Dynamic Reading Comprehension Visualization in Digital Course Texts

Dennis Menze

, Niels Seidel

and Slavisa Radovic

CATALPA – Center of Advanced Technology for Assisted Learning and Predictive Analytics,

Fernuniversit

at Hagen, Universit

atsstr. 27, 58097 Hagen, Germany

{dennis.menze, niels.seidel, slavisa.radovic}@fernuni-hagen.de

Keywords:

Reading Comprehension, Reading Analytics, Learning Analytics.

Abstract:

Both reading textbooks and answering quizzes lead to better recall of learning content and better learning

outcomes, especially when both forms are combined interactively. Nevertheless, existing solutions in learn-

ing management systems usually offer reading and quizzes separately. This work aims to improve this by

measuring and visualizing students’ reading progress and reading comprehension based on their answers to

automatically displayed questions about the text sections they have just read. In this paper, we present an

adaptive system for supporting reading comprehension. A randomized trial with 57 students showed high

engagement with both the reading material and embedded questions, demonstrating the technical feasibility of

integrating comprehension support into digital course texts. Finally, possible extensions to improve adaptivity,

interventions, automation, and measurement of reading comprehension are discussed.

1 INTRODUCTION

Textbook reading has been shown to be an impor-

tant factor in student test scores, both print and elec-

tronic (Daniel and Woody, 2013). While Landrum

et al. (2012) conﬁrmed that quiz score and ﬁnal grade

are signiﬁcantly positively correlated with the self-

reported percentage of completed reading of text-

books, Yang et al. (2021) suggested that repeated

testing could improve reading skills, reading engage-

ment, and reading comprehension (RC), leading to an

enhanced recall of learning content and key concepts.

Studies from educational science have shown that ad-

junct questions are a form of active learning that in-

creases attention to essential parts of the text, and ac-

tive processing of the topic leads to better learning

outcomes (Syed et al., 2020). RC is a complex pro-

cess with as many interpretations as there are of read-

ing because it is often viewed as its essence, essen-

tial for academic and lifelong learning, getting sci-

entiﬁc attention as a cognitive process despite this

fundamental importance only since the 1970s. Fur-

thermore, comprehension monitoring is an important

strategy to improve text understanding and is unlikely

to develop spontaneously (Panel, 2000).

So, both reading textbooks and answering quizzes

https://orcid.org/0000-0003-0002-868X

https://orcid.org/0000-0003-1209-5038

https://orcid.org/0000-0001-8840-6053

lead to better recall of learning content and bet-

ter learning outcomes, especially when both forms

are combined interactively (e.g., Peverly and Wood

(2001); Callender and McDaniel (2007); Panel

(2000); Yang et al. (2021)). Nevertheless, exist-

ing solutions in learning management systems, e.g.,

Moodle, usually offer reading and quizzes separately.

Moreover, previous research has focused on enriching

digital texts or visualizing concept maps or summary

learner metrics in dashboards but rarely on real-time

visualization of personal RC in the text itself.

In their study, Christhilf et al. (2022) implemented

per-paragraph reading strategy detection using man-

ual scoring of Constructed Response Protocols (self-

explanations of the text just read) approximately ev-

ery 35 words. Although RC is measured at this level

of detail, the study’s goal was to potentially provide

general feedback on ineffective strategy patterns, not

to visualize per-paragraph RC. For real-time analysis,

they suggested combining their approach with an au-

tomated system.

This work aims to address this gap by investigat-

ing how to measure individual students’ RC in learn-

ing management systems and how it can be visualized

adaptively to the learning progress and the individual

comprehension level. Thus, this study examines and

answers the following research question:

RQ1: How can digital learning management sys-

tems effectively integrate and visualize RC?

266

Menze, D., Seidel, N. and Radovic, S.

Dynamic Reading Comprehension Visualization in Digital Course Texts.

