push) (Horvitz, 1999; Kiani

et al., 2019). Visualization onboarding can be exter-

nal (Stoiber et al., 2022c), separate from the tool, or

in-situ (Stoiber et al., 2023a), i.e., integrated within

it. In-situ systems use either the “pull” or “push” ap-

proach for providing help (Horvitz, 1999; Kiani et al.,

2019).

So far, most visualization onboarding systems

can be categorized as external learning environ-

ments (Stoiber et al., 2022c), detached from the core

visual analytics tools. For example, Firat et al. (2020)

developed an interactive pedagogical treemap appli-

cation for training. Additionally, Peng et al. (2022)

present results of a study to evaluate six parallel co-

ordinate literacy modules based on Bloom’s taxon-

omy (Bloom et al., 1956) using videos, tests, and

tasks. Both examples provide a platform to enhance

students’ visualization literacy.

The few papers where in-situ visualization on-

boarding is observed explain the features of the vi-

sualization tool by exploring both push and pull

onboarding assistance, with the main focus on the

“pull” help-seeking behavior. Yalc¸ın (2016) pre-

sented HelpIn, a system designed for the Keshif vi-

sualization tool that demonstrates some aspects of

“push” and “pull” models, providing contextual in-

situ help (Chundury et al., 2023) without disrupting

user tasks. They utilize the “push” model through no-

tiﬁcations to suggest relevant help topics on the ﬂy.

The notiﬁcations are designed in such a way that users

are not disrupted in their exploration tasks. Similarly,

IBM Cognos (Analytics, 2022) and Advizor (Solu-

tions, 2022) employ step-by-step tours with tooltips,

overlays, and textual descriptions, respectively, thus

aligning with the “pull” model (Horvitz, 1999). How-

ever, these examples primarily focus on feature expla-

nation.

Stoiber et al. developed four visualization on-

boarding methods for visualization tools: a step-by-

step guide, scrollytelling, video tutorials, and in-situ

scrollytelling (Stoiber et al., 2022d; Stoiber, 2023).

They conducted studies with MTurk workers and stu-

dents to assess the impact of these methods on user

performance and experience. They also examined dif-

ferent instruction styles (concrete vs. abstract) in a

separate study (Stoiber et al., 2022b). Building on

their ﬁndings (Stoiber et al., 2022d, 2021, 2022c,b),

they recently proposed nine design actions for ef-

fective visualization onboarding, emphasizing in-situ

methods and considering factors like user engagement

and accepted annoyance (Stoiber et al., 2023a).

Furthermore, the literature reveals a variety of vi-

sualization onboarding concepts focusing on explor-

ing different teaching methods and learning types in

the broader context of visualization literacy. Tana-

hashi et al. (2016) contrast top-down and bottom-up

approaches, while Kwon and Lee (2016) highlight

the effectiveness of active learning strategies in un-

derstanding complex visualizations like parallel coor-

dinate plots. Ruchikachorn and Mueller (2015) em-

phasize the power of learning by analogy in grasping

unfamiliar visualization methods.

In summary, the current landscape of visualization

onboarding systems predominantly consists of exter-

nal learning environments with limited examples of

in-situ integration. Moreover, the subtle balance be-

tween providing sufﬁcient assistance and causing an-

noyance – a critical aspect of the push model – re-

mains an underexplored area.

2.3 Guidance in Visual Analytics

The concept of guidance was initially introduced by

Schulz et al., serving as an umbrella term encompass-

ing various aspects such as “recommender systems”,

“user support”, and “assistance” within Visual Ana-

lytics (VA) (Schulz et al., 2013). Ceneda et al. fur-

ther reﬁned this deﬁnition, describing guidance as a

computer-aided method designed to actively bridge a

user’s knowledge gap during interactive visual analyt-

ics sessions (Ceneda et al., 2017).

Their work also discusses the importance of unob-

trusiveness in guidance in VA, emphasizing the need

for guidance methods that do not hinder the analytical

process (Ceneda et al., 2017). They propose a model

of guidance that gradually narrows the gap hindering

the effective continuation of data analysis, highlight-

ing the signiﬁcance of unobtrusive guidance (Ceneda

et al., 2019a).

Several factors inﬂuence the optimal timing for

guidance and help-seeking behavior. Firstly, the com-

plexity of the analytical task and, secondly, the user’s

skill level play a crucial role in determining the tim-

ing for guidance (P

erez-Messina et al., 2022). In this

regard, Ceneda et al. also explored the issue of the

proper timing of guidance by using facial recognition

software and a machine learning model trained to de-

IVAPP 2025 - 16th International Conference on Information Visualization Theory and Applications

708

tect when to guide according to changes in the user’s

facial expressions (Ceneda et al., 2021).

Summary. The takeaways of this body of research

show that successful visualization onboarding sys-

tems should provide assistance that is contextually

relevant, minimally intrusive, and adaptable to in-

dividual user preferences and skill levels (Grayling,

2002; Silveira et al., 2001; Stoiber et al., 2023a;

Ceneda et al., 2019b). The degree of intrusiveness

accepted by the user of visualization onboarding sys-

tems is still an open question, particularly when it

comes to the utilization of pull or push models (Chun-

dury et al., 2023), which set the stage for further in-

vestigations. Therefore, we conducted a study to ﬁnd

ideas and concepts of visualization onboarding that

consider nudging and intrusiveness (see Section 1).

