Integrating ChatGPT in a Computer Science Course: Students

Perceptions and Suggestions

Kehinde Aruleba a,∗ , Ismaila Temitayo Sanusi b,∗ , George Obaidoc and Blessing Ogbuokirid
a School of Computing and Mathematical Sciences, University of Leicester, Leicester, LE1 7RH, UK, [email protected]
b School of Computing, University of Eastern, Joensuu, P.O.Box 111, 80101, Finland [email protected]
c Center for Human-Compatible Artificial Intelligence (CHAI), Berkeley Institute for Data Science (BIDS), University of
arXiv:2402.01640v1 [cs.CY] 22 Dec 2023

California, Berkeley, 94720, California, USA [email protected]

d Department of Mathematics and Statistics, York University, Toronto, ON M3J 1P3, Canada [email protected]

ARTICLE INFO ABSTRACT

Keywords: The integration of artificial intelligence tools such as ChatGPT in the education system has
ChatGPT gained attention in recent years. This experience report explores students’ perceptions and
Artificial intelligence suggestions for integrating ChatGPT in a computer science course. Following a ChatGPT activity
Plugged activities which includes code completion and analysis, seven students participated in in-depth interviews.
Findings from the transcribed interviews suggest that ChatGPT has the potential to enhance
learning experience including programming. They highlighted the tool’s ability to respond
immediately to queries and supporting personalised learning. However, they raise concerns
that heavy reliance on ChatGPT may adversely affect students’ critical thinking and problem-
solving skills. These findings show the importance of carefully balancing using ChatGPT
in computer science courses. The findings of this research have significant implications for
educators, curriculum designers and policymakers as they explore integrating AI tools into
educational contexts.

1. Introduction
Educational institutions have undergone transformative shifts in recent years by embracing digital technology to
support teaching and learning experiences (Aruleba, Jere and Matarirano, 2022a; Ayanwale, Sanusi, Adelana, Aruleba
and Oyelere, 2022; Aruleba, Jere and Matarirano, 2022b; Obaido, Agbo, Alvarado and Oyelere, 2023). This shift is
driven by various factors, including the rapid increase of online resources (Haleem, Javaid, Qadri and Suman, 2022), the
need for more accessible and flexible learning experiences (Yunusa, Sanusi, Dada, Oyelere, Agbo, Obaido and Aruleba,
2021; Criollo-C, Guerrero-Arias, Jaramillo-Alcázar and Luján-Mora, 2021), and the need to equip students with the
knowledge and skills needed for the 21st century (Agbo, Sanusi, Oyelere and Suhonen, 2021; Sanusi, Olaleye, Oyelere
and Dixon, 2022). Artificial intelligence (AI) systems have powered most of these shifts. AI-driven technologies
and tools such as intelligent tutoring systems and recommendation systems all offer personalised experiences and
support for students, administrators and educators (Xia, Chiu, Lee, Sanusi, Dai and Chai, 2022; Sanusi, Ayanwale and
Chiu, 2023). One area where these technologies have significantly impacted education is the development of chatbots
including generative AI (GenAI) systems which has shown to support students learning (Deriba, Sanusi and Sunday,
2023).
GenAI tools such ChatGPT are now becoming increasingly common, and they are being used across various
sectors. Research has established ChatGPT’s capabilities to offer significant gains across different sectors including
education (Dwivedi, Kshetri, Hughes, Slade, Jeyaraj, Kar, Baabdullah, Koohang, Raghavan, Ahuja et al., 2023;
Shoufan, 2023). The launch of ChatGPT which UNESCO (UNESCO, Sept 8, 2023) tagged the fastest growing app
in history popularized the GenAI concept among the general public. With ChatGPT’s distinct features and capability
of discussing any topic of interest, it is currently one of the most powerful AI applications. As a chatbot that engage
in convincing conversations with users, it is used to perform various tasks which includes essay writing, conducting
literature reviews, enhancing papers, and writing computer code (Owens Brian, Feb 20, 2023). Eva A.M. van Dis
(Feb 22, 2023) asserts that the abilities of ChatGPT are expected to rapidly expand as it continues to receive new data
through user interaction. These functionalities contribute to the contention of its role in education.
∗ Corresponding authors. [email protected] (K. Aruleba), [email protected] (I. Sanusi)
ORCID (s):

First Author et al.: Preprint submitted to Elsevier Page 1 of 13

 Integrating ChatGPT in a Computer Science Course: Students Perceptions and Suggestions

While this AI tool could be used to enhance and support learning, its rapid growth and adoption raises integrity
concerns such as cheating and plagiarism (Deriba et al., 2023). Educators are mostly concerned about ChatGPT’s
capabilities to produce an essay or coursework that will pass assessment or exams (Leinonen, Denny, MacNeil, Sarsa,
Bernstein, Kim, Tran and Hellas, 2023). While this concern is recognizable, the only way to address it is for educators
to explore approaches that can be used to integrate the tool in classroom practices. As ChatGPT, a free and web-based
tool which represents a tipping point in the development of AI (Vaughan Connolly, April 5, 2023) come to stay in the
educational system, we must figure out how to incorporate ChatGPT safely, effectively, and appropriately in teaching
practices. Hence, this study is interested in how ChatGPT can be integrated into teaching and learning in a computer
science (CS) course. We further explore the efficacy of using ChatGPT to learn programming in higher education
institution (HEI). We recognise that past research has explored the possibility of learning various concepts including
programming with ChatGPT (Leinonen et al., 2023; Ouh, Gan, Shim and Wlodkowski, 2023). However, more work
is required, particularly that newer version of the GenAI platform keeps unfolding. More importantly is the fact that
extensive research on the subject matter will assist education stakeholders to uncover strategies to effectively integrate
ChatGPT in CS programs at the postsecondary level.
This study is expected to provide insight into how ChatGPT can be incorporated into teaching and learning
computing in HEIs classroom especially computing education. We anticipate that the outcome of this research will
advance computer science education community’s knowledge among other field on the use of conversational chatbots
in facilitating learning. This paper is organised as follows. Having highlighted the purpose of this study in Section 1,
we discuss related work with respect to students perception of ChatGPT and its use for learning programming in
Section 2. In Section 3, we describe our methods and activities given to participants and the follow-up interviews. In
Section 4, we present our key findings from the interview and the student activities. We discussed our findings and offer
recommendations to relevant education stakeholders in Section 6 and concluded with Limitations and future research
directions.

