AI-Generated Programming Solutions: Impacts on Academic

Integrity and Good Practices

Chung Man Tang

, Vanessa S. C. Ng

, Henry M. F. Leung

and Joe C. H. Yuen

School of Computing and Information Science, Anglia Ruskin University, Cambridge CB1 1PT, U.K.

School of Engineering and Applied Science, Aston University, Birmingham B4 7ET, U.K.

Keywords: AI-Powered Chatbots, Academic Integrity, ChatGPT, Assessment, Educational Technology.

Abstract: As AI-powered chatbots, notably ChatGPT, gain widespread popularity, their integration into academic

settings raises concerns about preserving academic integrity. Students increasingly employ these chatbots to

generate answers for assignments and, notably, programming problems. Existing countermeasures, such as

plagiarism checkers equipped with AI writing detection capabilities, struggle to detect AI-generated computer

programs. To thoroughly examine this challenge, we conducted an experiment, presenting diverse

programming problems to ChatGPT. Alarming findings revealed its remarkable proficiency in generating

correct solutions across various topics, complexities of problems, and programming languages. To explore

the implications, we engaged a focus group of programming teachers, resulting in the identification of key

practices and strategies to respond to AI-generated work. These insights provide valuable guidance for

educators seeking to maintain integrity while adapting to the evolving role of AI in education.

1 INTRODUCTION

In recent years, AI-powered chatbots, particularly

exemplified by ChatGPT, a Large Language Model

(LLM), have surged in popularity. These intelligent

conversational agents have demonstrated remarkable

capabilities and the potential to enhance work

efficiency across various domains (Baidoo-Anu &

Ansah, 2023; Deng & Lin, 2022; Kalla & Smith,

2023). However, their widespread adoption has not

been without consequences, especially in the

academic realm (Hong, 2023; Rasul et al., 2023).

One of the striking developments is the utilization

of ChatGPT by students to facilitate their academic

work (Cotton, Cotton, & Shipway, 2023; Lo, 2023).

Beyond simple text-based assistance, students have

been found employing ChatGPT for generating

answers to assignments, essays, and more. In a rather

concerning trend, ChatGPT have even ventured into

the realm of computer programming, providing

solutions to a variety of programming problems

(Rahman & Watanobe, 2023; Surameery & Shakor,

https://orcid.org/0000-0003-2842-6196

https://orcid.org/0000-0002-2972-530X

https://orcid.org/0000-0002-7753-0136

https://orcid.org/0000-0001-9346-4987

2023). This has implications for academic integrity,

raising questions about how to combat academic

misconduct when AI-generated work becomes

indistinguishable from genuine student efforts

(Ventayen, 2023).

To tackle this pressing issue, some

countermeasures have emerged (Gao et al., 2022),

such as plagiarism checkers equipped with AI writing

detection capabilities (Turnitin, 2023). These tools

can detect and flag essays that appear to be generated

by AI. While they have been effective in handling

written assignments, they still fall short when it

comes to detecting AI-generated computer programs.

In light of these challenges, this paper embarks on

a comprehensive investigation to ascertain the extent

of ChatGPT’s proficiency in generating accurate

program solutions for programming assignments. To

achieve this, we collected a diverse set of

programming problems from textbooks available in

our university library, covering topics commonly

taught in elementary programming courses. Some

require code creation from scratch, while others

478

Tang, C., Ng, V., Leung, H. and Yuen, J.

AI-Generated Programming Solutions: Impacts on Academic Integrity and Good Practices.

DOI: 10.5220/0012563600003693

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 2, pages 478-485

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

demand the debugging or enhancement of provided

code. We also ensured diversity by soliciting

solutions for generic programming topics in widely

adopted languages like C/C++, Java, and Python, as

well as for language-specific problems.

Our findings were striking and, in some ways,

alarming. ChatGPT demonstrated the ability to

generate correct solutions for the majority of the

programming problems we presented, regardless of

the topic, language, coding requirements, or

complexity of the problem descriptions. To further

explore the implications and gather expert insights,

we presented our experimental process and results to

a focus group consisting of staff members who

specialize in teaching programming courses.