We present our investigation in the following, dis-

cussing the study design, participants, and derived re-

sults in detail.

3 INVESTIGATION

3.1 Study Design & Procedure

Study Design. We conducted the study with a dual

objective: to understand participants’ subjective ex-

periences with visualization onboarding systems, and

to gather innovative ideas for designing effective on-

boarding solutions for complex visualization systems

(see research questions section 1).

We designed a questionnaire and a user inquiry

using sketching with our study participants to an-

swer our research questions. We wanted to under-

stand which visualization onboarding solutions (e.g.,

tutorials, pop-ups, tooltips) the participants favored,

found effective, or disliked and what their attitudes

concerning the trade-off between helpfulness and in-

trusiveness were. From previous research, we know

that systems might be helpful but intrusive when they

offer unsolicited help messages indicating features

that have been overlooked by users. We also in-

cluded questions on how the participants thought pro-

grams might detect when users need assistance and

how users can realize their need for help. The ques-

tionnaire consisted of the following questions:

• What kind of onboarding system would you per-

sonally be most willing to use?

• What kind of systems might be helpful but unus-

able or intrusive?

• What level of annoyance is necessary?

• How could users realize that they need help?

• How could programs detect that the users need

help?

A practical component of the study, the sketching

sessions, involved participants sketching prototypes

for onboarding solutions tailored to a complex visu-

alization system featuring various visualization types

and ﬁltering options. This task was intended to re-

veal the range of approaches participants considered

effective for encouraging the use of onboarding sys-

tems. The goal of developing the prototype was to

determine whether they could devise concrete solu-

tions for the visualization onboarding process they ap-

preciated, and what kind of nudging they suggested.

Possible existing approaches to this problem were not

discussed with the participants so as not to inﬂuence

them in the development of creative solutions.

Procedure. The participants were given a document

consisting of the following parts: 1. a short descrip-

tion of the investigation, 2. an informed consent form,

3. a questionnaire, and 4. a brief description of the

task of developing the prototype. Before starting to

work, one of the co-authors met the participants and

gave them an introduction of about one hour regard-

ing the goals of the investigation and the study pro-

cess. The participants had eight weeks to complete

both tasks. They ﬁlled in the questionnaires individ-

ually but were allowed to develop the prototypes in

pairs because we assumed that if they could discuss

their solution with a colleague, they would reﬂect on

it more and create more creative solutions. Nine pairs

of participants cooperated to develop the prototype.

The other 47 participants developed the prototypes on

their own. Participants were required to create a pro-

totype consisting of a textual description and drafts of

screenshots.

3.2 Participants

As detailed in Table 1, we recruited 65 students (av-

erage age ≈ 25.7 years) from an international mas-

ter’s program in computer science at Vienna Univer-

sity of Technology, conducting the study within a lec-

ture setting (16 female, 49 male, the students came

from several different European countries). Upon ex-

amining their self-reported skills, it became clear that

most participants had a strong foundation in visualiza-

tion design, with about half also indicating experience

in UX design. In contrast, very few had experience

in designing help or support systems. The partici-

pants were master’s students in data science; many al-

ready worked part-time as computer scientists in that

Inconspicuous Guides: A Qualitative Study to Understand the Degree of Intrusiveness in Visualization Onboarding

709

Table 1: Distribution of proﬁciency and education levels. This table presents the 65 participants’ self-reported proﬁciency

in three areas: visualization design, UX design, and help/support system design.

Proﬁciency Viz Design Help Design UX Design Education n

Beginner 13 56 31 Bachelor’s 57

Intermediate 30 7 28 Master’s 5

Advanced 21 1 4 PhD 1

Expert 0 0 1 Undergraduate 1

area. These students were not only potential users

but also future developers of systems administrating

large datasets. As such, they had already developed

their ﬁrst ideas about designing such systems during

their studies. They might be called semi-experts, and

their ideas about motivating users to study IT systems

representing large datasets in detail are undoubtedly

relevant.

3.3 Data Analysis

We conducted a qualitative and exploratory investi-

gation to get an overview of the issues relevant to

the research question. We chose a qualitative con-

tent analysis approach (Schreier, 2012) to examine

the participants’ responses to the questionnaire and

the prototypes and ideas they generated. We were

able to break down the data into manageable, discrete

pieces of information by identifying recurring pat-

terns and themes, especially the kinds of visualization

onboarding solutions that the participants mentioned.

Through iterative rounds of reﬁnement, we developed

a nuanced understanding of the participants’ perspec-

tives and experiences as well as their preferences and

expectations. Two co-authors met repeatedly to dis-

cuss the data analysis, compare the results, and re-

solve diverging interpretations.

We developed a set of codes distinguishing be-

tween different features of visualization onboarding

systems. There are “push features” (e.g., pop-up

ﬁelds, compulsory tutorials), “pull features” (e.g., op-

tional tutorials, documentation), and features in be-

tween that can be designed either as “push” or as “pull

feature.” Tooltips, for example, can be interpreted as

a “pull feature”, but when they are opened because

a user moves the mouse unintentionally over the fea-

ture, it can be seen as “push feature”. Chatbots or in-

telligent recommender systems can be designed both

ways. They either give advice only when the user ac-

tivates them, or they are always active.