2. Related works
2.1. Student perception of ChatGPT in higher education
The emergence of ChatGPT holds profound implications for teaching and learning, particularly in HEIs. Studies
examining the perceptions of stakeholders, especially students, within the context of various fields, including computer
science, have been identified. Shoufan (2023) conducted a two-stage analysis involving 56 senior students in a
computer engineering program, revealing their appreciation for ChatGPT’s capabilities in study and work. However,
concerns about accuracy and the prerequisite background knowledge exist, emphasizing the necessity for educators
to guide effective usage and advocate for improvements in model accuracy. Similarly, Bonsu and Baffour-Koduah
(2023) surveyed 107 students in Ghana, reporting positive perceptions and a willingness to adopt ChatGPT in
education, despite no statistical relationship between perception and intention to use. Firat (2023) explored ChatGPT’s
implications through the perspective of 21 scholars and students across four countries, identifying key themes such
as the evolution of learning systems and ethical considerations. The study advocates for further research on ethics,
privacy, and effective AI integration in education, concluding with an emphasis on understanding the opportunities
and challenges of AI in higher education.
Additionally, a study by Ngo (2023) investigated how 200 university students perceive the use of ChatGPT for
learning, revealing a generally positive view with benefits like time efficiency, diverse information access, personalized
tutoring, and enhanced writing ideas. Concerns center around source evaluation, accurate citation, and language
precision, proposing solutions such as verifying ChatGPT responses, utilizing it as a reference tool, providing
guidelines, and fostering academic integrity to ensure ethical application in academia. Hamid, Zulkifli, Naimat, Yaacob
and Ng (2023) explored the use of ChatGPT in process-driven problem-based learning (PDPBL) in the Bachelor of
Pharmacy program. Results indicate that ChatGPT improves group collaboration, engagement, and motivation during
PDPBL sessions. Despite challenges in understanding ChatGPT’s information, most students recognize its potential to
replace traditional information-seeking methods. The study suggests that ChatGPT can enhance PDPBL in pharmacy
education, emphasizing the need for further research to validate the information provided and assess its impact on a
larger sample size.
Meanwhile, Elkhodr, Gide, Wu and Darwish (2023) investigated ChatGPT’s role as a classroom aid in higher
education, with three case studies involving ICT students. The findings reveal a positive perception of ChatGPT
as a valuable learning resource, with students expressing a willingness to use AI tools in the future. The study

First Author et al.: Preprint submitted to Elsevier Page 2 of 13

 Integrating ChatGPT in a Computer Science Course: Students Perceptions and Suggestions

suggests enhanced functionality, user flow, and content comprehension among ChatGPT users compared to those
relying solely on traditional search engines. In Woithe and Filipec (2023), the discussion of ChatGPT’s dynamics
in higher education, contribution, and implications was considered. Employing a constructivist-interpretivist approach
with qualitative methods, the study uncovers factors influencing ChatGPT adoption, its post-adoption role, student
and educator perspectives, and psychosocial effects on users. The trichotomous model underscores informed decision-
making for integrating ChatGPT into education, balancing technical benefits with learner impacts. The study proposes
UTAUT2 model extensions, prompting avenues for additional research in this evolving landscape.

2.2. ChatGPT for teaching and learning programming

In addition to studies exploring education stakeholders’ perceptions of ChatGPT in language learning Bin-Hady,
Al-Kadi, Hazaea and Ali (2023); Kohnke, Moorhouse and Zou (2023), there is a growing focus on the AI chatbot’s role
in supporting programming education. Rahman and Watanobe (2023) delves into ChatGPT’s impact on education and
programming learning, examining personalized feedback and interactive teaching opportunities along with challenges
like potential cheating and diminished critical thinking skills. The research includes coding experiments and surveys,
providing insights into ChatGPT’s benefits and challenges in programming education. In Surameery and Shakor (2023),
ChatGPT’s use for debugging is explored, emphasizing its capabilities and limitations. The research recommends
integrating ChatGPT into a broader debugging toolkit to optimize bug identification and resolution. Opara, Mfon-
Ette Theresa and Aduke (2023) investigated the introduction of ChatGPT in education, evaluating its benefits and
challenges, including personalized learning experiences and task automation. Recommendations stress the importance
of citing and referencing ChatGPT’s responses in educational contexts.
Ouh et al. (2023) assess ChatGPT’s efficacy in generating solutions for coding exercises in a Java programming
course. Findings show its accuracy but note challenges with non-textual descriptions. Leinonen et al. (2023) evaluate
ChatGPT’s potential in producing code explanations, finding LLM-created explanations to be significantly easier
to understand and more accurate than student-created ones. Gottipati, Shim and Shankararaman (2023) address
challenges in developing advanced programming skills, using ChatGPT as a support tool, showing initial evidence of
its effectiveness. Banić, Konecki and Konecki (2023) explore pair programming with ChatGPT as an educational aiding
tool, offering insights into this approach. Chan and Hu (2023) survey students in Hong Kong, indicating enthusiasm
for integrating ChatGPT in teaching, acknowledging its promise, and voicing concerns. Yilmaz and Yilmaz (2023)
explores ChatGPT’s impact on programming education, showing positive outcomes in computational thinking skills
and motivation, and suggesting potential benefits for integration.
In summary, several studies indicate acceptance among students and teachers for the use or inclusion of ChatGPT
as an aid to teaching and learning. This aligns with the purpose of this study, serving as an addition to knowledge with
a distinct approach.

3. Methodology
This qualitative study aims to understand students’ perception of integrating ChatGPT into the Computer Science
curriculum. The study involved recruiting seven year 2 Computer Science students to complete some programming and
database activities. After completing and submitting their attempts, they were then interviewed. All the interactions
were done remotely due to the semester holidays.