The consensus among the focus group members

was unequivocal: AI-generated programming

solutions pose a significant challenge to academic

integrity, and no easy solution exists to tackle this

issue within the specifications of programming

problems. While program descriptions alone couldn’t

sufficiently challenge ChatGPT, the focus group

engaged in an insightful discussion, resulting in the

identification of a list of good practices and strategies

to address the issue of AI-generated work.

In this paper, we will detail our experimental

methodology, present our findings, and discuss the

practical recommendations that have emerged from

the deliberations of our focus group. These

recommendations hold the potential to offer a way

forward in maintaining the integrity of programming

assignments while acknowledging the growing

influence of AI technologies in academia.

2 AI IN PROGRAM ASSESSMENT

In the ever-evolving landscape of science and

technology education, attaining proficiency in

programming skills stands as a paramount pursuit for

students in computer science and other fields.

Programming assignments and exercises play a

pivotal role in honing these skills, demanding

students to apply problem-solving techniques and

unleash their creativity (Krathwohl, 2002). They also

serve as vital assessment tasks for teachers to evaluate

students’ proficiency and provide constructive

feedback for improvement. However, this integral

aspect of learning programming poses challenges for

both students and teachers.

Students often grapple with intricate problems,

investing substantial time and effort in deciphering

and addressing programming challenges.

Simultaneously, teachers, burdened with assessing

numerous programming exercises and assignments,

encounter a time-consuming and potentially error-

prone process (Tang, Yu, & Poon, 2009). Traditional

methods, such as visually inspecting source code or

manually testing programs with predefined cases,

contribute significantly to the workload of

programming teachers.

Recognizing the imperative for efficiency and an

enriched learning experience, several universities

have embraced automated program assessment

systems over the past decades (Ala-Mutka, 2005;

Tang, Yu, & Poon, 2023). These systems not only

alleviate the workload for teachers but also provide

students with immediate feedback (Lee et al., 2018),

fostering a more engaging and interactive learning

environment.

A significant development in these automated

systems is the integration of plagiarism checkers (Ng,

Li, & Ngai, 2004; Yu, Poon, & Choy, 2006).

Plagiarism has historically been a concern, and these

checkers, designed to identify code similarity within

a batch of submissions, have proven effective in

curbing academic malpractice (Yu, Poon, & Choy,

2006).

Figure 1: Program source code generated by ChatGPT.

However, the landscape is evolving with the

advent of AI in recent years. A concerning trend is

AI-Generated Programming Solutions: Impacts on Academic Integrity and Good Practices

479

emerging, where students leverage AI tools to

effortlessly generate solutions for their programming

assignments. Figure 1 illustrates an instance where

ChatGPT swiftly generates source code for a task

like, “Write a C program that determines the type of

a triangle given the lengths of three sides.” The allure

of eliminating effort, with ChatGPT providing a

solution in seconds along with relevant comments,

poses a temptation for students to bypass the rigor of

genuine problem-solving.

Unlike conventional plagiarism checkers that

compare submitted code against the other

submissions in the pool or existing solutions, the

randomness inherent in AI-generated programs

complicates detection. Each generation may yield

different source code, rendering traditional methods

inadequate. While there have been advancements in

detecting AI-generated essays (Turnitin, 2023), the

lack of semantic meaning in program source code

impedes the application to function in the

programming domain.

As the academic community grapples with this

new challenge, it becomes imperative to explore

immediate solutions to address the current situation.

While researchers work towards identifying ways to

deter AI-generated programs, educators and

institutions must navigate the evolving landscape,

considering strategies to uphold academic integrity

and ensure a fair and rigorous evaluation of students’

programming abilities. More importantly, we should

explore the potential and power of AI in enhancing

the quality of teaching and learning, and assessments,

rather than merely considering it as a threat to

academic integrity (Rahman & Watanobe, 2023).

In conclusion, integrating AI tools into

programming education presents opportunities and

challenges (Meyer et al., 2023). Balancing the use of

AI for educational enhancement while safeguarding

the learning process integrity is crucial in adapting to

this technological shift.

