Using Artificial Intelligence in Higher Education

Mario Konecki

, Mladen Konecki

and Ivana Ogrizek Biškupić

Faculty of Organization and Informatics, University of Zagreb, Varaždin, Croatia

Algebra University College, Zagreb, Croatia

Keywords: Higher Education, Artificial Intelligence, Virtual Assistants, Chatbots, Opportunities, Challenges.

Abstract: Higher education always had its challenges that professors and other educators had to deal with. Along with

challenges there are also different opportunities that various technologies provide in the form of usable

educational aiding means. Sometimes the same technology can represent both opportunity and a challenge.

This is true for an emerging technology based on artificial intelligence that comes in the form of various

intelligent assistants and chatbots. One of the latest revolutions in this domain is artificial intelligence-based

tool ChatGPT. In this paper, opportunities and challenges of using ChatGPT and other virtual assistants and

chatbots are presented and discussed. Research results about using ChatGPT in higher education by students

is also given and elaborated.

1 INTRODUCTION

Rapid development of different technology has

resulted in various opportunities but also in different

challenges in higher education. On one side new

technology enables better visualization, more

effective performance, higher level of

automatization, quicker information retrieval, etc. On

the other side there are many challenges for educators

since the same technology can be used by students to

solve their tasks quicker and easier with the aid of

artificial intelligence, rather than using one’s own

knowledge and skills. This can consequently result in

students not understanding particular concepts and

processes. Based on this, it can be said that a special

set of challenges for educators has emerged with the

more applicable artificial intelligence tools, such as

are various virtual assistants and chatbots.

Mentioned artificial intelligence-based tools can

be of great value when used by professionals, and can

also benefit students when used properly. However,

if students opt for this it is possible to replace a

significant amount of effort by work results provided

by artificial intelligence (Cotton et al., 2023). In this

way students can skip some important steps in their

education which consequently leads to lower levels of

professional capabilities.

Virtual assistants and chatbots are rather new

addition to higher education in more widely used

capacity. Students tend to adopt new technologies

rather quickly, and their educators have to do the

same. This poses a challenge for educators since they

must incorporate this into their in most cases already

busy schedule.

Artificial intelligence in higher education opens

many questions for educators, such as:

 How should artificial intelligence-based tools

be used?

 What artificial intelligence-based tools should

be used for?

 How can artificial intelligence-based tools be

used to support education?

 How can unwanted uses of artificial

intelligence-based tools be prevented?

 What teaching models should be used when

considering artificial intelligence in education?

 How should tasks and assignments be modified

in order to utilize artificial intelligence and

possibilities of newly developed tools in this

domain?

 How should student work be checked to detect

authentical or artificial intelligence-based work

results?

As artificial intelligence enters higher education

educators will have to put a lot of effort into adapting

to this new paradigm and providing students with

adapted teaching models that will utilize new

possibilities while preserving genuine education

efforts.

136

Konecki, M., Konecki, M. and Biškupi

c, I.

Using Artiﬁcial Intelligence in Higher Education.

DOI: 10.5220/0012039700003470

In Proceedings of the 15th International Conference on Computer Supported Education (CSEDU 2023) - Volume 2, pages 136-141

ISBN: 978-989-758-641-5; ISSN: 2184-5026

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

2 ARTIFICIAL INTELLIGENCE

IN EDUCATION

The rapid development of technologies affects the

transformation of education in increasing capacity. In

this process many new aspects have emerged, such as

more rapid introduction of distance learning with and

without digital support, e-learning in different

development stages (1.0, 2.0 and semantic one), etc.

Along with new aspects, a specific set of challenges

has also been identified. Some of these challenges are

already known, but many challenges are new and

present something that is yet to be researched and

resolved. One such set of challenges results from

recent introduction of artificial intelligence in higher

education that includes assistive tools in the form of

various virtual assistants and chatbots, such as

ChatGPT (Kung et al., 2022).

Web technology has allowed learners to

communicate and access information and has enabled

meaningful interactivity through Web 2.0

technologies that empowered e-learning students

transforming them from recipients into active and

social contributors.