The codes were partly based on previous re-

search (Stoiber et al., 2022d) and partly derived

from the analyzed material. The following set of

codes was used for the analysis (see also Figure 1):

ML/chatbots, tooltips, video, tutorial, example, gami-

ﬁcation, help button, interactive walkthrough, pop-up,

together, doc/full search, test, compulsory tutorial, fo-

rum. Participants mainly suggested using either ma-

chine learning or chatbots (code ML/chatbots), but

the code encompasses any kind of AI-based support.

Some participants suggested offering practical exam-

ples (sometimes with solutions) to assist users. Gam-

iﬁcation was considered to motivate users more to use

visualization onboarding than simple tutorials. An in-

teractive walkthrough is supposed to be an interactive

form of tutorial. Pop-ups are always generated by the

system and are therefore “push features”. Doc/Full

search is meant to be documentation with full-text

search capabilities. Tests are entry tests to assess

users’ knowledge. A forum was sometimes suggested

so users could exchange experiences and help each

other. Some participants explicitly mentioned that a

speciﬁc subset of these features should be combined

to create an appropriate visualization onboarding sys-

tem (code: together).

3.4 Results

In this section, we present the results of this investi-

gation, divided into the questionnaire and prototype

results based on the sketching sections.

3.4.1 Results of the Questionnaire

The goal of the questionnaire was to clarify the per-

sonal opinions of the participants concerning nudging

and acceptable annoyance in onboarding systems.

Nudging Preferences. The ﬁrst question pertained

to the participents’ personal preferences concerning

the nudging behavior of visualization onboarding sys-

tems. Participants primarily chose in-situ tooltips as

the preferred form of nudging (36 participants out of

65). Tooltips are a very subtle and unobtrusive form

of nudging, which makes them seem very attractive.

Participants also appreciated onboarding systems that

include AI/ML elements (21 out of 65). Tutorials (14

out of 65) and learning by concrete and practical ex-

amples (13 out of 65) were mentioned as unobtru-

sive concepts. Eight participants explicitly said they

preferred to explore independently and use assistance

only when completely stuck.

IVAPP 2025 - 16th International Conference on Information Visualization Theory and Applications

710

Perceptions of Intrusiveness vs. Helpfulness. We

also asked participants about visualization onboard-

ing features that might be intrusive but still helpful.

The answers to this question were more general. 22

participants mentioned compulsory tutorials as help-

ful but intrusive, and 31 said that pop-ups were help-

ful. Participants said visualization onboarding should

not be nudging users constantly (e.g., by pop-ups that

appear frequently and are difﬁcult to eliminate). Visu-

alization onboarding should be manageable in length.

Additionally, it should not present unnecessary or un-

wanted information.

Degree of Intrusiveness. In a third question, we

asked about the necessary level of annoyance or intru-

siveness that would be acceptable for the participants.

In this case, the answers were sometimes very general

and provided no clear indication of speciﬁc features

that would or would not be acceptable. Those partic-

ipants who gave more precise answers said that short

and concise tutorials at the beginning would be ac-

ceptable (9 participants). Another possibility was to

use AI-based recommendations (9 participants) and,

surprisingly, pop-up messages (8 participants). In this

context, AI-based recommendation summarizes any

visualization onboarding system that uses machine

learning, for example, to track and react to user be-

haviors or AI concepts such as “virtual” assistants

to answer questions similar to ChatGPT. Participants

also mentioned that the amount of intrusiveness de-

pends on the users’ previous experience (10 partici-

pants) and the complexity of the tool (5 participants).

Detection of Help-Seeking Behavior. We also

asked the participants how users could ﬁnd out that

they needed help. 35 out of 65 participants said that

users notice that they need help when they cannot ﬁn-

ish their task, are confused, and show speciﬁc interac-

tion patterns (e.g., doing the same actions repeatedly

without result). 15 participants answered that users

should know they need help when receiving repeated

error messages from the system. Only 10 participants

mentioned that users should know they need help be-

cause of a general understanding of visualizations or

a test before using the visualization. The last ques-

tion was about the system’s ability to detect whether

users needed help. Almost all participants (64 out of

65) said the visualization system should monitor the

users’ activities and detect inappropriate behavior.

Lessons Learned. The study revealed a preference

for subtle in-situ nudging techniques in visualization

onboarding (RQ1), with a general skepticism towards

more intrusive methods. The participants showed an

Figure 1: Distribution of mentioned onboarding types (pro-

totypes).

interest in AI-based assistance but also had high ex-

pectations for system capabilities in detecting and re-

sponding to user needs. The participants said they

found intrusive features acceptable for inexperienced

users and complex visualizations (RQ2).

3.4.2 Results Prototype

Participants were asked to generate ideas or very

small prototypes for visualization onboarding consid-

ering nudging and intrusiveness (see Figure 1). Par-

ticipants were generally very reluctant to force users

to seek assistance. They tried very hard to ﬁnd so-

lutions that could, on the one hand, motivate users

to learn about the visualization but that, on the other

hand, did not interfere too much with the users’ work-

ﬂow.

Interestingly, the participants most frequently

mentioned any kind of features supported by artiﬁcial

intelligence (code ML/Chatbots) in a very broad sense

(37 out of 65 participants). Sometimes, we got the

impression that the participants had unrealistic ideas

about the capabilities of AI. Some suggested a kind

of personal assistant who unobtrusively looked over

their shoulder and reminded them cautiously and re-

strainedly about possible better solutions. Most had

at least some sort of chatbot in mind that would an-

swer questions. Additionally, it should be mentioned

that these AI-based features were usually not the main

component of the small prototype that the participants

suggested. Rather, they were additional features ac-

companying other onboarding methods.