3 ChatGPT EXPERIMENT

The experimental design we implemented comprises

a systematic evaluation of ChatGPT’s ability to

generate programming solutions. To commence the

experiment, we acquired a comprehensive set of

practice exercises from computer programming

textbooks, covering a wide range of elementary

programming topics. These exercises served as the

foundational elements for evaluating ChatGPT’s

performance. Our methodology involved inputting

these exercises into ChatGPT, guiding the model to

generate solution programs for each exercise in single

or multiple prominent programming languages:

C/C++, Java, and Python. This multilingual

evaluation was intentionally designed to assess the

model’s proficiency across varying programming

paradigms. After generating the solution programs,

we performed an exhaustive analysis to assess their

correctness. This scrutiny specifically targeted the

functional accuracy of the solution programs,

evaluating how well they aligned with the specified

requirements of the exercises. The findings were

subsequently presented to a focus group comprising

experienced programming teachers. Leveraging their

extensive expertise, we facilitated a thorough

discussion to gather valuable insights and

recommendations. This discourse, extending beyond

the evaluation of ChatGPT’s performance, included

proposals for refining the design of programming

problems and suggesting best practices for addressing

assessment challenges. In the subsequent subsections,

we delve into the specifics of our experiment design.

3.1 Exercise Acquisition

The exercises sourced from textbooks cover various

topics. Our selection criteria were twofold: exercises

had to produce tangible program source code

outcomes, and these outcomes needed to be

assessable (Tang, Yu, & Poon, 2023). Our staff

members, drawing on collective experience and

professional judgment, meticulously curated the

exercises to ensure their suitability and alignment

with specific learning outcomes (Biggs, 2003).

We carefully selected 50 programming exercises

from reputable textbooks (Deitel & Deitel, 2012 &

2016; Heathcote, 2017; Malik, 2013) with an

emphasis on diversity. Within this set, 30 exercises

presented concise problem descriptions, each focused

on a single programming problem within specific

topics. Covering fundamental concepts in elementary

programming courses, these exercises were evenly

distributed across four categories: fundamentals (e.g.,

variables, data types, operators, and expressions),

control structures (e.g., loops and conditionals), data

structures (e.g., arrays, lists, stacks, and queues), and

algorithms (e.g., searching, sorting, and recursion).

Importantly, these categories were language-

independent, allowing solutions in C/C++ (referred to

as C++ hereafter for simplicity), Java, and Python.

Each category comprised 5 exercises, totaling 20

exercises and 60 program solutions across three

languages. We also introduced two language-specific

categories: low-level memory manipulation (e.g.,

manual memory allocation/deallocation, pointers,

CSEDU 2024 - 16th International Conference on Computer Supported Education

480

value and reference semantics) exclusively designed

for C++, and Object-Oriented structures (e.g.,

classes, objects, inheritance, polymorphism, and

encapsulation) tailored for both C++ and Java. Each

language-specific category included 5 exercises,

contributing to a total of 10 exercises.

In addition to exercises with concise descriptions,

we introduced 10 exercises with more extensive

problem statements, addressing multiple problems

across various topics, primarily in the programming

languages of C++ and Java.

Beyond the aforementioned code-from-scratch

exercises, we included 10 exercises where the source

code was provided, requiring students to modify the

existing code. Notably, the source code for these

exercises was written in C++, chosen due to its

availability in our textbook collection.

Table 1: A summary of the selected exercises.

Exercise Type / Topic

Prog.

Lan

Ex.

Result.

Pro

Short exercises

–

Pro

rammin

basics C++ &

Java &

Python

5 15

–

Control structures 5 15

–

Data structures 5 15

–

lgorithms 5 15

–

emory manipulation C++ 5 5

–

O-O

tructures C++

Java 5 5

Lon

exercises C++ 8 8

Java 2 2

Modifying existing code C++ 10 10

Total 50 90

In summary, our exercise sampling strategy

underwent a thorough selection process, guided by

criteria ensuring assessability and alignment with

learning outcomes. The outcome is a diverse

collection of exercises designed to address various

programming concepts and languages. An overview

of the selected exercises, detailing topics, languages,

quantities for exercises and resulting programs, can

be found in Table 1, with comprehensive details for

each exercise included in Appendix.

3.2 Program Solution Generation

In the process of soliciting solutions from ChatGPT

for programming problems, we presented the

questions as chat queries. The specific version of

ChatGPT employed in this experiment was the latest

GPT-3.5. It’s worth noting that we consider the

results independent of the browser and platform used.