DOI: 10.5220/0013216100003932

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

In Proceedings of the 17th International Conference on Computer Supported Education (CSEDU 2025) - Volume 1, pages 266-273

ISBN: 978-989-758-746-7; ISSN: 2184-5026

To answer this question, we developed and com-

pared several design variants for displaying RC based

on ﬁndings from the literature review. The most

promising variant was then implemented in a proto-

type and evaluated in an experiment to answer the

second research question:

RQ2: Does displaying measured RC lead to more

learning activity (reading duration, question attempts)

or even better learning as measured by multiple-

choice questions?

2 RELATED WORK

Reading comprehension research has evolved from

early skill-based approaches to more nuanced under-

standings of how readers interact with texts. This sec-

tion synthesizes key ﬁndings that inform our design

choices, focusing on measurement approaches, exist-

ing solutions, and visualization techniques.

2.1 Understanding and Measuring

Reading Comprehension

Early work by Davis (1944) identiﬁed nine hierarchi-

cal mental skills essential for RC, ranging from basic

word recognition to inferring author intent. While this

skills-based perspective has been questioned, subse-

quent research has conﬁrmed that RC involves mul-

tiple interacting abilities (Panel, 2000). The simple

model proposed by Gough and Tunmer (1986) char-

acterizes RC as the product of decoding and listen-

ing comprehension, combining bottom-up word iden-

tiﬁcation with top-down semantic processing (Cutting

and Scarborough, 2006).

Modern approaches emphasize reading as an ac-

tive, constructive process requiring prior knowledge

and engagement with the text (Panel, 2000). Perfetti

et al. (2005) highlighted that deeper comprehension

involves building a mental model of the text’s mes-

sage through processes at word, sentence, and text

levels. Effective comprehension instruction methods

include comprehension monitoring and awareness,

visualization of relationships between ideas, question

answering with feedback, and reader-generated ques-

tions and summaries.

For measuring RC, traditional methods primar-

ily use multiple-choice questions that test different

levels of understanding, as deﬁned by Alonzo et al.

(2009): literal comprehension focusing on surface-

level understanding, inferential comprehension re-

quiring conclusions about author intent, and evalua-

tive comprehension connecting text to prior knowl-

edge. Fletcher (2006) emphasized that readability

signiﬁcantly inﬂuences both the inferences readers

can make and researchers’ comprehension assess-

ment. Recent research has expanded measurement

approaches to include analysis of gaze tracking data

(Caruso et al., 2022), linguistic features and readabil-

ity metrics (Crossley et al., 2017), and scrolling be-

haviour analysis (Gooding et al., 2021).

2.2 Digital Reading Support Systems

While existing learning management systems like

Moodle typically separate reading and quiz activities,

multiple studies have shown that combining reading

and quizzes leads to better learning outcomes (Pev-

erly and Wood, 2001; Callender and McDaniel, 2007;

Yang et al., 2021). Several approaches have emerged

to enhance digital reading experiences. Sun et al.

(2004) focused on visualizing book structure and nar-

rative threads, while Guerra et al. (2013) explored

social comparison of reading progress. Milliner and

Cote (2015) studied student engagement in a special-

ized, extensive reading system, ﬁnding high satisfac-

tion despite lower than expected reading completion.

Boticki et al. (2019) identiﬁed reading styles from

learning log data, distinguishing between receptive

reading (sequential and steady) and response reading

(active engagement with varying pace).

More recent approaches have explored automated

analysis methods. Bravo-Marquez et al. (2011)

used machine learning to classify text comprehen-

sion based on student-generated documents, while

Christhilf et al. (2022) implemented paragraph-level

reading strategy detection. Work by Thaker et al.

(2020) developed methods for automated identiﬁ-

cation of personalized content based on knowledge

gaps, and both Kim et al. (2020) and Wang and

Walker (2021) investigated concept mapping with ex-

pert comparison.

Particularly relevant to our work, Syed et al.

(2020) studied how adjunct questions displayed dur-

ing reading affect learning outcomes, combining gaze

tracking with questions during reading. While they

also placed questions in a side panel to avoid disrupt-

ing the reading ﬂow, they used short free-response

questions that required manual grading. Sun et al.