The next generation of online education is in high

capacity dictated by the academic need to personalise

learning together with the efficient automation

offered by artificial intelligence (Montebello, 2018)

and learning analytics. According to Montebello

(2018) the use and integration of artificial intelligence

in every aspect of e-learning model is crucial due to

its ability to capture the learners’ individual

characteristics and use them to personalise the e-

learning delivery while addressing typical e-learning

issues of isolation, motivation, and impersonal

environment.

Personalized learning for students can be

supported by 24/7 online feedback system which can

be quite demanding and expensive whether analytic

systems is used, or more teaching staff are hired.

Graesser and McDaniel (2017) suggest cost-effective,

and efficient virtual assistants/avatars and chatbots as

the interactive systems used to serve heterogeneous

roles to facilitate student learning, for example

intelligent tutoring system or mentor (Huff et al.,

2019).

Supported by LMS Moodle, Atif et al. (2021)

have developed and hosted a Pedagogical

Conversational Agent known as VIRTA (VIRTual

Agent) that simulates the teacher in providing support

in order to promote better understanding of a learning

unit and its requirements. Authors’ idea was to build

a character that students at universities would find

engaging, hence, they have created virtual

pedagogical agent as female character designed to be

employed in educational settings for instructional

purposes, to interact with learners using text-based

and audio-based communication (developed by using

Unity 3D game-building development environment).

The results implicate the benefits of presented

model as students continued to use VIRTA for

months after VIRTA was initially made which

indicates that students seek information when they

need it, and a small percentage of them found it

worthwhile to return to use it (Atife et al., 2021).

Tamayo at al. (2019) have implemented a virtual

assistant in the form of chatbot, a model called

EconBot that provides students with support in their

online learning of economics.

Diwan et al. (2023) have developed AI-based,

automatically generated learning content evaluated

from the aspects of grammar, semantic accuracy and

relevance, but even though the model had shown

promising results it has not been tested in the

classroom yet.

One of the novelties in the educational sector of

video games design is Codex developed by OpenAI.

Codex is a system that can autocomplete and generate

programming code by translating natural language to

a programming code. The system is proficient in more

than dozen programming languages (Python,

JavaScript, Go, Perl, PHP, Ruby, Swift, TypeScript,

and even Shell) and is able to interpret commands in

natural language and execute them on behalf of the

user (Chen, 2021). This opens the door for students

and developers to rely in a large extent on the

software instead their own knowledge and skills. The

same challenge comes to a design and art education

programmes due to a large extent of using OpenAI

Dall·E 2 for design of photos that never existed

(OpenAI Dall·E-2, 2022).

Journalism and Marketing students can also

benefit from using photos design by Artificial

intelligence since it can create impressive design, and

reduce costs, and skip the issues regarding publishing

rights that comes with publishing of real people. This

is possible because artificial intelligence can create

imaginary people without identity and in any city and

environment one can think of.

Educators face new challenges that come from the

usage of artificial intelligence in education almost

daily, and therefore it is challenging to analyse,

structure or suggest effective approaches and models

that can be used in education. The biggest challenge

lies in everyday machine and deep learning that

adapts artificial intelligence by enormous speed while

at the same time educators are slowly learning about

it and its effects.

Using Artiﬁcial Intelligence in Higher Education

137

3 ARTIFICIAL INTELLIGENCE

AND AI CONTENT

DETECTORS

Company OpenAI has become a worldwide market

leader in the context of artificial intelligence models

development. Their GPT-3 models can understand

and generate natural language suitable for different

purposes. But there are also many other tools based

on artificial intelligence.

Davinci is highly capable model intended for

applications requiring a lot of understanding of the

content, like summarization for a specific audience

and creative content generation explaining the

motives of characters, and Ada is the fastest. Curie is

capable for many nuanced tasks like sentiment

classification and summarization, and Babbage can

perform straightforward tasks and is efficient at

semantic search documents ranking and matching up

with search queries (OpenAI, 2023). Microsoft has

implemented OpenAI’s ChatGPT into its services,

such as Bing search engine, MS Office 365 and Azure

services (Microsoft, 2023). Google has recently

released Chatbot Bard AI (Google, 2023). The

announcement of the implementation of artificial

intelligence in the increasing number of applications

and tools suggests that the presence of artificial

intelligence is daily increasing in almost all segments

of usage. Many individuals and smaller companies

have also been developing artificial intelligence

solutions for years and they can be found in education

efforts for some time. Some authors even present

GPT as a co-author of their research paper (Thomas

et al., 2022; GPT et al., 2022).

