Innovation or Imitation? A Critical Analysis of AI-Authored vs.

Human-Authored Scientiﬁc Papers

Corinna H

ormann

1 a

, Lisa Kuka

1 b

, Anneliese Fraser

2 c

and Barbara Sabitzer

1 d

STEM Education, Johannes Kepler University Linz, Altenbergerstraße 69, Linz, Austria

Teacher Education Centre, University of Passau, Innstraße 41, Passau, Germany

Keywords:

Artiﬁcial Intelligence, Academic Publishing, Human vs AI Authorship, Digital Transformation.

Abstract:

The use of Artiﬁcial Intelligence (AI) in research and academic publishing has been a topic of growing inter-

est and debate. While some argue that AI-based systems have the potential to revolutionize the way scien-

tiﬁc papers and academic work are generated, others express concerns about the authenticity of AI-authored

papers. Several respected organizations have recently developed guidelines regarding the use of AI in schol-

arly manuscripts and publishing. This critical analysis will examine the advantages and disadvantages of

AI-authored scientiﬁc papers compared to those authored by humans. The underlying work describes the pro-

cess of creating different papers solely with the help of ChatGPT or Jenni AI and compares them to human

written drafts. Therefore, both AI tools were asked to generate scientiﬁc papers about the “History of Digi-

tal Education in Austria”, “A History of Women in Computer Science”, and “Modelling of Mental Arithmetic

Strategies Using UML”. In conclusion, it is indisputable that AI-driven tools signiﬁcantly facilitate the drafting

of outlines, titles, and the composition of papers. However, the creation of a high-quality scientiﬁc academic

publication still demands considerable human input, encompassing both creative effort and critical thinking,

to ensure depth, originality, and scholarly rigor.

1 INTRODUCTION

During the resulting Emergency Remote Teaching

(ERT) due to the COVID-19 pandemic, students tried

to ﬁnd new ways to interact with technology and

adapted to new tools (Vargo et al., 2021). Arti-

ﬁcial Intelligence (AI) technologies, including lan-

guage translation, plagiarism detection, grammar and

spelling checks, and AI-generated essay outlines, are

widely acknowledged as beneﬁcial tools that enhance

the writing process and assist scholars and students

in their academic endeavors. Furthermore, research

shows that AI improves students’ writing skills, sense

of self-efﬁcacy, and comprehension of academic in-

tegrity, all of which have a favorable impact on aca-

demic writing. Still, students seem to be worried

about how it will affect their ability to be creative,

think critically, and behave ethically (Malik et al.,

2023).

https://orcid.org/0000-0002-4770-6217

https://orcid.org/0000-0002-0000-5915

https://orcid.org/0009-0000-9400-0666

https://orcid.org/0000-0002-1304-6863

With the inclusion of Large Language Models

(LLMs) into Digital Writing Assistants (DWAs) like

Grammarly or WordTune, students became confused

whether it is legitimate to use these tools for academic

writing. It is even possible to generate citations along-

side the work created by the AI, which are referred to

as Automatic Article Generators (AAGs) (El-Sayed

Abd-Elaal and Mills, 2022). Researchers, educators,

and anti-plagiarism software all face additional work

when identifying a piece of work produced by AAGs.

More and more universities struggle with the con-

cerns related to the integration of LLMs in student

projects. Several authors already tried to answer the

question, whether or not using AI technologies is a vi-

olation of academic integrity (Perkins, 2023). The In-

ternational Center of Academic Integrity (ICAI), es-

tablished in 1992 by notable scholars, provides an au-

thorized deﬁnition of academic integrity. Its founder,

Don McCabe, is recognized for having popularized

the term. The Center ﬁrst deﬁned the “fundamen-

tal values of academic integrity” in 1999 as honesty,

trust, fairness, respect, and responsibility. In 2014,

they added the virtue of courage as a sixth compo-

nent. According to the ICAI, academic integrity is

534

Hörmann, C., Kuka, L., Fraser, A. and Sabitzer, B.

Innovation or Imitation? A Critical Analysis of AI-Authored vs. Human-Authored Scientiﬁc Papers.