(2018) demonstrated that online reading duration

strongly indicates reading motivation, which is cru-

cial for enhancing intensive reading behaviors. Their

study showed that groups with high reading duration

also had higher motivation and more phases of inten-

sive reading. However, these approaches require spe-

cialized equipment like eye trackers or manual grad-

ing effort or focus on broader aspects like reading

progress and engagement patterns rather than provid-

Dynamic Reading Comprehension Visualization in Digital Course Texts

267

ing immediate, automated section-level feedback on

comprehension.

Our work addresses this gap by developing and

testing a system that directly integrates questions

with reading material, provides real-time comprehen-

sion visualization, and adapts to individual reading

progress. This synthesis of prior work directly in-

forms our design choices: We adopt multiple-choice

questions as a proven, scalable assessment method,

place questions in a side panel to maintain reading

ﬂow, provide immediate section-speciﬁc comprehen-

sion feedback, and support individual progress mon-

itoring. These design decisions combine the most

effective elements from previous research while ad-

dressing the identiﬁed gap in immediate comprehen-

sion feedback.

3 DESIGN AND REALIZATION

Referring to RQ1, we aim to design and realize a pro-

totype of an adaptive system for supporting RC. In de-

signing our system, we drew on several key ﬁndings

from prior research to address three key questions:

(1) How can questions be displayed without disrupt-

ing the reading ﬂow? (2) How can reading compre-

hension be measured and visualized in real-time? (3)

How can the system adapt to different reading inten-

tions and comprehension levels?

Studies have shown that comprehension monitor-

ing is most effective when it occurs during reading

rather than afterwards (Panel, 2000). Following this

insight and research by Syed et al. (2020) showing

that interrupting the reading ﬂow can be detrimental

to comprehension, we implemented a system where

questions are displayed adaptively based on reading

progress in a side panel rather than embedding them

directly in the text.

While Moodle’s quiz activity already provides

functionality for displaying questions and answers,

the key design decision was how to integrate these

seamlessly with the reading experience. We consid-

ered three possible approaches for embedding ques-

tions: directly in the text under each section, in a

side panel, or as an overlay similar to chatbots. We

chose the side panel approach, following Syed et al.

(2020), as direct embedding would interrupt the read-

ing ﬂow—which is detrimental to RC (Foroughi et al.,

2015)—and an overlay might be perceived negatively

by students due to its association with marketing

tools. The side panel solution maintains a clean read-

ing experience while keeping questions readily acces-

sible.

Regarding the timing of the question display, we

considered different reading intentions identiﬁed by

Peckham and McCalla (2012): thorough reading for

comprehension, quick skimming for an overview, or

targeted searching for speciﬁc information. While

ideally, questions would only be shown during thor-

ough reading, distinguishing between these intentions

proved unnecessary in practice. When skimming,

questions fade too quickly and are distracting, and

students can actively control their engagement by

opening or closing the side panel.

RC is calculated as the ratio of correct answers to

questions after a section or to all questions on the cur-

rent page, a metric validated in multiple studies (Cal-

lender and McDaniel, 2007; Foroughi et al., 2015;

Panel, 2000). Real-time visualization through color-

coded progress bars was implemented following evi-

dence that immediate feedback supports comprehen-

sion monitoring (Panel, 2000). The distinction be-

tween comprehension levels is left to the teacher as

a recommendation when creating and assigning ques-

tions. Ideally, when the instructor ﬁlls a text with

many questions for each section, it results in a kind of

Figure 1: Screenshot of prototype: text (left), reading progress bar (middle), questions (right)

CSEDU 2025 - 17th International Conference on Computer Supported Education

268

text coverage for the students to monitor their learning

progress.

This was approached and implemented in a Moo-

dle environment for a study that collected data about

reading behavior and RC. The prototype extends an

existing Moodle plugin called Longpage (Seidel et al.,

2024), which provides functionalities that simplify

reading on screen or provide advantages that are not

available when reading printed works or PDF ﬁles:

students can annotate texts, mark and comment sec-

tions, and share this information with others.

The integration of a RC estimation into Longpage

breaks down into three main parts: an option for the

teacher to create and edit questions and assign them

to parts of the text, the display of the embedded ques-

tions for the student with the option to answer them,

and the display of the estimated RC for each text sec-

tion. Figure 1 shows a screenshot of the implemented

prototype.