3.1. Recruitment
In this study, we employed quota sampling to select participants from the pool of year 2 Computer Science students.
Our sample covers students from diverse demographics and academic abilities. The sample consisted of two female
and five male participants, reflecting a gender-balanced representation within the cohort. Additionally, participants
exhibited diversity in ethnic backgrounds, with one Black, four Asian, and two White participants. The authors made
an effort to ensure inclusivity by actively recruiting participants with disabilities. Two out of the seven participants in
the sample reported having disabilities, enhancing the study’s ability to explore the experiences and perspectives of
students with diverse abilities within the context of computing education. All participants consented to participate in
the study. As the study was conducted remotely, firstly, we individually contacted and gave participants a brief about
the research and shared questions and instructions about the activities.

First Author et al.: Preprint submitted to Elsevier Page 3 of 13

 Integrating ChatGPT in a Computer Science Course: Students Perceptions and Suggestions

3.2. Procedure
Due to the semester holiday, we developed a remote engagement protocol for the activities and interview (see
Section 3.5 and 3.4). All the interaction and communications happened via email and Microsoft Teams call as it was
the most convenient form of conversation for the participants. A brief email was sent to inquire about participants’
availability for the study and describe the activities involved in the study and their expectations of the interview. In case
of any concern, we provided participants the option to cancel or reschedule the call at any time during the study. The
activities section of the study was divided into two parts (see Section 3.4), and the interview protocol was divided into
five parts (see Section 3.5). Ethical approval was obtained before initiating the study. Informed consent was obtained
from each participant, ensuring confidentiality, anonymity, and the right to withdraw at any stage.

3.3. Data Collection and Analysis

The data retrieved in this study were gathered through the students’ artefacts, completed learning tasks, and
cognitive interviews. Participants were firstly told to submit screenshots of each of their codes alongside the code
generated by ChatGPT, including the prompts given to ChatGPT. Except for where the authors already provided Java
code snippets, all the activities and feedback submitted by students were written in different programming languages
due to the instruction that allows students to use a language of their choice.
Secondly, we followed up with participants a week after submitting the answers to the activities for interview
requests. We conducted Microsoft Teams based structured interviews about ChatGPT. The interview duration ranges
from 18 - 30 minutes. In total, we collected seven individual feedbacks on the activities and 184 minutes of interview
data. The interviews were recorded after obtaining consent and later transcribed. Subsequently, qualitative coding and
analysis were done. Based on the grounded theory approach, codes were developed based on emergent themes from
the interviews (Constantinou, Georgiou and Perdikogianni, 2017). These codes were refined through several coding
iterations until theoretical saturation was achieved. A simultaneous literature review process also informed the thematic
analysis. Two authors of this study were responsible for conducting all qualitative coding and analysis.
Throughout the process, a colleague provided iterative feedback on the coding and analysis. The initial step involved
reviewing our transcribed data several times to internalise the various interviewee views and accounts. All authors
spent several weeks interacting with the data, which helped to establish their credibility. The authors combined the
overlapping codes and eliminated the duplicates before carefully selecting the final codes. All of the authors actively
participated in several rounds of peer debriefing (Houghton, Casey, Shaw and Murphy, 2013) to settle significant
differences that arose during this procedure.

3.4. ChatGPT Activities

To understand the possibilities of integrating ChatGPT in a Computer Science course, we designed activities with
clear instructions for participants. These tasks were distributed to participants via email during a holiday period,
allowing for flexibility in completion time. Participants were told to attempt the questions independently and also
use ChatGPT to solve the questions after attempting them; this will be useful for code comparison, i.e. comparing
human code vs ChatGPT-generated code. They were further instructed to take screenshots of their code and that of
ChatGPT. The activities were divided into two categories, i.e., code completion and code analysis.

3.4.1. Code Completion

Participants were given a series of code completion tasks covering various programming languages and concepts,
including user input/output, SQL queries, and array manipulation. This task assessed participants’ ability to complete
code snippets and solve programming challenges. Some tasks require them to write the code from the beginning, and
others are partial code snippets with missing sections, and participants are required to solve the problem by completing
the code. The questions were carefully selected to cover a range of complexity levels, i.e., from basic to intermediate.
Each participant was presented with a total of five code completion tasks. Participants were explicitly informed not
to get help or support from any external source (such as Internet and textbooks) when attempting the tasks. However,
they were encouraged to think critically and attempt all questions independently.

3.4.2. Code analysis

Participants were given two codes and instructed to analyse them based on specific criteria. The criteria included
identifying potential bugs, suggesting code optimisations and solutions, and explaining the overall functionality of the
code. The code snippets were written in Java and administered individually to each participant. The tasks included

First Author et al.: Preprint submitted to Elsevier Page 4 of 13

 Integrating ChatGPT in a Computer Science Course: Students Perceptions and Suggestions

analysing output in a given string manipulation code and identifying and addressing potential issues in a given
calculator class, including error handling and data loss concerns.
Responses from participants, including code snippets and explanations, were collected via a secure cloud platform.
The data included screenshots of ChatGPT and participants’ code solutions and PDF documents containing their
explanations and analyses. No personal identifying information was collected to ensure anonymity.

3.5. Participants and Interview

This study was conducted in the context of a Computer Science course. We specifically engage our participants
with programming and database activities for the study duration. Students were interviewed with a semi-structured
interview procedure. These students have all completed and passed at least a programming and database course in the
past; this is expected to provide more indepth insights into how ChatGPT can be used in coding, hence contributing
additional critical perspective. The interview questions were divided into five different categories. This was done to
enhance the quality and effectiveness of the research process, and also to systematically explore different aspects of
the participants’ experiences and opinions. The categorisation aimed to provide structure to the interview process and
maintain focus during the interview. The categories are defined as follows:

• Activities Questions: Participants were asked to compare their code with ChatGPT-generated code, focusing on
structure, readability, and syntax. Creativity and innovation in the AI-generated code were explored, along with
strengths and weaknesses in both.
• General Questions: The broader implications of ChatGPT in education were discussed. Participants shared their
perspectives on how ChatGPT could assist educators, its integration into existing educational platforms, support
for students with disabilities, potential drawbacks, and the necessary preparation for ChatGPT classroom use.

• Ethical Questions: Participants reflected on the ethical considerations surrounding ChatGPT, expressing their
views on its impact on teaching, learning, and society. Opinions on students’ use of ChatGPT for learning tasks
were also discussed.
• Evaluating Questions (after the activities): Participants provided their overall impressions of ChatGPT and
discussed any concerns regarding its ethical implications that may have arisen during the comparison activities.