DOI: 10.5220/0012712400003693

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

In Proceedings of the 16th International Conference on Computer Supported Education (CSEDU 2024) - Volume 1, pages 534-541

ISBN: 978-989-758-697-2; ISSN: 2184-5026

a dedication to these principles (International Center

for Academic Integrity, 2021). Many other deﬁni-

tions exist and are feasible. However, this policy does

not include whether students perform academic mis-

conduct intentionally or not. However, in response

to the growing inﬂuence of AI in academic writing,

the Faculty of Business Administration at the Univer-

sity of Economics, Prague, has undertaken a signiﬁ-

cant curricular reform in winter 2023. This reform re-

sults in the discontinuation of traditional written bach-

elor’s theses in favor of a diverse array of practical as-

sessments, collectively termed “bachelor’s projects”

(Friedmann, 2023).

The results from a survey to determine students’

perceptions of AI tools, underlined the signiﬁcance

of appropriate instruction and guidelines for using

AI tools in order to prevent plagiarism problems

and maintain academic integrity (Kreps et al., 2022;

Holmes et al., 2021). Perkins (2023) even considers

LLMs as potential tools to reduce cognitive demands

that are required by a certain task and does not neces-

sarily see them as a violation of academic integrity.

This paper introduces an experiment of creating

different academic papers solely with the help of

ChatGPT or Jenni AI and comparing them to human-

written drafts. A brief introduction to the topic is

given in Section one, while Section two outlines the

development of AI in scientiﬁc and academic writ-

ing. The third section explains the AI tools used in the

experiment, whereas the methodology is described in

Chapter four. The conducted experiment, examining

human- and AI-authored work, is outlined in Section

ﬁve and discussed in Section six. Chapter seven pro-

vides a conclusion and an outlook on upcoming work.

2 AI TOOLS IN THE

UNDERLYING EXPERIMENT

For this experiment two different tools were used to

generate content for a scientiﬁc paper: ChatGPT-4

and Jenni AI (Ultimate Plan).

In the present study, ChatGPT-4, available at a

subscription cost of $ 20 per month, was selected

for its enhanced capabilities over its predecessor

ChatGPT-3. These improvements include more pre-

cise responses to queries, the ability to generate im-

ages through AI (integrating DALL · E), advanced

web browsing functionalities, and superior data anal-

ysis features.

The premium version of Jenni AI, priced at $ 20

per month, offers users an unrestricted word count al-

lowance daily. In contrast to the complimentary vari-

ant, Jenni AI imposes a limit of 200 AI-generated

words per day.

The signiﬁcant difference in creating scientiﬁc pa-

pers with ChatGPT to Jenni AI is that the latter makes

a whole draft with a single line of input, whereas

ChatGPT users have to request every single step and

reﬁne them. Still, both tools can be seen as AI co-

pilots that assist or augment human capabilities in var-

ious tasks or processes (see Figure 1).

Figure 1: Co-piloting with AI tools (Revell et al., 2023)

(adapted by the authors).

2.1 ChatGPT

ChatGPT by OpenAI was chosen for this experiment

because of obvious reasons. It currently is the most

well-known AI chatbot and a very potent tool. Nei-

ther is ChatGPT the ﬁrst GPT model, nor the ﬁrst lan-

guage model. However, it signiﬁcantly advanced nat-

ural language processing by popularizing huge lan-

guage models and quickening the use of artiﬁcial in-

telligence. In ChatGPT, the GPT stands for “Gen-

erative Pre-Trained Transformer”, a LLM that mim-

ics human speech through deep learning (Yosifova,

2023). ChatGPT can offer several signiﬁcant bene-

ﬁts in idea generation, outline creation, or writing and

editing support.

2.2 Jenni AI

In contrast to ChatGPT, Jenni AI is a speciﬁc tool for

academic purposes, enabling to correct grammatical

mistakes, provide citations in the work (which is not

possible with ChatGPT), and create strong arguments.

Still, it has to be noted that Jenni AI is also GPT

based. With its emphasis on academic writing and

integrated plagiarism detector, Jenni AI is ideal for

both academic essays and personal statements (Orig-

inality.ai, 2023). Jenni AI was created in 2016 by

David Park and Henry Mao. In 2020, Park relaunched

Jenni AI, and because to some new features and Tik-

Tok’s attention, more people started using it. Most

AI programs just reply to queries with long text pas-

Innovation or Imitation? A Critical Analysis of AI-Authored vs. Human-Authored Scientiﬁc Papers

535

sages. Jenni AI works on user’s behalf more closely

and learns from the input. She will also remind the

prompt engineer to write for him- or herself if he or

she depends too much on the AI’s help (Giles, 2023).