Another very frequently mentioned feature was

tooltips (34 out of 65 participants) (see Fig. 1). In con-

trast to pop-ups, tooltips are unobtrusive and only ap-

pear upon mouse-over. Tooltips are a borderline case

as far as the activity of the system is concerned. They

are relatively unobtrusive but can appear without the

Inconspicuous Guides: A Qualitative Study to Understand the Degree of Intrusiveness in Visualization Onboarding

711

user asking for help. In addition, the information of-

fered by tooltips is usually very short. It was often

mentioned throughout the investigation that help texts

should be short and concise and offer necessary infor-

mation only. Tooltips meet these requirements, which

makes them very popular. Other features that were

also mentioned quite often were videos (21 out of 65),

the use of concrete and practical examples (19 out of

65), optional tutorials (19 out of 65), and gamiﬁcation

(18 out of 65). Gamiﬁcation was seen as a possibility

to motivate users to use the visualization onboarding

system by giving them rewards or by using features

that increase enjoyment and a positive mood so that it

was not seen as a burden. Short initial tutorials, espe-

cially for inexperienced users, were seen as an accept-

able possibility for visualization onboarding. Some

participants also mentioned MS Word’s spell-check

as a good example of an unobtrusive way of showing

users how to correct errors.

Compulsory tutorials (5 out of 65), (ability) tests

at the beginning (5 out of 65), extensive documen-

tation with full-text search capabilities (6 out of 65),

and pop-ups (9 out of 65) were features of visualiza-

tion onboarding systems that were hardly mentioned

at all. Compulsory tutorials and pop-ups in particular

generated very negative comments among the partic-

ipants. Communication in a user forum or interactive

walkthroughs was not seen as attractive, either. Some

users intended to use several features combined (code:

together).

Lessons Learned. Summarizing the results of the

qualitative content analysis on the prototypes for vi-

sualization onboarding showed that tooltips, AI fea-

tures, and interactive methods such as videos and

practical examples (tutorials) were favored, as shown

in Figure 1. Conversely, features perceived as forc-

ing information onto the user, such as compulsory

tutorials and repetitive pop-ups, were regarded unfa-

vorably (RQ3). The results from the development of

prototypes aligned closely with the questionnaire re-

sponses.

3.4.3 Examples of Prototypes

The results of the overall analysis of the prototypes

developed by the participants conform to the results of

the analysis of the questionnaires. Nevertheless, the

prototypes, investigated individually, indicate that the

participants came up with a wide variety of solutions.

The four examples, three of which are illustrated

in Figure 2, Figure 3, and Figure 4, are discussed in

the following and show this nicely. They were chosen

for their clarity of presentation and lucid structure.

Example 1: Self-Test. First-time users of the visu-

alization receive a self-test (see Figure 2). The system

provides feedback about the user’s answers. If they

make a mistake, they are offered a demo video. These

videos should be short and to the point. While the

user is working, the system monitors their activities.

When the system detects an error, the user is asked

whether they need help. This is intrusive but helps the

users to solve their tasks quickly. Another possibility

is tooltips (called “hover info buttons” here), which

are less intrusive. Extensive documentation should

also be available when needed.

Example 2: e-Learning System. This solution of-

fers an onboarding system that is similar to an e-

learning system but aims to teach the speciﬁc infor-

mation that users need at the time they need it (see

Figure 3). The system tries to engage users by moti-

vating them to solve examples on their own. A sidebar

is added to the existing visualization on the left-hand

side of the screen. This sidebar contains challenges

that are ordered according to their difﬁculty. Each has

a clearly deﬁned goal and several steps to reach it.

Multiple-choice questions are shown if a user selects

a challenge (see Figure 3). Each question contains the

next step required to reach the solution. The system

explains the correct solution when the user clicks on

the wrong solution. In addition, a scoring system can

be introduced, adding an element of gamiﬁcation.

Example 3: Exemplary Workﬂows and Use Cases.

The system proposed by this participant consists of

various elements. It starts out with a short test at

the beginning to assess the user’s level of expertise.

This helps to avoid presenting hints and onboarding

tools to users who do not need them. The system

will also provide information about typical workﬂows

with numbers and explanatory text (see Figure 4).

This helps to inform the user on the correct sequence

of steps to perform a task. In addition, there will be a

concrete use case informing the user about interactive

features in particular. If the user does not fulﬁll the

task, feedback is provided about how to solve it suc-

cessfully. The onboarding system will also contain

videos.

Example 4: Chatbot. This prototype is unique be-

cause a chatbot forms the main element of the on-

boarding system. In contrast, most other prototypes

only use chatbots or AI-based features as assistive

features. This chatbot is supposed to accomplish sev-

eral different tasks. It answers questions about how

to complete speciﬁc tasks. In addition, it provides in-

formation about the data (e.g., the mean values of the

IVAPP 2025 - 16th International Conference on Information Visualization Theory and Applications

712

Figure 2: The example of this prototype is highly structured. Several features provide help (self-test game, automatic error

detection, videos and documentation, tooltips). These different features are related, and the system can move from one feature

to another when needed.