Moodle already provides the functionality to cre-

ate, preview, and edit questions and their answer op-

tions with the so-called question bank.

For the as-

signment of these questions to text sections, a collec-

tion of third-party plugins was installed: Embed ques-

tion atto button

makes it possible to assign questions

to sections. After putting the cursor at a speciﬁc posi-

tion in the text, the teacher can select a question from

the question bank. Then, a cryptic code will be in-

serted inside the text to identify the question. Embed

question ﬁlter

is a Moodle text ﬁlter plugin that con-

verts this cryptic code into HTML for rendering the

question inside of Longpage like in a quiz. Multi-

ple questions must be added sequentially without line

breaks so that the script can correctly associate sec-

tions and questions. The questions are hidden by de-

fault.

Using the Intersection Observer API

available in

a modern web browser, a custom script clones the

HTML code of a hidden question when the student

scrolls over it, pastes it to the side panel, and sets it

to visible. When the original hidden question inside

the text is scrolled outside the view, the cloned ques-

tion in the side panel will be removed. This way, for

the student, the question appears in the side panel as

long as the corresponding text section is visible. If

there are multiple questions per section, they can be

https://docs.moodle.org/311/en/question bank

(accessed 2025/01/10).

https://moodle.org/plugins/atto embedquestion

(accessed 2025/01/10).

https://moodle.org/plugins/ﬁlter embedquestion

(accessed 2025/01/10).

https://www.w3.org/TR/intersection-observer

(accessed 2025/01/10).

clicked through in a carousel by clicking on the left

and right arrows. Thus, students can answer as many

questions as they like until the contingent on ques-

tions is exhausted. With two arrows, up and down, it

is possible to jump to the next section with questions

available.

The reading progress indicator available in Long-

page is shown immediately on the right side of the

text as a bar. Its width signiﬁes how often fellow stu-

dents have read certain text sections. The prototype

displays RC by coloring this existing reading progress

bar according to the estimated comprehension level.

Hovering the mouse cursor over the bar displays the

estimated value.

When the student submits an answer, an AJAX

call is triggered so that the PHP function on the server

is executed to calculate the comprehension values for

the whole page. This is necessary because a question

can be referenced several times on the same page, so

a new answer could potentially change the estimated

RC in multiple sections. This function iterates over all

questions on the page, fetches the student’s last ﬁve

attempts no older than three months for each ques-

tion (to implement a moving average and knowledge

decay), and calculates the average scores. Finally,

all scores are returned to the client, which changes

the color and tooltip text of all reading progress bars

accordingly. An overall RC estimate for the page is

added to the sidebar.

4 EVALUATION

The prototype resulting from answering RQ1 in the

previous section was put to the test in an experiment

described in this section to answer RQ2, i.e., whether

displaying measured RC does lead to more learning

activity or even better learning.

4.1 Methods

4.1.1 Participants

To determine the number of participants needed to

evaluate the prototype, a statistical power analysis

was performed with a signiﬁcance level of α=0.05,

a power of 1-β=0.80. This resulted in a requirement

of n=128 participants to measure medium effects (Co-

hen’s d=0.50) and n=51 participants for large effects

(Cohen’s d=0.80). Subjects were recruited through

the FernUniversit

at Hagen survey pool, a mailing list

to which students can subscribe if they wish to partic-

ipate in university studies. An overview of the exper-

iment with an estimated duration of 30 minutes was

Dynamic Reading Comprehension Visualization in Digital Course Texts

269

Questionnaire

Pretest

Experimental group (n=28):

Read text with RC display

Control group (n=29):

Read text without RC display

Posttest

Figure 2: Flowchart of participants’ progress through the phases of the trial in a horizontal layout with centered elements.

provided in the recruitment email and on the study

homepage. Participation was open for one week in

October 2022. Participants were informed about pri-

vacy, content, objectives, data collected, their rights

as subjects, and the study duration and provided writ-

ten consent to participate per the EU General Data

Protection Regulation, the research ethics guidelines

of the American Psychological Association and the

German Psychological Association. The 57 partici-

pants were between 19 and 65 years old (M=46.79,

SD=12.99). 40 were female, 17 were male, and 0

was diverse. The highest level of education was fairly

evenly distributed and ranged from a completed ap-

prenticeship to a high school diploma to a university

degree.