To preserve the authenticity of the exercises, we

directly incorporated most of them from the original

textbooks. Our modifications were limited to

essential changes, like adjusting references to

additional information. For instance, we altered

phrases such as “modify the program in Figure X...”

to “modify the following program...” and “use the

provided function in Example Y...” to “use the

provided function below...,” ensuring clarity while

maintaining the core content.

To guide ChatGPT in generating solutions in

various programming languages, we incorporated

specific instructions within the exercises. Language

directives such as “in C” or “Write a [Python]

program” were inserted to prompt the model’s

language-specific responses. For example, the

prompt “Write a Java program to calculate and print

a list of all prime numbers from 1 to 100” indicates

the addition of the programming language “Java” to

prompt the generation of a Java program. Although

ChatGPT has the capability to generate new

responses if the user is not satisfied, our experiment

specifically focused on analysing only the initial

response produced by ChatGPT.

Following the generation of responses (solutions),

we documented and presented them to two

programming teachers for evaluation. The assessment

process involved visually inspecting the source code

and, in certain cases, executing the programs for

testing. The assessment outcomes were categorized as

“correct” if the program met all requirements,

“partially correct” if it fulfilled major requirements

but wasn’t entirely correct, or “incorrect” if it failed

to satisfy major requirements. The judgments were

based on the professional expertise of the staff. In

cases where there was a discrepancy in the

assessment results between the two assessors, a

moderation process was implemented to reconcile

and establish a consensus.

3.3 Results

Our assessment results reveal that ChatGPT

demonstrates a notably high proficiency in generating

accurate program solutions, surpassing our initial

expectations. While we expected effectiveness in

simpler exercises, our findings reveal its competence

extends to more complex tasks.

Table 2 provides a concise summary of the

assessment results. The first column categorizes

exercises and topics. In the second column, the

number of solutions under assessment is displayed,

while the third to fifth columns detail the assessment

outcomes. Specifically, the column marked with “

”

signifies the count of correct solutions, the column

marked with “

” indicates the count of partially

correct solutions, and the column marked with “

”

AI-Generated Programming Solutions: Impacts on Academic Integrity and Good Practices

481

represents the count of incorrect solutions. For

detailed assessment outcomes of each individual

exercise, refer to the Appendix.

Table 2: Assessment results for the generated solutions.

Exercise T

e / To

ic Assessed

  

Short exercises

–

Programming basics 15 15 0 0

–

Control structures 15 15 0 0

–

Data structures 15 15 0 0

–

orithms 15 15 0 0

–

emory manipulation 5 5 0 0

–

O-O structures 5 5 0 0

Long exercises 10 6 4 0

Modifying existing code 10 7 3 0

Total 90 83 7 0

ote:  = correct;  =

artiall

correct;  = incorrect.

Remarkably, for all short exercises, ChatGPT

consistently produced correct solutions across all

programming languages. This suggests its capability

to generate solutions spans various programming

topics and languages.

For longer questions, ChatGPT achieved correct

solutions for 6 out of 10 exercises and partially

correct solutions for the remaining 4 exercises,

without generating any outright incorrect solutions.

This prompts an exploration into factors contributing

to ChatGPT’s generation of partially correct

solutions, revealing potential implications for

refining exercise design to pose challenges for the

model. Two identified causes include ChatGPT’s

tendency to provide a “best solution” based on its

knowledge, potentially overlooking nuanced details

in complex exercise requirements. Additionally, the

absence of prior knowledge to exercise requirements

poses a challenge, as some exercises rely on applying

knowledge gained in previous exercises or examples.

ChatGPT, lacking such context initially, responds

based on its existing knowledge, which may not align

with exercise expectations.

On the positive note, to rectify solutions that are

not entirely correct, students are required to engage in

a critical evaluation of AI-generated solutions. They

need to discern the factors contributing to the inability

to generate a completely correct solution and

subsequently provide additional or rectify missing

information to guide ChatGPT in producing accurate

solutions. This iterative process not only prompts

students to adopt a bug-fixing approach but also

cultivates higher-order thinking skills, representing a

substantial educational benefit (Krathwohl, 2002).