Figure 3: This ﬁgure shows a screenshot of example 2. On the left-hand side are the challenges that users have to solve. Right

beside it are the multiple-choice questions to be answered by the users.

Figure 4: Description of a workﬂow with numbers and tooltips.

data) and the appropriate interpretation of the data.

One of its key features is its ability to become ac-

tive when it detects a user error (e.g., when the user

employs many different tools without ﬁnishing the

task), providing reassurance that users are always sup-

ported. In this case, the user is asked whether they

need help. In addition to the chatbot, the system also

contains an optional tutorial video and tooltips.

Inconspicuous Guides: A Qualitative Study to Understand the Degree of Intrusiveness in Visualization Onboarding

713

Lessons Learned: The overall analysis of the pro-

totypes indicates some clear tendencies among par-

ticipants. However, the analysis of the individual

prototypes shows that the participants’ solutions vary

considerably. The participants generally do not stick

to one idea but distinctly combine different features.

They typically combine more and less intrusive fea-

tures, sometimes targeted at more or less experienced

users. Machine learning or AI features (especially

chatbots) are usually added as an auxiliary feature.

Gamiﬁcation is often used with the explicit goal of

motivating users to interact with the onboarding sys-

tem and overcoming users’ annoyance with intrusive

onboarding. Step-wise introductions or practical ex-

amples are also seen as a possibility to engage users.

Again, the idea is to overcome the annoyance of the

users with more intrusive forms of onboarding.

4 DISCUSSION & FUTURE

WORK

This investigation aimed to identify what kind of vi-

sualization onboarding users of visualizations found

annoying and intrusive, and what would be more ac-

ceptable. This research question is based on previ-

ous research (Pohl et al., 2023) indicating that users

of visualizations are sometimes not aware that they

need help and, therefore, come to erroneous con-

clusions about the data presented to them. While

there is some research on users struggling with vi-

sualizations (B

orner et al., 2016; Galesic and Garcia-

Retamero, 2010; Rezaie et al., 2024; Lee et al., 2017),

this speciﬁc issue has, to the best of our knowledge,

not yet been addressed in research. In the next para-

graph, we present lessons learned based on our re-

search.

User Skepticism Towards Active Assistance. The

participants generally indicated that they were pretty

skeptical about visualization onboarding, especially

systems actively trying to assist them. Many partic-

ipants mentioned that they preferred to ﬁnd out how

the system worked themselves, following a trial-and-

error strategy (cf. Mahmud et al. (2020); Andrade

et al. (2009)).

Preference for Subtle, Contextualized Nudging.

Participants preferred unobtrusive onboarding meth-

ods such as tooltips, concise tutorials or videos, and

small pop-ups in unused screen areas. These meth-

ods were seen as minimally disruptive and helpful.

Some users also mentioned MS Word’s spell-check as

a positive example of a system’s unobtrusive but ac-

tive help. They argued that something similar should

be developed for visualizations (e.g., systems that in-

dicate the next necessary step in the interaction pro-

cess through subtle highlighting). Participants argued

that visualization onboarding should preferably be in

situ (integrated into the system) so that their workﬂow

would not be interrupted.

Pop-ups that appeared repeatedly, covered large

parts of the screen, or were difﬁcult to close were

seen as highly annoying and intrusive. Participants

also criticized long and complex tutorials with unnec-

essary information. Nudging should be adapted to the

context, they said.

Borderline Cases. While participants preferred less

intrusive solutions, they argued that in some cases, in-

trusiveness might be annoying but helpful. It depends

on the situation whether intrusiveness is acceptable.

Participants found it appropriate that the amount of

nudging for inexperienced users and during interac-

tion with complex visualizations could be higher. In

this case, compulsory tutorials might be helpful. Pop-

ups (e.g., with hints of the day) appearing in unused

areas of the screen were deemed acceptable provided

that this does not occur frequently, and that the pop-

ups disappear by themselves. In all these cases, the

text should be short. Some users would also ﬁnd it

acceptable for a chatbot to give advice, thus interrupt-

ing the workﬂow.

This ﬁnding suggests that onboarding systems

should be adaptive, offering more guidance to novices

and less to experienced users, thus tailoring the on-

boarding experience to the user’s level of expertise.

Potential of AI-driven Visualization Onboarding

Methods. Participants were open to the idea of AI-

driven onboarding systems that could monitor user

activity and provide context-sensitive hints or alter-

native solutions. In detail, they imagined that an AI

system might monitor their activities and offer hints

or alternative ways of actively solving a problem to

assist users. Participants mentioned wanting to ask

the AI questions and discuss solutions with this sys-

tem. AI-driven onboarding is a more acceptable form

of nudging than many others. To what extent this can

be realized with existing generative AI models, like

ChatGPT, is an open question. First attempts have

been made to solve this problem (Joshi et al., 2024;

Zhao et al., 2024), but future work will show whether

such ideas are realistic. This is certainly a promising

area for future research.

IVAPP 2025 - 16th International Conference on Information Visualization Theory and Applications

714

Misalignment Between Users’ Expectations and

Visualization Realities. Most participants appar-

ently use a model from more conventional systems

to reason about problems with visualizations. This

model assumes one clear and identiﬁable path to the

solution to a problem. The system can detect any de-

viation from this path and either show an error mes-

sage or an inappropriate result. As mentioned above,

this does not always apply to visualizations. Inter-

action with visualizations is an exploratory process

(Battle and Heer, 2019), and often no clear path to

the solution can be identiﬁed. Visualization systems,

therefore, are sometimes unable to identify appropri-

ate or inappropriate user behavior or assess the valid-

ity of the insights generated by them. Many partici-

pants expected the system to always provide feedback

when errors occured. Visualization onboarding sys-

tems should also raise awareness of this problem.