3 METHODOLOGY

This experiment is designed to critically analyze and

compare the quality, authenticity, and overall effec-

tiveness of scientiﬁc papers authored by Artiﬁcial In-

telligence (AI) tools, speciﬁcally ChatGPT and Jenni

AI, against those authored by human researchers. The

experiment focuses on three distinct topics:

1. History of Digital Education in Austria

2. History of Women in Computer Science

3. Modeling of Mental Arithmetic Strategies Using

UML

One of the human-authored papers (History of

Digital Education in Austria) was written in English

by the authors themselves and accepted at a subject

relevant international conference (International Con-

ference on Informatics in Schools) (H

ormann et al.,

2022). The two other works (History of Women in

Computer Science, Modeling of Mental Arithmetic

Strategies Using UML) were handed in in German as

Bachelor theses at the authors’ department and were

both graded with “Excellent”. The human authors

were briefed about the scope and expectations of the

experiment, ensuring their anonymity. To maintain

an unbiased comparison, the human authors indepen-

dently developed their texts without exposure to the

AI-generated content.

The underlying experiment utilizes two AI tools,

ChatGPT and Jenni AI. These tools were selected

based on their advanced language processing capabil-

ities and relevance in academic research. For each

topic, both AI tools were tasked to create compre-

hensive scientiﬁc text. The inputs provided were

standardized to ensure consistency in the information

available to both AI systems.

The AI-crafted papers were evaluated on multi-

ple dimensions including accuracy of content, depth

of analysis, coherence in structure, and adherence to

academic writing standards. Furthermore, each work

has been tested with the plagiarism detection tool

“Turnitin Similarity” which is commonly employed

in academic and educational settings to ensure the in-

tegrity and originality of written work.

The three human-authored papers, as well as the

six AI-created drafts, can be found online on GitHub

following the QR code in Figure 2 or the link https:

//github.com/corinnahoermann/AI vs Human Paper.

Figure 2: Sources of Human-Authored and AI-Drafted Pa-

pers.

4 EXPERIMENT

4.1 Paper 1: History of Digital

Education in Austria

4.1.1 Human-Authored Paper

The underlying human-authored paper is entitled

“From Non-Existent to Mandatory in Five Years –

The Journey of Digital Education in the Austrian

School System” and explores the evolution of the sub-

ject Digital Education in Austria, from the introduc-

tion of Computer Science in schools in 1985 to the

implementation of Digital Education as a mandatory

subject in the 2022/23 school year, encompassing dig-

ital competences, media literacy, and civic education.

The manuscript spans a total of ten pages, includ-

ing about 30,000 characters, and adheres to the con-

ventional structural framework typical of literature-

based academic publications:

(1) Introduction

(2) Computer Science in Austrian Schools

(3) Digital Education in Austria

(a) Masterplan for Digitalization

(b) 8-Point-Concept

(d) Compulsory Subject Digital Education in Aus-

tria

(4) Conclusion and Outlook

(5) References

4.1.2 AI-Authored Draft – Jenni AI

To initiate the process of composing a paper using

Jenni AI, the following steps are required: The ﬁrst

step is to deﬁne what “you are writing today?”. To

re-create the human-authored paper the prompt a re-

search paper about the history of the school sub-

ject “Digital Education” (digitale Grundbildung) in

CSEDU 2024 - 16th International Conference on Computer Supported Education

536

Austria was typed in. Jenni AI assessed the input

as “Great prompt” and the authors also ticked the

box “Outline Builder: automatically create document

headings”.