In conclusion, although ChatGPT exhibits

impressive capabilities in generating accurate

programs, our study highlights opportunities to

enhance exercises for better challenging the model.

Moreover, it emphasizes the pedagogical value for

students in critically evaluating and refining AI-

generated solutions.

4 DISCUSSIONS

The findings of our experiment were meticulously

presented to a focus group comprising four lecturers

and senior lecturers experienced in teaching computer

programming courses. The objective of this

comprehensive deliberation was to rigorously analyse

the outcomes of the experiment and explore their

potential implications on academic integrity and

assessment strategies within the field of computer

programming education.

To foster a purposeful exchange of ideas, we

initiated the discussion with structured questions

encompassing three key aspects:

1. Perception and Awareness:

• How do you perceive the role of AI,

specifically ChatGPT, in generating correct

solutions for programming exercises?

2. Challenges to Academic Integrity:

• In your opinion, what challenges does the

capability of AI in generating correct

solutions pose to academic integrity in

computer programming courses?

3. Impact on Assessment Strategies:

• How might the use of AI in generating correct

solutions influence your current assessment

strategies for programming assignments?

Due to space constraints, the fully transcribed

discussion summary, guided by the above questions,

couldn’t be fully included in this paper. Here are the

concise conclusions that emerged:

Participants unanimously agreed that AI

introduces challenges in discerning the authenticity of

program solutions. The impressive proficiency of

ChatGPT in generating solutions for intricate

exercises highlights the difficulty in crafting

exercises meant to challenge AI-generated solutions.

Nevertheless, the excessively “perfect” nature of AI-

generated solutions, particularly those utilizing

advanced practices and features, may serve as a

distinguishing factor from those of average

elementary programmers.

Acknowledging the proficiency of AI in

generating accurate solutions, participants expressed

heightened concerns about the potential exacerbation

of academic misconduct. The dependency on AI

inhibits the cultivation of essential problem-solving

skills, resulting in disparities in skill development

CSEDU 2024 - 16th International Conference on Computer Supported Education

482

among students. Assessments face accuracy

challenges, making it difficult to assess students’

genuine understanding and comprehension of

underlying concepts, thereby biasing the feedback

loop in facilitating effective learning. The widespread

use of AI may erode trust between teachers and

students, ultimately impacting the authenticity of

individual efforts.

In response to these challenges, participants

reached a consensus on the imperative to reassess

exercise design and assessment strategies to address

potential risks associated with the misuse of AI. This

shift in assessment focus emphasizes the need for a

deeper understanding of the problem-solving process,

necessitating comprehensive documentation and

explanation of students’ approaches. The inclusion of

coding interviews or presentations becomes crucial,

providing educators with opportunities to scrutinize

students’ genuine understanding of their solutions.

This dynamic approach ensures assessments not only

gauge correctness but also delve into the intricacies

of the process involved in arriving at a solution,

fostering active learning, and contributing to

assessment for learning.

Participants collectively acknowledged ChatGPT

as a robust tool with the inherent potential to enhance

the learning experience. For example, students can

utilize ChatGPT for immediate clarification on

programming concepts, syntax, and problem-solving

approaches. Additionally, ChatGPT’s availability

round the clock ensures students have continuous

access to assistance and information as needed. These

aspects echo the findings of Rahman & Watanobe

(2023) regarding threats and opportunities,

highlighting the potential for further exploration.

5 RECOMMENDATIONS

Based on in-depth focus group discussions, we offer

concise recommendations specifically centered on

exercise design practices and assessment strategies.

Broader topics such as curriculum design, policy, and

ethics are reserved for future exploration.

5.1 Exercise Design Practices

The focus group has outlined four essential exercise

design practices. First and foremost, they recommend

the creation of unique and open-ended problems. This

involves developing tailored programming exercises

that transcend generic responses by incorporating

real-world scenarios or specific requirements.

Utilizing open-ended questions prompts students to

provide comprehensive explanations and reasoning

alongside their code submissions. This approach

actively discourages reliance on AI code generation

and fosters a deeper understanding of programming

concepts, encouraging students to articulate the

intricate thought processes behind their code.