• Wrap-up Questions: The interview concluded with participants sharing insights gained from the study and their
plans for applying these insights to their coding practices or future projects.

The interviews were conducted online by one of the article’s authors, who has prior experience teaching a
programming course to the participants. This approach aimed to establish a level of familiarity and trust between
the interviewer and participants, creating an environment conducive to open and honest responses. The interviewer’s
teaching background also provided valuable context for understanding the participants’ perspectives within the context
of a programming course. However, efforts were made to minimize potential biases by maintaining a neutral stance
during the interview. The interviews were scheduled and conducted individually to ensure focused and undivided
attention.

4. Findings
Four themes were generated from the qualitative coding and analysis. These themes were further divided into
sub-themes. The themes are student perception on using ChatGPT for coding, student suggestions in incorporating
ChatGPT in teaching and learning, ethics and student general perception.

4.1. Student Perception on using ChatGPT for coding

Despite our interview, students perceived a range of benefits and drawbacks in using ChatGPT for coding. Some
of these are explained below.

First Author et al.: Preprint submitted to Elsevier Page 5 of 13

 Integrating ChatGPT in a Computer Science Course: Students Perceptions and Suggestions

4.1.1. Code structure

We asked participants if they noticed any differences between the structure of their code and that of ChatGPT;
most participants described the structure of ChatGPT-generated codes as straightforward and easy to follow, even for
novice programmers. P6, for example, said, “in terms of readability, ChatGPT would like split it up. So, for example,
I remember the first question when it was taking first and last name; it did first name and last name on different lines
and it made it more concise question for the user to answer.” Similarly, P1 said, “in one of the tasks, I have created
an extra variable to do the calculations. However, ChatGPT generated code combined it in one variable, and I am
shocked; this made the code easier and shorter than mine.”
Consistent with findings from Khoury, Avila, Brunelle and Camara (2023), one of the participants noted that
ChatGPT codes can be “generic and insure” (P3), and there is not much creativity in there. Despite this comment from
P3, other participants, such as P1 and P6 believed thet ChatGPT’s code could be innovative and improved over time,
showing the potential for ChatGPT to introduce creative approaches to coding tasks. Overall, the students’ perceptions
of using ChatGPT for their coding activities were generally positive. These findings suggest that this tool is promising
and could assist and enhance coding tasks while also challenging students to think creatively about problem-solving.

4.1.2. Code comment and explanation

The majority of the participants reported that ChatGPT-generated codes are well commented and believed that
this can potentially improve student programming skills. P1 stated that “ChatGPT provided detailed comments in a
straightforward approach.” Multiple participants also shared this sentiment, indicating that ChatGPT’s code comments
are generally perceived as helpful in learning. More specifically, participants emphasised and appreciated that these
comments were generated automatically without prompting ChatGPT to add comments to the code (see Figure 1). Five
participants highlighted this feature of ChatGPT. P1 noted that even though they are a “very good programmer, I rarely
add comments to my code even when I am asked to; I just add two or three words as comments.”
The result suggested mixed reactions when asked if they think ChatGPT default comments could be better if they
prompted it to add comments to the code. P3 said that could lead to “confusion for the user because ChatGPT comments
explain even the simplest things (variables), which makes the comments too much.” Similarly, P5 also raised concerns
about the accuracy of the comments “if ChatGPT is instructed to add comment, it will go too in-depth and might not
be correct at the end". These concerns indicate that while comprehensive comments can be helpful, they must also
be accurate to maintain meaning. P7 noted that despite their practice of adding comments to their code, they believed
that their comments might not match ChatGPT standards. This indicates that some participants perceived ChatGPT
comments to be of higher quality than their own. The differences between user and AI-generated comments show the
importance of aligning the AI’s commenting technique with user code practices and preferences.
Some participants, such as P4 perceived ChatGPT’s comments as more than just explanations – they viewed them
as educational tools that could facilitate and support learning programming. P3, P6, P7 shared a similar view - they
believe the comments are like explanations and descriptive enough, which will help them learn programming quickly.
ChatGPT’s comments have the potential to not only help in code understanding but also facilitate the learning of
programming concepts.

4.1.3. Self-learning of programming

Our findings showed that all participants agreed that tools like ChatGPT can significantly enhance self-learning
in programming. P4 stated, "ChatGPT’s ability to instantly respond to their queries has been instrumental in their
learning journey.” Four participants highlighted the tool’s potential to ease the challenges encountered by students
who struggle with programming. They emphasised that ChatGPT can make the lives of such students easier when
used appropriately. P7 noted that the feature of ChatGPT to simplify complex concepts and real-time assistance is
valuable for those encountering programming difficulties. Similarly to some of the findings in Yilmaz and Yilmaz
(2023), P2 said they had effectively used ChatGPT to “accelerate their programming learning journey.” According to
P7, ChatGPT has the ability to empower students in their creative pursuits, “it is a tool that offers guidance, helping
students tap into their creativity and innovation.”

4.2. Student suggestions in incorporating ChatGPT in teaching and learning

All the participants highlighted that ChatGPT is a highly effective tool for supporting their learning process. More
specifically, they noted that ChatGPT has the potential to assist lecturers in the development of learning materials and
assessments.

First Author et al.: Preprint submitted to Elsevier Page 6 of 13

 Integrating ChatGPT in a Computer Science Course: Students Perceptions and Suggestions

Figure 1: P6 - ChatGPT generated code with comments

4.2.1. Teaching
The findings revealed suggestions from participants on how ChatGPT can be used to help in teaching. All
participants pointed out that lecturers can use ChatGPT to generate various activities for students based on their
strengths. P1 said, “. . . during my internship as a teacher for A-level students, I used ChatGPT to generate tasks
and lab questions for some of the lessons, which was quite helpful.” P4 suggested that schools should offer a course
for students on how to use ChatGPT and its implications. This education and training would ensure that students can
harness the tool’s full potential.
Similarly, P5 said, “. . . universities can create resources to support our lecturers effectively using ChatGPT –
training, guidelines or best practices.” Some participants suggested using ChatGPT in the classroom as an alternative
to Courseware Platforms will provide more interactive and adaptive learning experiences. According to the participants,
the tool can be integrated into learning management systems (LMS) such as Blackboard and Canvas; they echoed that
adding it to the online forum section of LMS will greatly help lecturers. P6 highlighted that “when you ask questions
in online forums, you either wait for a fellow student or the lecturer to respond. I believe integrating ChatGPT will be
beneficial because it answers you immediately, and you do not have to wait for hours.”