Even though this topic requires more attention and

should be discussed more widely, what is equally

relevant and concerning is not being able to detect

student authorship in comparison to artificial

intelligence-based authorship. This represents quite a

challenge in education due to the need for very

sophisticated verification of the authenticity and

originality of the content.

In order to detect work that has resulted from

using artificial intelligence-based tools, such as

ChatGPT, several different approaches can be taken,

including manual analysis. Some of these approaches

are (Cotton et al., 2023):

 Look for patterns or irregularities in the

language: chatbots often have limited language

abilities and may produce text that is not quite

human-like, with repetitive phrases or words,

or with odd or inconsistent use of language.

Examining the language used in the work can

help to identify whether it was likely written by

a chatbot.

 Check for sources and citations: chatbots are

not capable of conducting original research or

producing new ideas, so work that has been

written by a chatbot is unlikely to include

proper citations or references to sources.

Examining the sources and citations in the

work can help to identify whether it was likely

written by a chatbot.

 Check for originality: chatbots are not capable

of producing original work, so work that has

been written by a chatbot is likely to be very

similar to existing sources. Checking the work

for originality, either through manual review or

using plagiarism detection tools, can help to

identify whether it was likely written by a

chatbot.

 Check for factual errors: while AI language

models can produce coherent text, they may not

always produce text that is factually accurate.

Checking the essay for factual errors or

inconsistencies could be an indication that the

text was generated by a machine.

 Check the grammar and spelling: human

writing may contain errors and mistakes, such

as typos or grammatical errors, while writing

generated by AI may be more error-free.

However, this can vary depending on the

quality of the AI language model and the input

data it was trained on.

 Use language analysis tools: some tools (e.g.

GPT-2 Output Detector Demo) are designed to

analyse the language used in written work and

to identify patterns or irregularities that might

indicate that the work was produced by a

chatbot.

 Finally, human writing tends to be more

contextually aware and responsive to the needs

of the audience, while writing generated by AI

may be more generic and less tailored to a

specific context. This can impact the

effectiveness and clarity of the writing.

There are several artificial intelligence content

detectors that can be used in higher education.

One of the most popular software for checking

similarities is Turnitin Originality. It has been

recently published that their software in development

can detect artificial intelligence-assisted writing and

artificial intelligence writing generated by tools such

as ChatGPT (Caren, 2022).

Another potential solution for detection of

artificial intelligence-based content has been

presented by Princeton student Edward Tian as

CSEDU 2023 - 15th International Conference on Computer Supported Education

138

GPTZero with built-ins for educators with one

possibility among others, to get a holistic score for

how much of the document is written by artificial

intelligence (GPTZero, 2023).

Some other tools for detection of artificial

intelligence-based content include (Originality.ai,

2023; OpenAI’s AI text classifier, 2023; Unicheck,

2023; Copyleaks AI Content Detector, 2023; Writer

AI Content Detector, 2023; Crossplag AI Content

Detector, 2023; Sapling AI Content Detector, 2023):

 Originality.ai

 OpenAI’s AI text classifier

 Unicheck’s Emma

 Copyleaks AI Content Detector

 Writer AI Content Detector

 Crossplag AI Content Detector

 Sapling AI Content Detector

 Etc.

For the purpose of this paper, and particularly to

benefit academic work with students, some of the

possibilities of artificial intelligence that students

regularly use have been analysed. One of the

frequently used ChatGPT usage options is creating a

written essay. In order to test how well different

artificial intelligence content detectors work a query

for the ChatGPT to create a short student essay on a

specific subject has been used. Created essay has been

checked by using artificial intelligence content

detector software to determine how efficient

detection of artificial intelligence as the author of the

text is.

There was no indication of such detection in the

first report made by current version of Turnitin

Originality (Figure 1). It should be mentioned that

current version will likely soon be updated with new

detection algorithms designed especially for artificial

intelligence-based content, according to

announcements (Turnitin, 2023).

Afterwards, the same essay was tested by

GPTZero, which successfully determined an artificial

intelligence authorship (Figure 2). OpenAI’s AI text

classifier has also successfully detected that text is

likely written by artificial intelligence (Figure 3).