4.1.2 Procedure

In this study, a randomized controlled trial with a

between-subjects design was conducted. Block ran-

domization (Suresh, 2011) with a block size of two

was used to ensure equal sample sizes in the experi-

mental and control groups (see Figure 2). A course

was set up in a Moodle learning environment for the

experiment.

First, a short demographic questionnaire had to be

completed, asking for age, gender, and highest level

of education. Then, 8 multiple-choice questions were

posed in a pretest to determine prior knowledge on

the topic of learning analytics. Participants were then

asked to read a text about ethical issues and dilem-

mas in learning analytics, an excerpt from Slade and

Prinsloo (2013), translated into German (Flesch read-

ing ease: 61.0, medium readability). The speciﬁc

topic of the text was chosen because, on the one hand,

students themselves are affected by learning analytics

and can, therefore, beneﬁt from the reﬂection during

the study. On the other hand, it was assumed that eth-

ical texts would be more accessible across subjects.

The text had an estimated reading time of 12–17 min-

utes and was presented using the Longpage plugin,

with most of the secondary features (comments, an-

notations, etc.) disabled.

The functions described in Section 3 for measur-

ing RC were activated. However, only the experi-

mental group had RC visualization enabled. On the

other hand, the control group did not even see read-

ing progress bars next to the text but were still shown

the multiple-choice questions per section as the ex-

perimental group and could answer them in the same

way. Both were given questions because, as men-

tioned, studies have already shown that additional

questions lead to higher RC. So, this should not be

studied again; it should be assumed. The 20 questions

were the 8 questions from the pretest plus 12 more. In

the posttest, the same 20 questions are asked again.

4.1.3 Data Collection and Pre-Processing

User interactions within the Moodle environment

were captured in the database, including quiz and

reading activities. The latter were real usage data

on users’ reading behavior from the Intersection Ob-

server (see Section 3), which ﬁres log events as soon

as a text section becomes visible within the viewport

of the users’ display. Text sections had a unique iden-

tiﬁer and contained individual paragraphs and head-

lines. The dataset consisted of about 440.000 log

entries from reading and question attempts. Partici-

pants who spent less than 10 minutes, and thus about

three standard deviations below the minimum esti-

mated reading time of 14 minutes (and more than

three standard deviations below the mean estimated

reading time of 15.5 minutes) on the text page were

excluded from the study, as it was assumed that a sub-

stantive engagement with the text, questions, and RC

estimation was not possible in this short time. In the

actual experiment, there were n=57 participants, 28 in

the experimental group and 29 in the control group.

To analyze potential differences between the

groups, the following confounding variables were

controlled by the questionnaire: age, gender, and

highest level of education; by the pretest: prior knowl-

edge of ethical topics in learning analytics; and by the

data analysis: time of day and day of week, maximum

time spent per participant on the text page, maximum

text section reached, number of scroll events (see Fig-

CSEDU 2025 - 17th International Conference on Computer Supported Education

270

Table 1: Reading and question statistics per participant.

Experimental group (N=28) Control group (N=29)

Mean SD Mean SD

Reading time (min) 27.95 12.41 27.94 11.85

Sections read (%) 100.00 0.00 98.10 10.21

No. of scroll events 117.50 77.01 122.10 69.91

Time per scroll event (s) 23.24 22.78 22.10 24.17

Question attempts 15.61 12.88 16.62 9.58

Figure 3: Total number of scroll events per section with positions of adjunct questions marked.

ure 3), time per scroll event, and number of response

attempts during reading.

From the knowledge tests, pretest scores, posttest

scores, and RC scores were determined for each

group. Pretest scores were calculated from the 8

pretest questions, posttest scores from the same 8

questions, and RC scores from the 12 additional ques-

tions asked in the posttest.