Potential of Combining Different Features of On-

boarding. In developing their prototypes, partici-

pants systematically combined various features of on-

boarding systems. They usually combined more and

less intrusive features to provide different possibilities

for user onboarding. Using gamiﬁcation and practical

examples can counterbalance the negative effects of

more intrusive onboarding.

4.1 Limitation

While our study was limited by the demography of

the participants, who were primarily data science stu-

dents, it is important to note that these students are

the future data scientists, many of whom also work

part-time in the ﬁeld. We also conducted an investiga-

tion with professionals from a software development

company and found that they were also very skepti-

cal about more intrusive visualization onboarding and

preferred subtle nudging (tooltips and very short tuto-

rials) (Potzmann et al., 2023). This aligns with other

research ﬁndings (Stoiber et al., 2023b; Baur and

Stephaner, 2018). Further research should address

this issue more systematically, offering a wealth of

potential for future exploration.

Our approach mainly addresses the attitudes of

potential designers/users and a sketch-based approach

similar to constructive visualization by Huron et al.

(2014). An alternative approach would be to ask study

participants to solve tasks and use observation to de-

termine whether users apply onboarding, and what

speciﬁc features they use. However, we would like

to point out that this is a very challenging approach.

Researchers need detailed knowledge about the users’

level of experience to construct appropriate study ma-

terial. If this material is too easy so that participants

do not need onboarding, no appropriate results can

be gained. In addition, tasks have to be developed

speciﬁcally so that users are compelled to use various

onboarding features. Furthermore, to get detailed in-

sights about the users’ cognitive strategies, thinking

aloud or conducting interviews would be necessary to

explain why users adopt certain features of the on-

boarding system, or not. Nevertheless, the results of

such an investigation would be interesting.

5 CONCLUSION

We conducted an empirical study on the acceptance

of more intrusive forms of visualization onboarding.

The participants of our study accepted only very sub-

tle forms of intrusion. It seems that AI-based ap-

proaches might be more acceptable than traditional

forms of assistance. In contrast to existing visualiza-

tion onboarding solutions in visualization, we suggest

visualization onboarding integrated into the system.

This also conforms to research on help systems.

ACKNOWLEDGEMENTS

This work was funded by the BMK under the ICT of

the Future program (#874018), the Austrian Science

Fund (10.55776/I5622), as well as by the Austrian

Federal Ministry of Labour and Economy, the Na-

tional Foundation for Research, Technology and De-

velopment, and the Christian Doppler Research Asso-

ciation. For the purpose of open access, the author has

applied a CC BY public copyright license to any Au-

thor Accepted Manuscript version arising from this

submission.

REFERENCES

Analytics, I. C. (2022). IBM Cognos Analytics. https://

www.ibm.com/at-de/products/cognos-analytics. Ac-

cessed: 2022-01-31.

Andrade, O. D., Bean, N., and Novick, D. G. (2009). The

macro-structure of use of help. In Proceedings of

the 27th ACM International Conference on Design of

Communication, SIGDOC ’09, page 143–150. ACM.

Battle, L. and Heer, J. (2019). Characterizing exploratory

visual analysis: A literature review and evaluation of

analytic provenance in tableau. In Computer graph-

ics forum, volume 38, pages 145–159. Wiley Online

Library.

Inconspicuous Guides: A Qualitative Study to Understand the Degree of Intrusiveness in Visualization Onboarding

715

Baur, D. and Stephaner, M. (2018). Everything except the

chart. http://webvis-do.minik.us/. Accessed: 2023-

01-11.

Bloom, B., Engelhart, M., Furst, E., W.Hill, and Krath-

wohl, D. (1956). Taxonomy of Educational Objec-

tives, Handbook 1: Cognitive Domain. Longman,

New York; New York; London, 2nd edition edition

edition.

orner, K., Maltese, A., Balliet, R. N., and Heimlich, J.

(2016). Investigating aspects of data visualization

literacy using 20 information visualizations and 273

science museum visitors. Information Visualization,

15(3):198–213.

Ceneda, D., Arleo, A., Gschwandtner, T., and Miksch, S.

(2021). Show me your face: towards an automated

method to provide timely guidance in visual analyt-

ics. IEEE Transactions on Visualization and Com-

puter Graphics, 28(12):4570–4581.

Ceneda, D., Gschwandtner, T., May, T., Miksch, S., Schulz,

H.-J., Streit, M., and Tominski, C. (2017). Character-

izing guidance in visual analytics. IEEE Transactions

on Visualization and Computer Graphics, 23:111–

120.

Ceneda, D., Gschwandtner, T., and Miksch, S. (2019a). A

Review of Guidance Approaches in Visual Data Anal-

ysis: A Multifocal Perspective. Computer Graphics

Forum, 38(3):861–879.

Ceneda, D., Gschwandtner, T., and Miksch, S. (2019b). You

get by with a little help: The effects of variable guid-

ance degrees on performance and mental state. Visual

Informatics, 3(4):177–191.