Immediately, the AI-generated document, entitled

“Austrian Digital Education History”, was presented,

exhibiting an organizational framework. The struc-

ture of the paper was delineated as follows:

(1) Introduction to Digital Education in Austria

(2) Historical Development of Digitale Grundbildung

(3) Evolution of Digital Education Curriculum in

Austrian Schools

(4) Key Milestones in Austria’s Digital Education

Journey

(5) Impact of Digital Education on Austrian School

System

(6) Challenges and Solutions in Implementing Digi-

tale Grundbildung

(7) Case Studies: Successful Implementations of

Digital Education in Austria

(8) Future Outlook: Prospects for Digital Education

in Austria

(9) Conclusion: Reﬂections on the History of Digital

Education in Austria

(10) References

Next, the authors just let the tool ﬁll the sections

with text without editing it any further and Jenni AI

came up with seven pages in the ﬁrst version. As there

was no abstract, the authors added the headline man-

ually and again let the tool ﬁll the section with text

on its own. In total, the AI drafted seven pages with

about 24,000 characters, as well as eight valid refer-

ences. Thirty percent of the information was ﬂagged

by the “Turnitin” software as possibly copied, which

was the highest of all papers.

4.1.3 AI-Authored Draft – ChatGPT

For creating a paper in ChatGPT the authors used

a similar prompt: I want to write a research paper

about the history of the school subject “Digital Edu-

cation” (digitale Grundbildung) in Austria and Chat-

GPT provided tips to structure the work:

(1) Introduction

(2) Historical Overview

(3) Conclusion

There was no abstract and no references provided.

Still, the authors separately asked for an abstract.

When prompting: Can you create a title for this pa-

per for me?, ChatGPT provided “Navigating the Dig-

ital Wave: A Historical Analysis of Digital Education

in Austria”. The prompt engineers then had to sep-

arately tell ChatGPT to write the beforehand deﬁned

sections, like: Can you write the introduction for me

concerning this paper?

It has to be mentioned that in this case ChatGPT

was not able to provide valid references as it cannot

access or verify some sources, such as the latest sci-

entiﬁc journals, books, or articles. Still, ChatGPT

drafted ten pages with about 30,000 characters, while

the “Turnitin” software detected 6 % of potential pla-

giarism.

Moreover, the authors instructed ChatGPT to pro-

duce an image that visually conveys the paper in order

to illustrate it (see Figure 3).

Figure 3: Image created by ChatGPT visually represent-

ing “Navigating the Digital Wave: A Historical Analysis of

Digital Education in Austria”.

4.2 Paper 2: History of Women in

Computer Science

4.2.1 Human-Authored Paper

The bachelor thesis with the title “The Female Side of

Computer Science” comprises 46 pages (about 85,500

characters) in total and highlights the often over-

looked contributions of women in the ﬁeld of Com-

puter Science, despite their active and ongoing in-

volvement in its rapid advancements. It aims to show-

case a selection of pioneering women, detailing their

lives and achievements, and examining how their sig-

niﬁcant research has shaped and continues to inﬂu-

ence the ﬁeld. It is outlined like the following:

(1) Introduction

Innovation or Imitation? A Critical Analysis of AI-Authored vs. Human-Authored Scientiﬁc Papers

537

(2) Famous Female Computer Scientists

(3) Summary and Conclusion

(4) References

4.2.2 AI-Authored Draft – Jenni AI

Again the authors prompted a single line: I want to

write a thesis about the most important women in

Computer Science and conﬁrmed the outline builder.

The work was entitled “Inﬂuential Women in Com-

puter Science” and had the following structure:

(1) Introduction to Women in Computer Science

(2) Historical Context of Women’s Contributions to

Computing

(3) Challenges Faced by Women in Computer Sci-

ence

(4) Key Female Figures in the Development of Com-

puter Science

(5) The Impact of Women on Computer Science Evo-

lution

(6) Challenges and Triumphs of Women in the Tech

Industry

(7) Case Studies: Inﬂuential Women in Computer

Science

(8) Future Outlook: Encouraging Female Participa-

tion in Computing

(9) Conclusion: Recognizing Women’s Role in Shap-

ing Computer Science

The abstract was again missing but prompted by the

authors.

Like before, the text was generated automatically

and the AI tool came up with a total of seven pages,

consisting of approximately 24,000 characters. De-

spite the activation of the “auto cite from new sources

- external sources will be considered” feature, Jenni

AI did not include any external references in its out-

put. Moreover, the content was found to contain 24 %

instances of plagiarism according to the “Turnitin”

software.