Secondly, they recommend adopting Test-Driven

Development (TDD) (Janzen & Saiedian, 2005). This

method prompts students to write tests before coding,

fostering critical thinking in problem-solving. TDD

guides students to address smaller problems first,

ensuring code correctness and cultivating a deep

understanding of the problem domain. This approach

nurtures an analytical problem-solving mindset,

steering students away from predetermined solutions

and reducing reliance on AI assistance.

The third recommendation involves the use of

versioned assignments. They suggest mandating the

use of version control systems, such as Git,

particularly for larger assignments. This entails

requiring students to submit their work incrementally

and commit their source code at different stages of

development. The purpose is to promote continuous

engagement with the artifact and make it challenging

for students to rely solely on AI-generated solutions

for the entirety of the assignment.

Lastly, peer collaboration is encouraged as a key

exercise design practice. This entails fostering

collaborative learning through group assignments in

pairs or small groups while maintaining a focus on

individual accountability. Such an environment

facilitates active discussions and the sharing of ideas,

code, and problem-solving approaches among

students. Beyond enriching the learning experience,

this collaborative approach acts as a deterrent to

copying solutions, cultivating a shared understanding

and responsibility among students.

5.2 Assessment Strategies

In addition to exercise design practices, the focus

group has introduced assessment strategies tailored to

the integration of AI. Foremost among these

strategies is the endorsement of live coding

assessments. This approach requires students to

actively solve programming problems during their lab

or practical classes, thereby showcasing their coding

proficiency and problem-solving skills to their

programming teachers. The inherent nature of live

assessments renders them inherently less susceptible

to plagiarism and AI assistance, particularly within a

closely monitored environment. This format also

serves as an optimal platform for teachers to provide

AI-Generated Programming Solutions: Impacts on Academic Integrity and Good Practices

483

immediate feedback, facilitating assessment for

learning efficiently.

Building upon the efficacy of live coding

assessments, the next recommended strategy involves

the incorporation of viva voce presentations. Verbal

explanations necessitate a profound understanding of

students’ programs. Even when presented with an AI-

generated solution, students must delve into an

intensive examination of the code to comprehend the

underlying operations and theories. This exemplifies

how AI can serve as a powerful learning tool.

Simultaneously, viva voce assessments empower

teachers to accurately measure the depth of a

student’s comprehension and evaluate their ability to

articulate and elucidate their thought processes.

In addition to the above strategies, we advocate

for the implementation of portfolio-based

assessments. Students are encouraged to curate and

maintain a coding portfolio throughout the course,

with assessments grounded in the progression and

improvement demonstrated within this evolving

portfolio. This multifaceted strategy adds an

additional layer of complexity, creating a more

challenging environment for students to rely on

external AI-generated solutions. This approach aligns

with our commitment to ensuring assessments

authentically reflect the unique learning journeys of

individual students.

Collectively, these practices and strategies are

designed to foster genuine assessment for learning,

discourage plagiarism, and ensure that assessments

accurately reflect the skills and understanding

acquired by students in programming courses.

6 CONCLUSIONS

The recent decades have witnessed a significant leap

in computer technology, with AI emerging as

transformative forces that permeate our daily lives.

This relentless progression signifies not merely a

passing trend but an irreversible evolution, poised to

reshape educational landscapes and assume a pivotal

role in the realm of teaching and learning.

While the integration of AI introduces a

conundrum of challenges, particularly in preserving

academic integrity, it also holds the promise of

enhancing educational experiences. This study,

delving into the success of AI in generating precise

solutions for programming exercises and

assignments, underscores the undeniable influence of

these technologies on academic pursuits. The

challenge lies not just in acknowledging this

influence but in navigating its intricacies and

mitigating potential pitfalls.

The insights gleaned from our focus group

discussions have culminated in a robust set of

recommendations designed to address the nuances of

AI-generated work. These guidelines serve as a

proactive response, offering strategies to maintain

academic integrity while harnessing the potential of

AI in educational settings.

Embracing the power of AI, rather than resisting

it, can herald a new era in education. By leveraging

the capabilities of these technologies, we can sculpt

an educational landscape that is not just adaptive but

transformative. The journey ahead involves a delicate

balance—navigating challenges while harnessing the

boundless potential of AI to foster a future of enriched

and innovative education. As we step into this

transformative era, it becomes imperative to stay

proactive, ensuring that AI becomes an ally in the

educational journey rather than a hindrance.