4.2.2. Learning
All the participants stated that ChatGPT has the potential to help students learning process; this is consistent
with the findings in Shoufan (2023). A common theme among participants was ChatGPT’s diverse problem-solving
approaches. They expressed their excitement about the tool’s ability to solve a single question in multiple approaches;
these approaches enhance problem-solving and critical thinking skills by exposing students to different approaches
and methods. P7 noted that “. . . I can use the regenerate button several times to have multiple versions of a question.
Reading these different versions of answers to a single question will broaden my knowledge.” They all found ChatGPT
to be helpful in gaining a deeper understanding of the learning material. According to P2, “I prefer chatGPT to access
reading resources than traditional textbooks provided by lecturers.” This shift towards interactive and digital materials
is seen as a more flexible and engaging learning method for most participants.
ChatGPT could supplement lecturers’ explanations (P2, P3, P7). Due to the limited time allocated for lectures,
these participants said they used the tool to further understand concepts that were not explained in detail in class (see

First Author et al.: Preprint submitted to Elsevier Page 7 of 13

 Integrating ChatGPT in a Computer Science Course: Students Perceptions and Suggestions

Figure 2: P4 - ChatGPT explaining the output of a code

Figure 2). P1 also highlighted efficient code debugging, “. . . working on an assignment and having bugs or errors
that would take hours to go through Stackoverflow and all of these websites, ChatGPT will just help me out and really
explain what the problem was.” Despite the positives about ChatGPT, P3 raised concerns about the accuracy of the
tool responses. They feel the responses might sometimes be “. . . outdated and incorrect information”. This suggests
that the responses should be carefully reviewed and evaluated for assessments or learning tasks.

4.2.3. Student support

Most participants said they now prefer to get help from ChatGPT instead of search engines because they may
find the solution to the errors or bugs in their code using search engines but do not understand how this solution
works. However, with ChatGPT, it explains the errors and bugs as shown in Figure 3. P2 said, “. . . it is valuable
because you do not need to go through all Python documentation to know exactly what functions you need to use”.
Consistent with Fuchs (2023), P2 emphasises the importance of having an “. . . individual/customised personalised
feedback mechanism” and the significance of having access to an AI-based tutor system they can approach whenever
they need help (P4). All participants agreed that the tool will effectively support students with disabilities. P7 notably
stated that it will support those with dyslexia, and the tool’s flexibility in writing styles will be beneficial in addressing
students’ specific disability-related needs.
P3 acknowledge they have used ChatGPT in “. . . writing emails and grammar checks.” This feature was said to
support students learning to speak and write in English potentially. P1, P2, P3, and P7 stated that ChatGPT explains
code effectively, and this has stopped them from copying and pasting code they know little about from the internet.
Some of the participants raised concerns. P1 and P3 said that when the question is not constructed properly, the
responses from ChatGPT “. . . confused them.” Similar to Sok and Heng (2023), all participants mentioned concerns
about students becoming overreliant on the tool. This will reduce students’ critical thinking and problem-solving skills.
P2 and P6 stated that overreliance on ChatGPT could lead to reduced attention during lectures and encourage students
to think less when solving problems. It is essential that integrating ChatGPT in educational systems should aim to have
a balance between providing personalised support and promoting self-reliance.

First Author et al.: Preprint submitted to Elsevier Page 8 of 13

 Integrating ChatGPT in a Computer Science Course: Students Perceptions and Suggestions

Figure 3: P2 - Error identification and possible solutions explained by ChatGPT

4.3. Ethics
All the study participants are concerned about the ethical implications of using ChatGPT in the teaching and
learning process. While plagiarism appears to be the prominent issue which resonates with the concerns of different
stakeholders, especially in the academic (Anders, 2023) settings, other issues related to privacy, data protection and
misinformation were raised.

4.3.1. Plagiarism
Consistent with earlier concerns in literature (Anders, 2023; Cotton, Cotton and Shipway, 2023), the students
reported copying and pasting of content directly from the chatbot constitute a challenge for effective usage of the AI
for learning. The participants believe the AI app is now being used by students to generate resources and they take
credit for work that is not theirs. For instance, P4 stated that "one of the ethical implications is students taking credit
for work that is not theirs, which is a very bad practise in and outside of academic institutions." P5 also opined that
"...some people might claim ChatGPT output as their work." With the concerns about cheating in mind, P4 suggested
that "educational institutions should offer a short course on ChatGPT usage and as part of the training, students should
be prompted on how they would feel claiming credit for work that is not theirs."

4.3.2. Privacy
The study participants identified privacy as a major concern. This concern is linked to the storage of the users data
on the AI app. As stated by P2, "using chatGPT raises concerns because all of the users data is stored in it, whatever
you type in the chatbot." Perhaps some of the participants are not aware that new ways to manage personal data in
ChatGPT now exist OpenAI (April 25, 2023). According to OpenAI (April 25, 2023) it is possible as from April 2023
that end users’ to turn off chat history so that their data won’t be used to train the Chatbot models by default.

4.3.3. Data protection

Similar to privacy concern, two participants specifically mentioned data protection. In what appears to be two
sides of same coin with privacy principles, data protection to the participants is how to limit ChatGPT’s collection and

First Author et al.: Preprint submitted to Elsevier Page 9 of 13

 Integrating ChatGPT in a Computer Science Course: Students Perceptions and Suggestions

retention of personal information. P3 expressed his concern by mentioning that "... if the user passes a university data,
that will go directly into the database and anyone will be able to get that information."

4.4. Student general perception

From the participants interview data, we identified their general perception about ChatGPT which we categorized
as strengths and weaknesses.