4.2 Results

The anonymized data and analysis scripts used are

publicly available

The differences in reading and question statistics

per participant can be seen in Table 1. These were

not signiﬁcant in each case. All participants read an

average of about 99.04 % (SD=7.28 %) of the text, in

an average of 27.94 minutes (SD=12.02 minutes), and

answered an average of 80.61 % (SD=56.16 %) of the

questions.

The confounding variables identiﬁed through the

questionnaire and data analysis revealed no statisti-

cally signiﬁcant differences between the two groups,

indicating successful randomization.

The differences in pretest, posttest, and RC, all

of which were not statistically signiﬁcant, are shown

in Table 2. While the experimental group showed

slightly higher scores across all measures, these dif-

ferences did not reach statistical signiﬁcance.

4.3 Discussion

While the study showed no statistically signiﬁcant

differences in learning outcomes between groups,

several important insights were gained. First, the

https://osf.io/ftx7h/?view only=0e6a854ddb4845bf

8fdde4228021647d (accessed 2025/01/10).

high completion rates of both reading (98-100 %) and

question-answering (81 %) suggest strong engage-

ment with the system across both conditions. This

indicates that the basic design principles of the pro-

totype - integrating questions with text and provid-

ing an unobtrusive interface - successfully maintained

student engagement.

The lack of signiﬁcant differences between groups

may be attributed to several factors. First, the rela-

tively short exposure time (less than one hour) and

limited content (one medium-length text with twenty

questions) may have been insufﬁcient to demonstrate

the potential beneﬁts of comprehension visualization.

In a real educational setting, students would interact

with the system over an entire semester across mul-

tiple texts, allowing for repeated exposure and prac-

tice effects. However, the unobservable differences

between the two groups may also be due to the fact

that learning analytics instruments do not have such a

decisive inﬂuence on learning activities and learning

success (Hellings and Haelermans, 2022).

Additionally, both groups received embedded

questions during reading, which prior research has

shown to be beneﬁcial for comprehension (Peverly

and Wood, 2001; Callender and McDaniel, 2007).

This design choice may have reduced the potential

impact of the comprehension visualization, as both

groups beneﬁted from active engagement with the

content through questioning. Future studies might

consider comparing the current system against a con-

trol condition with no embedded questions to isolate

the effects of comprehension visualization better.

Despite these limitations, the study provided valu-

able validation of the technical implementation. The

prototype demonstrated robust performance across

multiple platforms with minimal user guidance, effec-

tively serving as a validation prior to potential longer-

term deployment in educational settings. The suc-

Dynamic Reading Comprehension Visualization in Digital Course Texts

271

Table 2: Test statistics.

Experimental group (N=28) Control group (N=29)

Mean SD Mean SD

Pretest grades (%) 59.00 14.61 57.79 13.44

Posttest grades (%) 69.11 14.43 68.04 14.81

RC grades (%) 72.70 15.71 71.24 11.71

cessful implementation of the logging system and ex-

perimental framework also provides a foundation for

future studies as the tool evolves.

5 CONCLUSION AND OUTLOOK

This work demonstrated how reading comprehen-

sion can be modeled and visualized in digital texts

(RQ1). The developed prototype successfully in-

tegrated questions that were adaptive to reading

progress and provided real-time visualization of com-

prehension scores. Ideally, when the instructor ﬁlls a

text with many questions for each section, it results

in a kind of text coverage for the students to monitor

their learning progress. This was approached and im-

plemented in a Moodle environment for a study that

collected data about reading behavior and RC. While

our experimental evaluation (RQ2) did not show sig-

niﬁcant differences in learning outcomes, it validated

the approach’s technical feasibility and user accep-

tance, with high engagement rates across both exper-

imental conditions.

The study’s ﬁndings suggest several promising di-

rections for future research. First, a longer-term ﬁeld

study is needed to evaluate the system’s impact over

an entire semester, where the cumulative effects of

comprehension visualization might become more ap-

parent. This would also allow for investigating how

students’ interaction patterns with the system evolve

over time.

The prototype could be enhanced in several ways

to provide more personalized and adaptive support.