REFERENCES

Ala-Mutka, K.M.(2005). A survey of automated assessment

approaches for programming assignments. Computer

Science Education, 15(2), 83-102.

Baidoo-Anu, D., & Ansah, L. O. (2023). Education in the

era of generative artificial intelligence (AI):

Understanding the potential benefits of ChatGPT in

promoting teaching and learning. J. of AI, 7(1), 52-62.

Biggs, J. (2003). Aligning teaching for constructing

learning. Higher Education Academy, 1(4), 1-4.

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.

Deitel, P. & Deitel, H. (2012). Java how to program,

Pearson. 9

Edition.

Deitel, P. & Deitel, H. (2016). C How to program with an

introduction to C++, Pearson, 8

Edition.

Deng, J., & Lin, Y. (2022). The benefits and challenges of

ChatGPT: An overview. Frontiers in Computing and

Intelligent Systems, 2(2), 81-83.

Gao, C. A., Howard, F. M., Markov, N. S., Dyer, E. C.,

Ramesh, S., Luo, Y., & Pearson, A. T. (2022).

Comparing scientific abstracts generated by ChatGPT

to original abstracts using an artificial intelligence

output detector, plagiarism detector, and blinded human

reviewers. BioRxiv, 2022-12.

Heathcote, P. M. (2017). Learning to program in Python,

PG Online Limited, 1

Edition.

Hong, W. C. H. (2023). The impact of ChatGPT on foreign

language teaching and learning: opportunities in

education and research. Journal of Educational

Technology and Innovation, 5(1).

CSEDU 2024 - 16th International Conference on Computer Supported Education

484

Janzen, D., & Saiedian, H. (2005). Test-driven

development concepts, taxonomy, and future

direction. Computer, 38(9), 43-50.

Kalla, D., & Smith, N. (2023). Study and analysis of

ChatGPT and its impact on different fields of

study. International Journal of Innovative Science and

Research Technology, 8(3).

Krathwohl, D. R. (2002). A revision of Bloom’s taxonomy:

An overview. Theory into practice, 41(4), 212-218.

Lee, V. C., Yu, Y. T., Tang, C. M., Wong, T. L., & Poon,

C. K. (2018). ViDA: A virtual debugging advisor for

supporting learning in computer programming

courses. Journal of Computer Assisted Learning, 34(3),

243-258.

Lo, C. K. (2023). What is the impact of ChatGPT on

education? A rapid review of the literature. Education

Sciences, 13(4), 410.

Malik, D. S. (2013). C++ programming: from problem

analysis to program design, Cengage Learning, 6

Ed.

Meyer, J.G., Urbanowicz, R.J., Martin, P.C., O’Connor, K.,

Li, R., Peng, P.C., ... & Moore, J.H. (2023). ChatGPT

and large language models in academia: opportunities

and challenges. BioData Mining, 16(1), 20.

Ng, S. C., Li, T. S., & Ngai, H. S. (2004). Plagiarism

detection in programming assignments, in Murphy, D.,

Carr, R., Taylor, J., & Wong, T. M. (Eds), Distance

Education and Technology: Issues and Practice, Open

University of Hong Kong Press, 2004, pp. 366-377.

ISBN: 962-7707-47-3

Rahman, M. M., & Watanobe, Y. (2023). ChatGPT for

education and research: Opportunities, threats, and

strategies. Applied Sciences, 13(9), 5783.

Rasul, T., Nair, S., Kalendra, D., Robin, M., de Oliveira

Santini, F., Ladeira, W. J., ... & Heathcote, L. (2023).

The role of ChatGPT in higher education: Benefits,

challenges, and future research directions. Journal of

Applied Learning and Teaching, 6(1).

Surameery, N. M. S., & Shakor, M. Y. (2023). Use

ChatGPT to solve programming bugs. International

Journal of Information Technology & Computer

Engineering (IJITC) ISSN: 2455-5290, 3(01), 17-22.