Chundury, P., Yalc¸in, M. A., Crabtree, J., Mahurkar, A.,

Shulman, L. M., and Elmqvist, N. (2023). Contex-

tual in situ help for visual data interfaces. Information

Visualization, 22(1):69–84.

Dhanoa, V., Walchshofer, C., Hinterreiter, A., Stitz, H.,

Groeller, E., and Streit, M. (2022). A process model

for dashboard onboarding. Computer Graphics Forum

(EuroVis ’22), 41(3):501–513.

Firat, E. E., Denisova, A., and Laramee, R. S. (2020).

Treemap Literacy: A Classroom-Based Investigation.

In Romero, M. and Sousa Santos, B., editors, Euro-

graphics 2020 - Education Papers, pages 29–38, n.p.

The Eurographics Association.

Galesic, M. and Garcia-Retamero, R. (2010). Graph Liter-

acy: A Cross-Cultural Comparison. Medical Decision

Making, 31(3):444–457. Publisher: SAGE Publica-

tionsSage CA: Los Angeles, CA.

Grayling, T. (2002). If we build it, will they come? a usabil-

ity test of two browser-based embedded help systems.

Technical communication, 49(2):193–209.

Horvitz, E. (1999). Principles of mixed-initiative user inter-

faces. In Proceedings of the SIGCHI Conference on

Human Factors in Computing Systems, CHI ’99, page

159–166, New York, NY, USA. ACM.

Huron, S., Jansen, Y., and Carpendale, S. (2014). Con-

structing visual representations: Investigating the use

of tangible tokens. IEEE transactions on visualization

and computer graphics, 20(12):2102–2111.

Joshi, A., Srinivas, C., Firat, E. E., and Laramee, R. S.

(2024). Evaluating the recommendations of llms to

teach a visualization technique using bloom’s taxon-

omy. Technical report, University of Nottingham.

Khurana, A., Subramonyam, H., and Chilana, P. K. (2024).

Why and when llm-based assistants can go wrong: In-

vestigating the effectiveness of prompt-based interac-

tions for software help-seeking. In Proceedings of IUI,

IUI ’24, page 288–303, New York, NY, USA. ACM.

Kiani, K., Cui, G., Bunt, A., McGrenere, J., and Chilana,

P. K. (2019). Beyond ”one-size-ﬁts-all”: Understand-

ing the diversity in how software newcomers discover

and make use of help resources. In Proceedings of

the 2019 CHI Conference on Human Factors in Com-

puting Systems, CHI ’19, page 1–14, New York, NY,

USA. Association for Computing Machinery.

Kwon, B. C. and Lee, B. (2016). A comparative evalua-

tion on online learning approaches using parallel co-

ordinate visualization. In Proceedings of the SIGCHI

Conference on Human Factors in Computing Systems,

CHI ’16, pages 993–997, New York, NY, USA. ACM

Press.

Lee, S., Kim, S.-H., Hung, Y.-H., Lam, H., Kang, Y.-A., and

Yi, J. S. (2016). How do people make sense of unfa-

miliar visualizations?: A grounded model of novice’s

information visualization sensemaking. IEEE Trans-

actions on Visualization and Computer Graphics,

22(1):499–508.

Lee, S., Kim, S.-H., and Kwon, B. C. (2017). VLAT:

Development of a visualization literacy assessment

test. IEEE Transactions on Visualization and Com-

puter Graphics, 23(1):551–560.

Mahmud, S., Alvina, J., Chilana, P. K., Bunt, A., and Mc-

Grenere, J. (2020). Learning through exploration:

How children, adults, and older adults interact with

a new feature-rich application. In Proceedings of the

2020 CHI Conference on Human Factors in Comput-

ing Systems, CHI ’20, page 1–14, New York, NY,

USA. Association for Computing Machinery.

Pazzani, M. J. and Billsus, D. (2007). Content-Based

Recommendation Systems, pages 325–341. Springer

Berlin Heidelberg, Berlin, Heidelberg.

Peng, I., Firat, E. E., Laramee, R. S., and Joshi, A. (2022).

Evaluating Bloom’s Taxonomy-based Learning Mod-

ules for Parallel Coordinates Literacy. In Bourdin, J.-

J. and Paquette, E., editors, Eurographics 2022 - Edu-

cation Papers, pages 21–29, Eindhoven, The Nether-

lands. The Eurographics Association.

Pohl, M., Potzmann, B., Stoiber, C., and Aigner, W.

(2023). Visualization Onboarding Designed by Users:

an Empirical Study. In International Conference on

Cognition and Exploratory Learning in Digital Age

(CELDA), pages 3–10, Funchal, Madeira, Portugal.

IADIS Press.

Potzmann, B., Pohl, M., Moitzi, D., and Girardi, D. (2023).

Using design actions for a heuristic evaluation of hid-

den. Technical report, Technische Universit

at Wien.

erez-Messina, I., Ceneda, D., El-Assady, M., Miksch, S.,

and Sperrle, F. (2022). A Typology of Guidance

Tasks in Mixed-Initiative Visual Analytics Environ-

ments. Computer Graphics Forum, 41(3):465–476.

IVAPP 2025 - 16th International Conference on Information Visualization Theory and Applications

716

Resnick, P. and Varian, H. R. (1997). Recommender sys-

tems. Commun. ACM, 40(3):56–58.