4.2.3 AI-Authored Draft – ChatGPT

Like before, the authors asked ChatGPT to write a

bachelor thesis about the most important women in

computer science and to create a title. The AI tool

came up with “Pioneering Code: The Untold Stories

of Women in Computer Science”. The outline that

was provided, looked like the following:

(1) Introduction

(2) Historical Context and Gender Dynamics in

STEM

(3) The Pioneers of Computer Science

(4) The ENIAC Programmers and Post-War Comput-

ing

(5) Modern Trailblazers in Computer Science

(6) Contemporary Issues and Progress

(7) Conclusion

Fourteen pages, or about 43,000 characters, were

produced by ChatGPT in all, while the “Turnitin”

software identiﬁed 17 % of the content as potentially

plagiarized.

To represent the thesis visually, the authors told

ChatGPT to create an image that represents this thesis

(see Figure 4).

Figure 4: Image created by ChatGPT visually representing

“Pioneering Code: The Untold Stories of Women in Com-

puter Science”.

4.3 Paper 3: Modeling of Mental

Arithmetic Strategies Using UML

4.3.1 Human-Authored Paper

The last human-authored paper that was examined

holds the title “Modeling of Mental Arithmetic Strate-

gies Using UML” and explores the integration of Uni-

ﬁed Modeling Language (UML) into Mathematics

education, speciﬁcally in the context of mental arith-

metic. It includes an analysis of mental arithmetic

strategies, represented in UML diagrams to visually

demonstrate their structure and relationships, while

also discussing the limitations of the study, noting that

it establishes a connection between mental arithmetic

and UML diagrams. The structure is outlined as fol-

lows:

(1) Introduction and Deﬁnition of Terms

CSEDU 2024 - 16th International Conference on Computer Supported Education

538

(2) What is UML and How Can It Be Used in Mathe-

matics Lessons?

(3) Mental Arithmetic Strategies

(4) Summary and Outlook

(5) References

It spans 31 pages and consists of approximately

44,000 characters.

4.3.2 AI-Authored Draft – Jenni AI

When Jenni AI was asked to create a bachelor thesis

about the integration of Uniﬁed Modeling Language

(UML) into Mathematics education, speciﬁcally in

the context of mental arithmetic it produced six pages

with approximately 17,000 characters and four reli-

able references. Additionally, the “Turnitin” program

revealed that there were 13 % instances of plagiarism

in the content. The following outline for the thesis en-

titled “UML Integration in Mental Arithmetic: Bach-

elor Thesis” was created by Jenni AI:

(1) Introduction

(2) Exploring the Role of UML in Enhancing Mental

Arithmetic

(3) Theoretical Framework: UML Meets Mathemat-

ics Pedagogy

(4) Case Studies: UML Application in Learning

Mental Arithmetic

(5) Methodology for Integrating UML into Math Cur-

ricula

(6) Assessing the Impact of UML on Mathematical

Cognitive Development

(7) Challenges and Solutions in the Implementation

of UML Tools

(8) Conclusion and Outlook

(9) References

The prompt engineers also told Jenni AI to cre-

ate an example using an UML diagram how mental

arithmetic could be displayed but the tool did not un-

derstand the instruction.

4.3.3 AI-Authored Draft – ChatGPT

A total of 18 pages with about 63,000 characters

was generated by ChatGPT when asked to create a

bachelor thesis (several sub-prompts were typed in

of course). However, the “Turnitin” program found

5 % of possible plagiarism. The tool drafted the title

“Integrating Uniﬁed Modeling Language (UML) into

Mathematics Education: A Theoretical Approach to

Enhancing Mental Arithmetic Skills” and outlined the

thesis the following way:

(1) Introduction

(2) Literature Review

(3) Theoretical Framework for UML Integration in

Mathematics Education

(4) Hypothetical Application in Education

(5) Potential Impact and Hypothetical Outcomes

(6) Conclusion

When ChatGPT was asked to create an example

using an UML diagram how mental arithmetic could

be displayed, the AI drafted a picture investigating the

calculation (5 + 3) × 2 (see Figure 5). Sadly, neither

the text nor the diagram make sense.

Figure 5: UML diagram outlining the process of solving

(5 +3)× 2 created by ChatGPT.

5 DISCUSSION

Jenni AI does not draft an abstract by default, despite

being one of the most popular text editors for creat-

ing research papers or other academic work. How-

ever, when this AI tool creates one, the output is valid.