4.4.1. ChatGPT strengths

With regards to the AI chatbot strength, the participants highlighted several positive impact of its use. The
notable take-aways from the students responses include: ChatGPT’s capability to support personal study plan, assisting
struggling learners, and usefulness for English language writing, especially non-native speakers.
According to p2, "... is very effective for personal study time because AI can provide things that are personal to you
and it is also available for you all the time." Some of the participants hold the believe that it assist in clarifying lessons
taught after classes. P3 opine that "ChatGPT will be a good follow-up tool after classroom sessions to enhance learning,
especially struggling learners." In line with existing research that ChatGPT promotes English language learning (Bin-
Hady et al., 2023; Kohnke et al., 2023), P5 stated that ChatGPT "is pretty accurate especially in correcting grammatical
mistakes and [he] usually uses it to try to see if mistake was made or not."
Overall, the participants rated the AI chatbot as "time saver unlike web searching, p4" with the capability to "give
real-time feedback to users prompt and customised learning materials" P6. There is also a general consensus that
ChatGPT is a valuable tool for learning and improving ones knowledge.

4.4.2. ChatGPT weaknesses

Despite the strengths attributed to ChatGPT usage and its effectiveness in improving learning, the students have
concerns regarding heavy reliance on the tool. The participants prominent considerations include: lessening of critical
thinking & problem solving capacity, negative effect on students creativity and responsibility, and misinformation.
Some participants expressed the opinion that the constant usage of ChatGPT to complete task or exercises may
limit learners problem solving skills. For instance P2 stated that "....just looking up code on ChatGPT and like
getting examples from that instead of you know trying out the code yourself would lessen your critical thinking and
your problem solving capacity." Relatedly, P4 reason that "relying on ChatGPT heavily will hinder the individual’s
creativity and innovation such that students will not be able to come up with fully authentic solutions or attempt
difficult problems." P6 also believe that continual usage will impact significantly on ones responsibility. The participant
specifically described what he meant by responsibility in this context as "the sense that you [users] are not responsible
for it [the content] if ChatGPT generated it." Participant 1 & 3’s responses suggested that without information
verification, misinformation can set in based on heavy reliance on ChatGPT.

5. Discussion
In recent years, the integration of artificial intelligence and natural language processing technologies into
educational settings has gained momentum. One significant application of these technologies is the incorporation of
chatbots, such as ChatGPT, into the curriculum of computer science courses. In this study, our aim was to explore
students’ perceptions and gather their suggestions regarding the integration of ChatGPT as an instructional tool in a
computer science course. We specifically investigated students’ perceptions and suggestions for integrating the GenAI
tool into computer science courses following ChatGPT coding activities. Through interviews, we gathered the thoughts
of seven participants involved in our study. The conversations with the study participants focused on their experiences
with coding using ChatGPT, ethical concerns, suggestions for integrating ChatGPT into their courses, and their general
opinions about its usage. Our findings suggest that students believe ChatGPT is a promising tool for learning, including
programming. Additionally, students suggested adopting a well-defined model for the incorporation of ChatGPT into
their courses. Furthermore, we identified from the students’ feedback that heavy reliance on ChatGPT may diminish
critical thinking and problem-solving skills. The ethical considerations raised by the participants include plagiarism,
privacy, and data protection.
Our findings align with previous studies that have examined the use of ChatGPT in educational settings. Similar
to the positive perceptions reported in studies exploring the use of ChatGPT for student support and engagement
(Shoufan, 2023; Bernabei, Colabianchi, Falegnami and Costantino, 2023), our participants appreciated the convenience

First Author et al.: Preprint submitted to Elsevier Page 10 of 13

 Integrating ChatGPT in a Computer Science Course: Students Perceptions and Suggestions

and accessibility of ChatGPT. This consistency suggests that ChatGPT can serve as a valuable resource to enhance
programming experiences by providing timely and accessible assistance outside of traditional classroom hours.
ChatGPT’s availability for on-demand support contributes to fostering autonomous learners, particularly in the
context of programming. Our students recognized ChatGPT’s role in enhancing their self-learning experiences within
programming. This finding aligns with the broader literature on self-directed learning (Yilmaz and Yilmaz, 2023;
Baskara, 2023), highlighting the potential of ChatGPT to empower learners to take ownership of their educational
journeys and foster a more independent and self-driven approach to programming education. Another advantage
identified in our study was ChatGPT’s utility in assisting with code restructuring. This aligns with previous research on
automated code refactoring tools, emphasizing the potential for ChatGPT systems to aid programmers in optimizing
and enhancing code quality (Cao, Li, Wen and Cheung, 2023). ChatGPT’s contributions to code restructuring align
with the broader goals of improving code maintainability and readability.
However, our study also uncovered challenges and concerns identified in prior research related to the integration
of GenAI in education. The inconsistencies in ChatGPT’s responses, as highlighted by our participants, have been
a common issue reported in previous studies (Joshi, Budhiraja, Dev, Kadia, Ataullah, Mitra, Kumar and Akolekar,
2023; Currie, 2023). These inconsistencies underscore the need for ongoing refinement and optimization of ChatGPT
responses to ensure accuracy and reliability, particularly in educational contexts where the quality of information
is critical. Such inconsistencies may reinforce misinformation, which can be counterproductive to learning because
students are less likely to identify these errors. Our study revealed a potential challenge associated with the integration
of ChatGPT: the potential reduction of critical thinking among students. This concern is consistent with research on
the potential drawbacks of excessive reliance on GenAI in education (Shoufan, 2023). Overreliance on ChatGPT for
code-related tasks may inadvertently hinder students’ abilities to engage in deep critical thinking and problem-solving,
essential skills in computer science. Such overreliance may negatively impact learning and lead to academic dishonesty
(Kasneci, Seßler, Küchemann, Bannert, Dementieva, Fischer, Gasser, Groh, Günnemann, Hüllermeier et al., 2023).
This concern highlights the importance of carefully designed integration strategies that strike a balance between the
use of AI tools and the traditional teaching methods employed by educators.
It is crucial to acknowledge the limitation related to data privacy concerns. While ChatGPT serves as a valuable
educational tool, its operation may involve the collection and retention of personal information. This limitation aligns
with the broader discourse on data privacy and security in ChatGPT (Wu, Duan and Ni, 2023). Striking a balance
between providing personalized learning experiences and safeguarding the privacy of students’ data is a paramount
challenge in the integration of ChatGPT in education. This limitation necessitates the implementation of robust data
protection measures, ensuring compliance with relevant regulations such as the GDPR in Europe or the Family
Educational Rights and Privacy Act (FERPA) in the United States. Educators and institutions must adopt stringent
policies and practices that limit ChatGPT’s collection and retention of personal information to only what is essential for
its educational functions. Moreover, transparent communication with students about data usage and privacy safeguards
is imperative to build trust and maintain ethical standards in AI-enhanced education.
In light of our participants’ feedback and suggestions, our study contributes to the ongoing discourse surrounding
the role of AI in education. Recommendations for customization, personalization, increased interactivity, and trans-
parency align with the evolving landscape of GenAI-enhanced educational technologies (Leinonen et al., 2023). These
suggestions not only reflect the desire for more tailored and engaging learning experiences but also emphasize the
importance of clear communication and guidance regarding the role and limitations of ChatGPT in the educational
process..