Questions could be preselected based on individual

comprehension levels (Alonzo et al., 2009), ques-

tion difﬁculty, broader learning proﬁles (Yang et al.,

2021), and temporal engagement patterns. The sys-

tem could implement intelligent fading of questions

when comprehension is high, provide targeted help

when performance is low, and suggest speciﬁc sec-

tions for review based on answer patterns (Thaker

et al., 2020). Future versions could incorporate au-

tomated question generation and semantic matching

of content to reduce instructor workload while main-

taining pedagogical quality. RC measurement could

be enhanced by incorporating additional factors such

as text-marking patterns (Yang et al., 2021), reading

frequency, and broader course activities (Hoppe et al.,

2021).

A dashboard showing individual student compre-

hension patterns could enable more targeted interven-

tions for instructors. This could be particularly valu-

able in identifying struggling students early in the

semester. The current prototype is a foundation for

these enhancements, demonstrating robust technical

performance and strong user engagement.

While our initial results did not show signiﬁcant

learning gains, they provide valuable insights for de-

signing digital reading support systems and highlight

important considerations for future research. The

high engagement rates suggest that students are re-

ceptive to integrated RC support, warranting further

investigation of how such systems can be optimized

to maximize their educational impact.

ACKNOWLEDGEMENTS

This research was supported by the Research Clus-

ter “CATALPA – Center of Advanced Technology for

Assisted Learning and Predictive Analytics” of the

FernUniversit

at in Hagen, Germany.

REFERENCES

Alonzo, J., Basaraba, D., Tindal, G., and Carriveau, R. S.

(2009). They read, but how well do they understand?:

An empirical look at the nuances of measuring read-

ing comprehension. Assessment for Effective Inter-

vention, 35:34–44.

Boticki, I., Ogata, H., Tomiek, K., Akcapinar, G., Flanagan,

B., Majumdar, R., and Hasnine, N. (2019). Identifying

reading styles from e-book log data. pages 312–317.

Asia-Paciﬁc Society for Computers in Education.

Bravo-Marquez, F., L’Huillier, G., Moya, P., R

ıos, S. A.,

and Vel

asquez, J. D. (2011). An automatic text com-

prehension classiﬁer based on mental models and la-

tent semantic features.

Callender, A. A. and McDaniel, M. A. (2007). The beneﬁts

of embedded question adjuncts for low and high struc-

ture builders. Journal of Educational Psychology, 99.

Caruso, M., Peacock, C., Southwell, R., Zhou, G., and

D’Mello, S. (2022). Going deep and far: Gaze-based

CSEDU 2025 - 17th International Conference on Computer Supported Education

272

models predict multiple depths of comprehension dur-

ing and one week following reading. Intl. EDM Soci-

ety.

Christhilf, K., Newton, N., Butterfuss, R., McCarthy, K. S.,

Allen, L. K., Magliano, J. P., and McNamara, D. S.

(2022). Using markov models and random walks to

examine strategy use of more or less successful com-

prehenders. International Educational Data Mining

Society.

Crossley, S. A., Skalicky, S., Dascalu, M., McNamara,

D. S., and Kyle, K. (2017). Predicting text compre-

hension, processing, and familiarity in adult readers:

New approaches to readability formulas. Discourse

Processes, 54:340–359.

Cutting, L. E. and Scarborough, H. S. (2006). Predic-

tion of reading comprehension: Relative contributions

of word recognition, language proﬁciency, and other

cognitive skills can depend on how comprehension is

measured. Scientiﬁc Studies of Reading, 10:277–299.

Daniel, D. B. and Woody, W. D. (2013). E-textbooks at

what cost? Performance and use of electronic v. print

texts. Computers and Education, 62:18–23.

Davis, F. B. (1944). Fundamental factors of comprehension

in reading. Psychometrika, 9:185–197.

Fletcher, J. M. (2006). Measuring reading comprehension.

Scientiﬁc Studies of Reading, 10:323–330.

Foroughi, C. K., Werner, N. E., Barrag

an, D., and Boehm-

Davis, D. A. (2015). Interruptions disrupt reading

comprehension. Journal of Experimental Psychology,

144.