Moreover, the text of Jenni AI was somewhat super-

ﬁcial, characterized by a tendency to reiterate points

without substantial depth. Jenni AI takes a lot of work

to produce long texts, as one must repeatedly motivate

the AI to draft additional text. Furthermore, it can

be observed that the titles Jenni AI creates are simple

and could be more creative. Still, the option to utilize

valid references is a huge help.

In contrary to Jenni AI, ChatGPT does not draft a

whole paper by prompting a single sentence. The user

has to specify many prompt lines and copy the results

into a text-processing application on their own. The

Innovation or Imitation? A Critical Analysis of AI-Authored vs. Human-Authored Scientiﬁc Papers

539

text output by ChatGPT contains many more charac-

ters and does not reiterate points or statements. How-

ever, ChatGPT is not able to provide any valid refer-

ences at all.

Nevertheless, it is essential to acknowledge that

each of the six papers drafted by AI systems, whether

it was Jenni AI or ChatGPT, adhered to the estab-

lished standards of academic writing, including as-

pects such as clarity and coherence, but excluding ci-

tation accuracy.

The plagiarism checker’s response to AI-

generated texts showed an interesting pattern. In

texts drafted by ChatGPT, approximately 9 % on

average (5 - 17 %) of the content was ﬂagged for

potential plagiarism. This relatively low percentage

suggests a degree of originality in AI-generated

content. However, a notable increase in potential

plagiarism was observed in Jenni AI generated texts,

where the detection rate was signiﬁcantly higher with

about 22 % on average (13 - 30 %). The discrepancy

in plagiarism detection between texts generated by

Jenni AI and ChatGPT raises important questions

concerning the data sources and methods employed

by different AI tools. It implies that Jenni AI may

depend more on pre-existing text sources, which

then raise the rate of content matching. On the other

hand, texts produced by models like ChatGPT have

a lower detection rate, which suggests an advanced

approach to content generation that may involve more

innovative information combinations.

5.1 Discussion of Topic 1: History of

Digital Education in Austria

Each paper follows a traditional academic structure

but focuses on different aspects of digital education

in Austria. The ones by ChatGPT and Jenni AI are

more focused on historical and policy perspectives,

while the human-authored paper includes a compara-

tive analysis and speciﬁc case studies.

5.2 Discussion of Topic 2: History of

Women in Computer Science

Every paper adopts a somewhat similar structure,

starting with an introduction and background, mov-

ing into detailed discussions of key female ﬁgures,

and concluding with future outlooks and recommen-

dations. However, the focus and depth of content dif-

fer, with some emphasizing biographical case stud-

ies and others more on the overall impact and chal-

lenges. In the assessment of the auto-generated con-

tent of both AI tools, it could be observed that the cur-

riculum vitae of the selected female scientists could

be used without modiﬁcation, demonstrating a satis-

factory level of detail and accuracy. However, Jenni

AI produced a somehow confusing outline, by cre-

ating two different chapters dealing with challenges

of women in Computer Science (“Challenges Faced

by Women in Computer Science”, “Challenges and

Triumphs of Women in the Tech Industry”) and an-

other two for describing female ﬁgures (“Key Female

Figures in the Development of Computer Science”,

“Case Studies: Inﬂuential Women in Computer Sci-

ence”).

5.3 Discussion of Topic 3: Modeling of

Mental Arithmetic Strategies Using

UML

Each paper adopts a structured academic approach,

starting with an introduction and theoretical back-

ground, moving into the application and analysis of

UML in educational contexts, and concluding with

implications and future outlooks. However, they vary

in their focus areas, with two papers concentrating

more on mental arithmetic, while the third paper by

ChatGPT takes a broader view of UML’s integra-

tion in mathematics education. However, the human-

authored one is the only one that includes valid UML

diagrams.

6 CONCLUSION AND OUTLOOK

The exploration of AI-authored versus human-

authored scientiﬁc texts in this experiment highlights

the evolving landscape of academic writing. AI tools,

represented by ChatGPT and Jenni AI, demonstrate

proﬁciency in generating structurally sound and co-

herent academic texts. However, their capabilities

are currently best utilized as augmentative tools rather

than replacements for human intellect and creativity.