Crossplag AI Content Detector has not detected

artificial intelligence writing (Figure 4), and Writer

AI Content Detector found artificial intelligence

content (Figure 5).

Obtained results have been tested in several

iterations with different artificial intelligence created

content, and consequent results have confirmed the

first-round data.

Figure 1: Turnitin Originality software.

Figure 2: GPTZero report.

Figure 3: OpenAI’s AI text classifier report.

Figure 4: Crossplag AI Content Detector.

Using Artiﬁcial Intelligence in Higher Education

139

Figure 5: Writer AI Content Detector.

It can be concluded that detecting of artificial

intelligence-based content will remain an ongoing

development and research effort and it is safe to

presume that all mentioned tools will get even better

and more accurate in time. However, even now there

are some good and usable options that can help

educators to detect artificial intelligence footprint in

created content.

An important step in future efforts of educators

will be finding the right way of supporting

development of students’ knowledge and skills. Some

authors suggest that the focus should be put on

improving students’ creativity and critical thinking

rather than general skills (Zhai, 2022). Artificial

intelligence has just begun to impact the world of

higher education in a more concrete and wider scale.

The challenges that will become more and more

prominent will affect both teachers and students.

More research efforts are needed in order to find the

best possible approach and best education methods to

benefit both teachers and students via new and

applicable education tools.

4 STUDENTS’ EXPERIENCE

WITH CHATGPT

In order to conclude about students’ experience with

ChatGPT, research was conducted by using designed

questionnaire. The goal of the research was to

determine how familiar are students with artificial

intelligence and ChatGPT, for which purpose they

use ChatGPT and whether they think artificial

intelligence-based tools should be a standard part of

their higher education.

All questionnaire answers were based on the

Likert scale ranging from “Strongly disagree” (1) to

“Strongly agree” (5). 67 information technology

students have participated in the research. Research

results are shown in Table 1.

Table 1: Research results.

Statement

𝒙̅

I am familiar with the

concepts of artificial

intelligence

3,67 1,35

I am familiar with the

ChatGPT software tool

3,69 1,37

I use ChatGPT in my

her education

2,25 1,29

I use ChatGPT to help

me understand the

learnin

content

2,24 1,43

I use ChatGPT to help

me solve assi

ned tasks

2,07 1,39

I think that ChatGPT

should be a standard tool

in hi

her education

3,21 1,47

Research results have shown that students are

fairly well informed about artificial intelligence and

ChatGPT. It has also been shown that artificial

intelligence-based tools have started to appear in

higher student population as an aiding mean of

choice. Since ChatGPT has fairly recently been

positioned as an artificial intelligence tool suitable for

wider use, it is understandable and expected that

majority of students still do not use this tool in great

capacity, but it is notable that this trend has started

and that students are starting to adopt this artificial

intelligence tools in their higher education activities.

It will be interesting to track the popularity of this

kind of aiding software in student population as a part

of future research efforts.

5 CONCLUSION

Higher education is an important driving force of

development since it produces new professionals who

are necessary for all types of further development.

Higher education has had its challenges for years,

along with opportunities that came with new

technology.

One of the most recent advancements in

technology is once again opportunity and challenge at

the same time. Artificial intelligence and various

tools that are based on it can benefit both

professionals and students. It can save time, raise

efficiency, make learning more engaging, etc.

However, at the same time it offers students the

possibility to skip important lessons and rely on the

results of artificial intelligence too much.

One of the important challenges of educators will

be to get to know the possibilities of artificial

intelligence in education, to find the right approaches

CSEDU 2023 - 15th International Conference on Computer Supported Education

140

and models that will utilize the possibilities and

benefits of artificial intelligence and at the same time

promote development of one’s own knowledge and

skills.

In this paper, opportunities and challenges of

using ChatGPT and other virtual assistants and

chatbots have been presented and discussed, along

with brief analysis of current artificial intelligence

content detectors and possibilities of detecting

artificial intelligence authorship.

Further and more extensive analysis of artificial

intelligence content detectors, trends and popularity

of artificial intelligence-based tools among students

and teachers, as well as analysing different

approaches and education models will be part of

future research efforts.

REFERENCES

Atif, A., Jha, M., Richards, D., & Bilgin, A. A. (2021).