Gooding, S., Berzak, Y., Mak, T., and Shariﬁ, M. (2021).

Predicting text readability from scrolling interactions.

Association for Computational Linguistics.

Gough, P. B. and Tunmer, W. E. (1986). Decoding, reading,

and reading disability. Remedial and Special Educa-

tion, 7:6–10.

Guerra, J., Parra, D., and Brusilovsky, P. (2013). Encour-

aging online student reading with social visualization

support. Proceedings of the Workshops at the 16th

International Conference on Artiﬁcial Intelligence in

Education AIED, pages 47–50.

Hellings, J. and Haelermans, C. (2022). The effect of pro-

viding learning analytics on student behaviour and

performance in programming: a randomised con-

trolled experiment. Higher Education, 83:1–18.

Hoppe, H. U., Doberstein, D., and Hecking, T. (2021).

Using sequence analysis to determine the well-

functioning of small groups in large online courses.

International Journal of Artiﬁcial Intelligence in Edu-

cation, 31:680–699.

Kim, M. K., Gaul, C. J., Bundrage, C. N., and Madathany,

R. J. (2020). Technology supported reading com-

prehension: a design research of the student mental

model analyzer for research and teaching (smart) tech-

nology. Interact. Learn. Environ.

Landrum, R. E., Gurung, R. A., and Spann, N. (2012).

Assessments of textbook usage and the relationship

to student course performance. College Teaching,

60:17–24.

Milliner, B. and Cote, T. (2015). One year of extensive read-

ing on mobile devices: engagement and impressions.

pages 404–409. Research-publishing.net.

Panel, N. R. (2000). Teaching children to read: An

evidence-based assessment of the scientiﬁc research

literature on reading and its implications for reading

instruction. NIH Publication No. 00-4769, 7.

Peckham, T. and McCalla, G. (2012). Mining student be-

havior patterns in reading comprehension tasks.

Perfetti, C. A., Landi, N., and Oakhill, J. (2005). The acqui-

sition of reading comprehension skill. pages 227–247.

Peverly, S. T. and Wood, R. (2001). The effects of ad-

junct questions and feedback on improving the read-

ing comprehension skills of learning-disabled adoles-

cents. Contemporary Educational Psychology, 26:25–

43.

Seidel, N., Stritzinger, A., Menze, D., and Friedrich, K.

(2024). Longpage - A Moodle activity plugin de-

signed to assist learners with reading extended texts.

Slade, S. and Prinsloo, P. (2013). Learning analytics: Ethi-

cal issues and dilemmas. American Behavioral Scien-

tist, 57:1510–1529.

Sun, J. C. Y., Lin, C. T., and Chou, C. (2018). Applying

learning analytics to explore the effects of motivation

on online students’ reading behavioral patterns. In-

ternational Review of Research in Open and Distance

Learning, 19:209–227.

Sun, Y., Harper, D. J., and Watt, S. N. K. (2004). De-

sign of an e-book user interface and visualizations to

support reading for comprehension. pages 510–511.

ACM Press.

Suresh, K. (2011). An overview of randomization tech-

niques: An unbiased assessment of outcome in clin-

ical research. Journal of Human Reproductive Sci-

ences, 4:8.

Syed, R., Collins-Thompson, K., Bennett, P. N., Teng, M.,

Williams, S., Tay, D. W. W., and Iqbal, S. (2020). Im-

proving learning outcomes with gaze tracking and au-

tomatic question generation. pages 1693–1703. ACM.

Thaker, K. M., Zhang, L., He, D., and Brusilovsky, P.

(2020). Recommending remedial readings using stu-

dent knowledge state. International Conference on

Education Data Mining.

Wang, S. and Walker, E. (2021). Providing adaptive feed-

back in concept mapping to improve reading compre-

hension. Conference on Human Factors in Computing

Systems - Proceedings.

Yang, A. C., Chen, I. Y., Flanagan, B., and Ogata, H. (2021).

Automatic generation of cloze items for repeated test-

ing to improve reading comprehension. Educational

Technology and Society, 24:147–158.

Dynamic Reading Comprehension Visualization in Digital Course Texts

273