The experiment’s investigation into the capacities

and limitations of AI in academic writing demon-

strates the current state and potential of AI tools in

scholarly work. While AI tools like ChatGPT and

Jenni AI demonstrate remarkable abilities in draft-

ing structured, coherent papers and theses, they still

require human oversight for depth, originality, and

academic rigor. Particularly, the underlying work re-

veals that AI can efﬁciently generate outlines, titles,

and even complete drafts, but these outputs often lack

in understanding and critical analysis that human ex-

pertise brings. Moreover, the experiment highlights

a signiﬁcant difference in the approach of AI tools in

creating content. For instance, Jenni AI’s ability to

CSEDU 2024 - 16th International Conference on Computer Supported Education

540

draft entire papers with minimal input contrasts with

ChatGPT’s requirement for detailed prompts at each

step. This distinction underscores the diverse method-

ologies inherent in AI tools and their varied applica-

tions in academic writing.

The ethical implications of AI in academic author-

ship also form a critical part of future discussion, as

there is a vast need for clear guidelines regarding the

use of AI in scholarly work. This now is especially

important as AI tools become more and more acces-

sible.

To further grasp the consequences of AI in edu-

cational contexts, more empirical research is neces-

sary, particularly in fostering creativity, critical think-

ing, and maintaining academic integrity. Addition-

ally, the exploration of AI’s potential in aiding di-

verse academic tasks, such as data analysis and lit-

erature review, is recommended as a future research

direction. In conclusion, while AI represents a sig-

niﬁcant advancement in academic writing, its role re-

mains supplementary to human intellect and creativ-

ity. The ﬁndings of this experiment lay a foundation

for future exploration into the evolving relationship

between AI and human authorship in academia, em-

phasizing a need for an ethical balanced approach to

integrate AI tools in scholarly practices.

REFERENCES

El-Sayed Abd-Elaal, S. H. G. and Mills, J. E. (2022). As-

sisting academics to identify computer generated writ-

ing. European Journal of Engineering Education,

47(5):725–745.

Friedmann, S. (2023). The Faculty of Economics cancels

written bachelor theses.

Giles, L. (2023). Review of Jenni.ai: An Essay writer for

Students.

Holmes, W., Porayska-Pomsta, K., Holstein, K., Suther-

land, E., Baker, T., Shum, S. B., Santos, O. C., Ro-

drigo, M. T., Cukurova, M., Ig, ., Bittencourt, I., and

Koedinger, K. R. (2021). Ethics of AI in education:

Towards a community-wide framework. International

Journal of Artiﬁcial Intelligence in Education.

ormann, C., Schmidthaler, E., Kuka, L., Rottenhofer, M.,

and Sabitzer, B. (2022). From non-existent to manda-

tory in ﬁve years – the journey of digital education

in the austrian school system. Local Proceedings of

the 15th International Conference on Informatics in

Schools: Situation, Evolution, and Perspectives, IS-

SEP 2022.

International Center for Academic Integrity (2021). The

fundamental values of academic integrity third edi-

tion.

Kreps, S., McCain, R. M., and Brundage, M. (2022). All

the news that’s ﬁt to fabricate: AI-generated text as a

tool of media misinformation. Journal of Experimen-

tal Political Science, 9(1):104–117.

Malik, A. R., Pratiwi, Y., Andajani, K., Numertayasa, I. W.,

Suharti, S., and Darwis, A. (2023). Exploring artiﬁ-

cial intelligence in academic essay: Higher education

student’s perspective. International Journal of Educa-

tional Research Open, 5.

Originality.ai (2023). Jenni AI review – supercharge your

next research paper.

Perkins, M. (2023). Academic integrity considerations of

AI large language models in the post-pandemic era:

ChatGPT and beyond. Journal of University Teaching

& Learning Practice, 20.

Revell, T., Yeadon, W., Cahilly-Bretzin, G., Clarke, I., Man-

ning, G., Jones, J., Mulley, C., Pascual, R., Bradley,

N., Thomas, D., and Leneghan, F. (2023). ChatGPT

versus human essayists: An exploration of the impact

of artiicial intelligence for authorship and academic

integrity in the humanities.

Vargo, D., Zhu, L., Benwell, B., and Yan, Z. (2021). Digi-

tal technology use during covid-19 pandemic: A rapid

review. Human Behavior and Emerging Technologies,

3(1):13–24.

Yosifova, A. (2023). The evolution of ChatGPT: History

and future.

Innovation or Imitation? A Critical Analysis of AI-Authored vs. Human-Authored Scientiﬁc Papers

541