6. Limitation and Future work

We identified some limitations in our study. First, the coding activities that our participants completed were not
complex as reflected in their responses to our interviews. Designing complex task may generate more insight. Second,
this study did not consider the ChatGPT version adopted by each participants to complete the exercises. Future research
may explore how different versions of ChatGPT or other GenAI tools perform in completing coding tasks. Third, we
worked with only seven students in this study which account for the use of a qualitative method. Future research may
conduct similar study with relatively large sample and analyzing the outputs from human vs AI.

First Author et al.: Preprint submitted to Elsevier Page 11 of 13

 Integrating ChatGPT in a Computer Science Course: Students Perceptions and Suggestions

7. Acknowledgement
We thank the students who participated in this study.

References
Agbo, F.J., Sanusi, I.T., Oyelere, S.S., Suhonen, J., 2021. Application of virtual reality in computer science education: A systemic review based on
bibliometric and content analysis methods. Education Sciences 11, 142.
Anders, B.A., 2023. Is using chatgpt cheating, plagiarism, both, neither, or forward thinking? Patterns 4.
Aruleba, K., Jere, N., Matarirano, O., 2022a. An evaluation of technology adoption during remote teaching and learning at tertiary institution by
gender. IEEE Transactions on Computational Social Systems .
Aruleba, K., Jere, N., Matarirano, O., 2022b. Technology adoption readiness in disadvantaged universities during covid-19 pandemic in south africa.
International Journal of Higher Education 11, 172–180.
Ayanwale, M.A., Sanusi, I.T., Adelana, O.P., Aruleba, K.D., Oyelere, S.S., 2022. Teachers’ readiness and intention to teach artificial intelligence in
schools. Computers and Education: Artificial Intelligence 3, 100099.
Banić, B., Konecki, M., Konecki, M., 2023. Pair programming education aided by chatgpt, in: 2023 46th MIPRO ICT and Electronics Convention
(MIPRO), IEEE. pp. 911–915.
Baskara, F.R., 2023. The promises and pitfalls of using chat gpt for self-determined learning in higher education: An argumentative review, in:
Prosiding Seminar Nasional Fakultas Tarbiyah dan Ilmu Keguruan IAIM Sinjai, pp. 95–101.
Bernabei, M., Colabianchi, S., Falegnami, A., Costantino, F., 2023. Students’ use of large language models in engineering education: A case study
on technology acceptance, perceptions, efficacy, and detection chances. Computers and Education: Artificial Intelligence , 100172.
Bin-Hady, W.R.A., Al-Kadi, A., Hazaea, A., Ali, J.K.M., 2023. Exploring the dimensions of chatgpt in english language learning: A global
perspective. Library Hi Tech .
Bonsu, E., Baffour-Koduah, D., 2023. From the consumers’ side: Determining students’ perception and intention to use chatgptin ghanaian higher
education. Available at SSRN 4387107 .
Cao, J., Li, M., Wen, M., Cheung, S.c., 2023. A study on prompt design, advantages and limitations of chatgpt for deep learning program repair.
arXiv preprint arXiv:2304.08191 .
Chan, C.K.Y., Hu, W., 2023. Students’ voices on generative ai: Perceptions, benefits, and challenges in higher education. arXiv preprint
arXiv:2305.00290 .
Constantinou, C.S., Georgiou, M., Perdikogianni, M., 2017. A comparative method for themes saturation (comets) in qualitative interviews.
Qualitative research 17, 571–588.
Cotton, D.R., Cotton, P.A., Shipway, J.R., 2023. Chatting and cheating: Ensuring academic integrity in the era of chatgpt. Innovations in Education
and Teaching International , 1–12.
Criollo-C, S., Guerrero-Arias, A., Jaramillo-Alcázar, Á., Luján-Mora, S., 2021. Mobile learning technologies for education: Benefits and pending
issues. Applied Sciences 11, 4111.
Currie, G.M., 2023. Academic integrity and artificial intelligence: is chatgpt hype, hero or heresy?, in: Seminars in Nuclear Medicine, Elsevier.
Deriba, F.G., Sanusi, I.T., Sunday, A.O., 2023. Enhancing computer programming education using chatgpt: A mini review, in: In 23rd Koli Calling
International Conference on Computing Education Research (Koli Calling ’23), ACM.
Eva A.M. van Dis, Johan Bollen, W.Z.R.v.R.C.L.B.N., Feb 22, 2023. Chatgpt: Five priorities for research. URL: https://www.nature.com/
articles/d41586-023-00288-7.
Dwivedi, Y.K., Kshetri, N., Hughes, L., Slade, E.L., Jeyaraj, A., Kar, A.K., Baabdullah, A.M., Koohang, A., Raghavan, V., Ahuja, M., et al., 2023.
“so what if chatgpt wrote it?” multidisciplinary perspectives on opportunities, challenges and implications of generative conversational ai for
research, practice and policy. International Journal of Information Management 71, 102642.
Elkhodr, M., Gide, E., Wu, R., Darwish, O., 2023. Ict students’ perceptions towards chatgpt: An experimental reflective lab analysis. STEM
Education 3, 70–88.
Firat, M., 2023. What chatgpt means for universities: Perceptions of scholars and students. Journal of Applied Learning and Teaching 6.
Fuchs, K., 2023. Exploring the opportunities and challenges of nlp models in higher education: is chat gpt a blessing or a curse?, in: Frontiers in
Education, Frontiers. p. 1166682.
Gottipati, S., Shim, K.J., Shankararaman, V., 2023. Ai for connectivism learning-undergraduate students’ experiences of chatgpt in advanced
programming courses .
Haleem, A., Javaid, M., Qadri, M.A., Suman, R., 2022. Understanding the role of digital technologies in education: A review. Sustainable Operations
and Computers 3, 275–285.
Hamid, H., Zulkifli, K., Naimat, F., Yaacob, N.L.C., Ng, K.W., 2023. Exploratory study on student perception on the use of chat ai in process-driven
problem-based learning. Currents in Pharmacy Teaching and Learning .
Houghton, C., Casey, D., Shaw, D., Murphy, K., 2013. Rigour in qualitative case-study research. Nurse researcher 20.
Joshi, I., Budhiraja, R., Dev, H., Kadia, J., Ataullah, M.O., Mitra, S., Kumar, D., Akolekar, H.D., 2023. Chatgpt–a blessing or a curse for
undergraduate computer science students and instructors? arXiv preprint arXiv:2304.14993 .
Kasneci, E., Seßler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., Gasser, U., Groh, G., Günnemann, S., Hüllermeier, E., et al., 2023.
Chatgpt for good? on opportunities and challenges of large language models for education. Learning and individual differences 103, 102274.
Khoury, R., Avila, A.R., Brunelle, J., Camara, B.M., 2023. How secure is code generated by chatgpt? arXiv preprint arXiv:2304.09655 .
Kohnke, L., Moorhouse, B.L., Zou, D., 2023. Chatgpt for language teaching and learning. RELC Journal , 00336882231162868.
Leinonen, J., Denny, P., MacNeil, S., Sarsa, S., Bernstein, S., Kim, J., Tran, A., Hellas, A., 2023. Comparing code explanations created by students
and large language models. arXiv preprint arXiv:2304.03938 .