Artificial Intelligence (AI)-enabled remote learning and

teaching using Pedagogical Conversational Agents and

Learning Analytics. In Intelligent systems and learning

data analytics in online education (pp. 3-29). Academic

Press.

Caren, C. (2022). AI writing: The challenge and

opportunity in front of education now. https://

www.turnitin.com/blog/ai-writing-the-challenge-and-

opportunity-in-front-of-education-now

Chen, Mark et al. (2021) Evaluating Large Language

Models Trained on Code. Cornell University

Chen, M et al. (2021). Evaluating large language models

trained on code. arXiv preprint arXiv:2107.03374.

https://arxiv.org/pdf/2107.03374.pdf,

https://doi.org/10.48550/arXiv.2107.03374

Cotton, D. R., Cotton, P. A., & Shipway, J. R. (2023).

Chatting and Cheating. Ensuring academic integrity in

the era of ChatGPT.

Diwan, C., Srinivasa, S., Suri, G., Agarwal, S., & Ram, P.

(2023). AI-based learning content generation and

learning pathway augmentation to increase learner

engagement. Computers and Education: Artificial

Intelligence, 4, 100110. https://doi.org/10.1016/

j.caeai.2022.100110

GPT, Thunström, A. O. & Steingrimsson, S. (2022)

Preprint at HAL. https://hal.science/hal-03701250

Graesser, A., & McDaniel, B. (2017). Conversational

agents can provide formative assessment, constructive

learning, and adaptive instruction. In The future of

assessment (pp. 85-112). Routledge.

Huff, E. W., Mack, N. A., Cummings, R., Womack, K.,

Gosha, K., & Gilbert, J. E. (2019). Evaluating the

usability of pervasive conversational user interfaces for

virtual mentoring. In Learning and Collaboration

Technologies. Ubiquitous and Virtual Environments

for Learning and Collaboration: 6th International

Conference, LCT 2019, Held as Part of the 21st HCI

International Conference, HCII 2019, Orlando, FL,

USA, July 26–31, 2019, Proceedings, Part II 21 (pp. 80-

98). Springer International Publishing.

Kung, T. H. et al. (2023). Performance of ChatGPT on

USMLE: Potential for AI-assisted medical education

using large language models. PLOS Digital Health,

2(2), e0000198.

Montebello, M. (2018). AI Injected e-Learning: The Future

of Online Education. Studies in Computational

Intelligence 745, Springer International Publishing

Tamayo, P. A., Herrero, A., Martín, J., Navarro, C., &

Tránchez, J. M. (2020). Design of a chatbot as a

distance learning assistant. Open Praxis, 12(1), 145-

153.

Thomas, I. S., Wang, J., & GPT-3 (2022) What Makes Us

Human: An Artificial Intelligence Answers Life's

Biggest Questions. Sounds True, Boulder.

Zhai, X. (2022). ChatGPT user experience: Implications for

education. Available at SSRN 4312418.

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4

312418

OpenAI Dall·E-2 (2022). DALL·E Now

Available Without Waitlist. https://openai.com/blog/

dall-e-now-available-without-waitlist/

OpenAI (2023). Models. https://platform.openai.

com/docs/models/overview

Microsoft (2023). Microsoft and OpenAI extend

partnership - The Official Microsoft Blog.

https://blogs.microsoft.com/blog/2023/01/23/microsoft

andopenaiextendpartnership/

Google (2023). An important next step on our AI journey.

https://blog.google/technology/ai/bard-google-ai-searc

h-updates/

GPTZero (2023). GPTZero. https://gptzero.me/

Originality.ai (2023). https://originality.ai/

OpenAI’s AI text classifier (2023). https://platform.

openai.com/ai-text-classifier

Unicheck (2023). https://unicheck.com/

Copyleaks AI Content Detector (2023). https://copyleaks.

com/features/ai-content-detector

Writer AI Content Detector (2023). https://writer.com/ai-

content-detector/

Crossplag AI Content Detector (2023). https://crossplag.

com/ai-content-detector/

Sapling AI Content Detector (2023). https://sapling.

ai/utilities/ai-content-detector

Turnitin (2023). Turnitin announces AI writing detector and

AI writing resource center for educators. https://www.

turnitin.com/press/turnitin-announces-ai-writing-detec

tor-and-ai-writing-resource-center-for-educators.

Using Artiﬁcial Intelligence in Higher Education

141