Tang, C. M., Yu, Y. T., & Poon, C. K. (2009). An approach

towards automatic testing of student programs using

token patterns. In 17th International Conference on

Computers in Education, ICCE 2009 (pp. 188-190).

Tang, C. M., Yu, Y. T., & Poon, C. K. (2023). An

automated system with a versatile test oracle for

assessing student programs. Computer Applications in

Engineering Education, 31(1), 176-199.

Turnitin. (2023, November 11). Turnitin’s AI detector

capabilities.https://www.turnitin.com/solutions/topics/

ai-writing/ai-detector/

Ventayen, R. J. M. (2023). OpenAI ChatGPT generated

results: Similarity index of artificial intelligence-based

contents. Available at SSRN 4332664.

Yu, Y. T., Poon, C. K., & Choy, M. (2006, October).

Experiences with PASS: Developing and using a

programming assignment assessment system. In 2006

Sixth International Conference on Quality Software

(QSIC’06) (pp. 360-368). IEEE.

APPENDIX

A list of selected exercises from textbooks.

Ex.

No.

Exercise Description

Text-

ook

†

Page Result

Short exercise

ic: Fundamentals

2.23 Lar

est and Smallest Inte

ers C1 98 

10 Avera

e score C2 118 

2.15 Arithmetic J 102 

2.18 Dis

Sha

es with Asterisks J 103 

2 Circle’s area and circumference P 16 

ic: Control structures

3.34 Flo

d’s Trian

le C1 142 

4.12 Prime Numbers C1 183 

6 Sum of evens and odds C2 330 

4.18 Credit Limits Calculato

J 182 

4.22 Tabular Out

J 183 

ic: Data structures

6.14 Union of Sets C1 298 

12.10 Reversin

the Words of a Sentence C1 544 

4 Countin

scores in ran

es C2 585 

22.11 Palindrome Teste

J 970 

2 Student Lis

P 36 

ic: Al

orithms

5.39 Recursive Greatest Common Diviso

C1 241 

5.41 Recursive Prime C1 242 

18.17 Print an Arra

Backward J 832 

19.5 Bubble Sor

J 862 

1Ana

ram P 49 

ic: Memor

mani

ulation

12.8 Insertin

into an Ordered List C1 544 

12.19 De

th of a Binar

Tree C1 547 

12.20 Recursivel

Print a List Backward C1 548 

13 Circular linked lists C2 1148 

14.6 D

namic Arra

Allocation C1 580 

ic: Ob

-oriented structures

19.10 Account Inheritance Hierarch

C1 798 

20.16 CarbonFootprint Abstract Class:

Pol

mor

hism

C1 843 

23.3 O

erator Overloads in Tem

lates C1 917 

9.8

uadrilateral Inheritance Hierarch

J 429 

21.7 Generic isE

ualTo Method J 939 

Lon

exercises

3.16 Mort

e Calculato

C1 137 

5.36 Towers of Hanoi C1 240 

7.17 Simulation: The Tortoise and the Hare C1 353 

7.22 Maze Traversal C1 356 

7.32 Pollin

C1 363 

12.12 Infix-to-Postfix Converte

C1 544 

15 Memor

Game C2 588 

16 Air

lane Seatin

Assi

nmen

C2 589 

16.5 Random Sentences J 748 

22.26 Insert/Delete An

where in a Linked Lis

J 975 

Modif

existin

code

3.27 Validatin

User In

C1 140 

4.15 Modified Com

ound-Interest Pro

ram C1 183 

4.22 Avera

e Grade C1 185 

5.20 Displaying a Rectangle of Any

Characte

C1 238 

5.33 Guess the Number Modification C1 240 

6.32 Linear Search C1 305 

12.16 Allowin

licates in a Binar

Tree C1 547 

12.23 Level Order Binar

Tree Traversal C1 548 

16.11 Modif

Class GradeBoo

C1 657 

17.13 Enhancin

Class Rectan

le C1 711 

†

Textbooks: C1: Deitel & Deitel (2016); C2: Malik (2013);

J: Deitel & Deitel (2012); P: Heathcote (2017).

Result:  = correct;  =

artiall

correct;  = incorrec

AI-Generated Programming Solutions: Impacts on Academic Integrity and Good Practices

485