First Author et al.: Preprint submitted to Elsevier Page 12 of 13

 Integrating ChatGPT in a Computer Science Course: Students Perceptions and Suggestions

Ngo, T.T.A., 2023. The perception by university students of the use of chatgpt in education. International Journal of Emerging Technologies in
Learning (Online) 18, 4.
Obaido, G., Agbo, F.J., Alvarado, C., Oyelere, S.S., 2023. Analysis of attrition studies within the computer sciences. IEEE Access .
Opara, E., Mfon-Ette Theresa, A., Aduke, T.C., 2023. Chatgpt for teaching, learning and research: Prospects and challenges. Opara Emmanuel
Chinonso, Adalikwu Mfon-Ette Theresa, Tolorunleke Caroline Aduke (2023). ChatGPT for Teaching, Learning and Research: Prospects and
Challenges. Glob Acad J Humanit Soc Sci 5.
OpenAI, April 25, 2023. New ways to manage your data in chatgpt. URL: https://openai.com/blog/
new-ways-to-manage-your-data-in-chatgpt.
Ouh, E.L., Gan, B.K.S., Shim, K.J., Wlodkowski, S., 2023. Chatgpt, can you generate solutions for my coding exercises? an evaluation on its
effectiveness in an undergraduate java programming course. arXiv preprint arXiv:2305.13680 .
Owens Brian, N., Feb 20, 2023. How nature readers are using chatgpt. URL: https://www.nature.com/articles/d41586-023-00500-8.
Rahman, M.M., Watanobe, Y., 2023. Chatgpt for education and research: Opportunities, threats, and strategies. Applied Sciences 13, 5783.
Sanusi, I.T., Ayanwale, M.A., Chiu, T.K., 2023. Investigating the moderating effects of social good and confidence on teachers’ intention to prepare
school students for artificial intelligence education. Education and Information Technologies , 1–23.
Sanusi, I.T., Olaleye, S.A., Oyelere, S.S., Dixon, R.A., 2022. Investigating learners’ competencies for artificial intelligence education in an african
k-12 setting. Computers and Education Open 3, 100083.
Shoufan, A., 2023. Exploring students’ perceptions of chatgpt: Thematic analysis and follow-up survey. IEEE Access .
Sok, S., Heng, K., 2023. Chatgpt for education and research: A review of benefits and risks. Available at SSRN 4378735 .
Surameery, N.M.S., Shakor, M.Y., 2023. Use chat gpt to solve programming bugs. International Journal of Information Technology & Computer
Engineering (IJITC) ISSN: 2455-5290 3, 17–22.
UNESCO, Sept 8, 2023. Guidance for generative ai in education and research. URL: https://unesdoc.unesco.org/ark:/48223/
pf0000386693.
Vaughan Connolly, Steve Watson, U.o.C., April 5, 2023. Chatgpt (we need to talk). URL: https://www.cam.ac.uk/stories/
ChatGPT-and-education.
Woithe, J., Filipec, O., 2023. Understanding the adoption, perception, and learning impact of chatgpt in higher education: A qualitative exploratory
case study analyzing students’ perspectives and experiences with the ai-based large language model.
Wu, X., Duan, R., Ni, J., 2023. Unveiling security, privacy, and ethical concerns of chatgpt. arXiv preprint arXiv:2307.14192 .
Xia, Q., Chiu, T.K., Lee, M., Sanusi, I.T., Dai, Y., Chai, C.S., 2022. A self-determination theory (sdt) design approach for inclusive and diverse
artificial intelligence (ai) education. Computers & Education 189, 104582.
Yilmaz, R., Yilmaz, F.G.K., 2023. Augmented intelligence in programming learning: Examining student views on the use of chatgpt for programming
learning. Computers in Human Behavior: Artificial Humans 1, 100005.
Yunusa, A.A., Sanusi, I.T., Dada, O.A., Oyelere, S.S., Agbo, F.J., Obaido, G., Aruleba, K., 2021. The impact of the covid-19 pandemic on higher
education in nigeria: university lecturers’ perspectives. IjEDict-International Journal of Education and Development Using Information and
Communication Technology 17, 43–66.

First Author et al.: Preprint submitted to Elsevier Page 13 of 13