Rezaie, M., Tory, M., and Carpendale, S. (2024). Struggles

and strategies in understanding information visualiza-

tions. IEEE Transactions on Visualization and Com-

puter Graphics, 30(6):3035–3048.

Roberts, J. C., Headleand, C., and Ritsos, P. D. (2015).

Sketching designs using the ﬁve design-sheet method-

ology. IEEE transactions on visualization and com-

puter graphics, 22(1):419–428.

Ruchikachorn, P. and Mueller, K. (2015). Learning

visualizations by analogy: Promoting visual liter-

acy through visualization morphing. IEEE Trans-

actions on Visualization and Computer Graphics,

21(9):1028–1044.

Schreier, M. (2012). Qualitative content analysis in prac-

tice. Sage publications, Newbury Park, California.

Schulz, H.-J., Streit, M., May, T., and Tominski, C. (2013).

Towards a characterization of guidance in visualiza-

tion. In Poster at IEEE Conference on Information

Visualization (InfoVis), volume 2.

Silveira, M. S., de Souza, C. S., and Barbosa, S. D. (2001).

Semiotic engineering contributions for designing on-

line help systems. In Proceedings of the 19th an-

nual international conference on Computer documen-

tation, pages 31–38.

Solutions, A. (2022). Advizor Solutions. https://www.

advizorsolutions.com/platform/. Accessed: 2022-01-

31.

Sperrle, F., Jeitler, A., Bernard, J., Keim, D., and El-Assady,

M. (2021). Co-adaptive visual data analysis and guid-

ance processes. Computers & Graphics, 100:93–105.

Stoiber, C. (2023). Visualization Onboarding - Supporting

Users in Understanding Unfamiliar Visual Represen-

tations. PhD thesis, TU Wien, Wien,

Osterreich.

Stoiber, C., Ceneda, D., Wagner, M., Schetinger, V.,

Gschwandtner, T., Streit, M., Miksch, S., and Aigner,

W. (2022a). Perspectives of visualization onboarding

and guidance in VA. Visual Informatics, 6(1):68–83.

Stoiber, C., Grassinger, F., and Aigner, W. (2022b). Ab-

stract and concrete materials: What to use for visual-

ization onboarding for a treemap visualization? In

Proceedings of VINCI, VINCI ’22, New York, NY,

USA. ACM.

Stoiber, C., Pohl, M., and Aigner, W. (2023a). Design

Actions for the Design of Visualization Onboarding

Methods. In IEEE VIS Workshop on Visualization

Education, Literacy, and Activities, Melbourne, Aus-

tralia. IEEE.

Stoiber, C., Radkohl, S., Grassinger, F., Moitzi, D., Stitz,

H., Goldgruber, E., Girardi, D., and Aigner, W.

(2023b). Authoring tool for data journalists integrat-

ing self-explanatory visualization onboarding concept

for a treemap visualization. In Proceedings of the 15th

Biannual Conference of the Italian SIGCHI Chapter,

CHItaly ’23, New York, NY, USA. Association for

Computing Machinery.

Stoiber, C., Wagner, M., Grassinger, F., Pohl, M., Stitz,

H., Streit, M., Potzmann, B., and Aigner, W. (2022c).

Visualization Onboarding Grounded in Educational

Theories. In Visualization Psychology. Springer Na-

ture, n.p.

Stoiber, C., Walchshofer, C., Grassinger, F., Stitz, H., Streit,

M., and Aigner, W. (2021). Design and compara-

tive evaluation of visualization onboarding methods.

In Proceedings of the 14th International Symposium

on Visual Information Communication and Interac-

tion, VINCI ’21, New York, NY, USA. Association

for Computing Machinery.

Stoiber, C., Walchshofer, C., Pohl, M., Potzmann, B.,

Grassinger, F., Stitz, H., Streit, M., and Aigner, W.

(2022d). Comparative evaluations of visualization on-

boarding methods. Visual Informatics, 6(4):34–50.

Tanahashi, Y., Leaf, N., and Ma, K. (2016). A study

on designing effective introductory materials for in-

formation visualization. Computer Graphics Forum,

35(7):117–126.

Vouligny, L. and Robert, J.-M. (2005). Online help system

design based on the situated action theory. In Pro-

ceedings of the 2005 Latin American conference on

Human-computer interaction, pages 64–75.

Xu, F. F., Vasilescu, B., and Neubig, G. (2022). In-ide code

generation from natural language: Promise and chal-

lenges. ACM Trans. Softw. Eng. Methodol., 31(2).

Yalc¸ın, M. A. (2016). A Systematic and Minimalist Ap-

proach to Lower Barriers in Visual Data Exploration.

Dissertation, University of Maryland, College Park.

Zhao, Y., Zhang, Y., Zhang, Y., Zhao, X., Wang, J., Shao,

Z., Turkay, C., and Chen, S. (2024). Leva: Using large

language models to enhance visual analytics. IEEE

Transactions on Visualization and Computer Graph-

ics, pages 1–17.

Zhou, Z., Wang, W., Guo, M., Wang, Y., and Gotz, D.

(2023). A Design Space for Surfacing Content Rec-

ommendations in Visual Analytic Platforms. IEEE

Transactions on Visualization and Computer Graph-

ics, 29(1):84–94.

Inconspicuous Guides: A Qualitative Study to Understand the Degree of Intrusiveness in Visualization Onboarding

717