ChatGPT to Generate Computer Science Problem Sets

International Journal on Integrating Technology in Education (IJITE) Vol.12, No.2, June 2023
DOI:10.5121/ijite.2023.12202 19
USING CHATGPT TO GENERATE COMPUTER
SCIENCE PROBLEM SETS
Tyler Flaagan
The Beacom College of Computer and Cyber Sciences, Dakota State University,
Madison, SD, USA
ABSTRACT
The topic of this paper is the exploration of using ChatGPT as a tool to assist educators in creating
innovative computer science problem sets for programming classes. The motivation for this research stems
from the challenges educators face in continually developing engaging and relevant problem sets that
stimulate students’ critical thinking and align with the evolving landscape of the computer science field. By
harnessing the capabilities of ChatGPT, we aim to provide a solution that can alleviate the burden on
educators and enable them to create dynamic, diverse, and effective learning experiences for their
students.
KEYWORDS
ChatGPT, machine learning, computer science, programming languages, problem sets.
1. INTRODUCTION
ChatGPT, an advanced language model developed by OpenAI, has gained significant attention in
recent months as a powerful tool for a variety of applications. It is based on the GPT-4
architecture and has demonstrated remarkable capabilities in generating human-like text [1]. In
this paper, we explore the potential of ChatGPT to assist computer science instructors in creating
innovative problem sets. ChatGPT very little in the way of information published due to its
novelty [2].
The motivation for this research lies in addressing the pressing challenges educators face in their
quest to develop engaging and relevant problem sets that stimulate students' critical thinking and
align with the rapidly evolving landscape of the computer science field. As technology advances
at an unprecedented pace, educators often need help to keep up with new trends and industry
developments while ensuring that their teaching methods and materials remain compelling and
captivating for students. Furthermore, the need to promote critical thinking and problem-solving
skills in the context of real-world applications adds a layer of complexity to the task. By
exploring innovative solutions such as ChatGPT, this research aims to empower educators with
the tools and resources necessary to overcome these challenges, ultimately fostering a more
enriching and dynamic learning environment for students in computer science courses.
ChatGPT has already begun to impact academia One such perspective discusses the potential
implications of publishing scholarly content using artificial intelligence that delve into copyright,
attribution, plagiarism, and authorship [2]. ChatGPT has shown an incredible ability to pass
traditional higher-education assessments, causing some to call for bans while taking tests [3].
Attempting to mitigate the negative uses of advanced language model tools will be difficult, if
not impossible, since similar products becoming available is inevitable. Other problems include a

20
lack of human interaction, limited understanding, bias in training data, lack of creativity,
dependency on data, lack of contextual understanding, limited ability to personalize instruction,
and privacy [4].
Advanced tools can positively impact students as well. One such example is the use of AI for
intelligent student tutoring [5]. Outside of tutoring, others have also discussed it’s use for solving
programming bugs to aid in software development [2]. They way in which it is used in an
educational environment will decide whether its impact is positive or negative [6].
2. BACKGROUND
The rapid advancements in computer science have made it increasingly more work for educators
to keep up with the pace and create engaging, relevant problem sets for their students. This
challenge often leads to a heavy workload and a need for more diverse educational resources that
cater to the ever-changing needs of programming classes. With the advent of advanced language
models like ChatGPT, there lies an opportunity to revolutionize the way educators develop
computer science problem sets, enabling them to adapt to the evolving landscape of the field
more efficiently.
ChatGPT, a powerful language model based on the GPT-4 architecture, has demonstrated its
capabilities in generating human-like text, understanding context, and providing creative
solutions across various domains. By harnessing these capabilities, educators can create a wide
array of problem sets that are tailored to their student's needs and consider current industry trends
and real-world scenarios. This allows for a more dynamic and diverse learning experience,
fostering critical thinking and problem-solving skills among students.
Integrating ChatGPT into creating computer science problem sets offers several advantages:
1. It can substantially reduce the time and effort educators spend developing new problem
sets, allowing them to focus on other essential aspects of teaching, such as providing
personalized feedback and facilitating student discussions.
2. ChatGPT can be fine-tuned to generate problems based on specific topics, difficulty
levels, or learning objectives, ensuring that the generated problems align with the
curriculum and educational goals.
3. ChatGPT can help educators consistently introduce novel and engaging problems,
maintaining students' interest and motivation throughout the course.
In an ideal scenario, an instructor could reuse the same examples semester after semester as
fundamental computer science knowledge units rarely change. In practice, some issues make the
reuse of problems less than ideal. One such issue is if a student needs to retake the same course.
The cause for retaking a course comes from various issues that do not need to be explored here,
but a student retaking the same course may already have work completed, and if they are
retaking, they may not have a strong understanding of the content. Another is that students may
post solutions online. There are two common reasons why a student may post online. First, they
are posting to homework-helping sites and forums such as homeworkhelper.com and chegg.com,
thus providing easy access to information that would fall under academic dishonesty. There are
positive reasons that a student may post online as well. An example is a student posting code to
code repository sites like Github.com. Posting to publicly available code repository sites allows
students to create a code portfolio to demonstrate proficiency to potential employers.
In conclusion, using ChatGPT to generate computer science problem sets holds immense
potential in transforming programming education. By addressing the challenges educators face in

21
developing relevant and engaging problems, ChatGPT can help bridge the gap between the
dynamic nature of the computer science field and students' learning experiences. Future research
needs to explore the practical implications of ChatGPT-based problem generation, assess its
effectiveness, and develop best practices that educators can adopt globally.
2.1. ChatGPT
ChatGPT, a cutting-edge language model, has recently gained significant attention as a powerful
tool for various applications across various fields. To effectively utilize ChatGPT, users can start
by posing a vague question and provide additional context in a subsequent prompt after receiving
the initial response. This allows the AI to recall and build upon previous conversations with the
user, ensuring a more coherent and meaningful interaction.
Users can refine the discussion during the conversation by providing clarifying comments or
feedback. This ensures that the AI understands the user's intent better, enabling more accurate
and informative responses. Additionally, ChatGPT can reject inappropriate requests, although
this feature does not impact the subject matter of this paper.
However, it is essential to recognize the limitations of ChatGPT. For instance, when generating
problem sets, the AI may occasionally produce incorrect information. Consequently, users should
verify the problem sets for accuracy and identify potential edge cases that need addressing.
Furthermore, ChatGPT may generate harmful instructions or biased information, a prevalent
concern in artificial intelligence and machine learning.
Another limitation was ChatGPT's knowledge cut off in September 2021. As a result, AI may
need more awareness of events, advancements, or innovations that have taken place after this
date. Despite this constraint, most fundamental computer science concepts remain unchanged,
allowing ChatGPT to continue serving as a valuable resource for generating problem sets. For
example, teaching an introductory concept like the variable declaration in the C programming
language has stayed the same since the end of ChatGPT's training period in 2021. Even with the
limitations mentioned earlier, ChatGPT remains a valuable tool for educational purposes and
various other applications, contributing significantly to artificial intelligence and machine
learning. ChatGPT is a sophisticated language model that has attracted considerable attention
lately as an influential instrument for numerous purposes across diverse fields.
3. RELATED WORKS
A myriad of sources exists for computer science problem sets, catering to the needs of both
students and instructors. Project Euler, one such resource, offers a collection of challenging
mathematical and computational problems [7]. These problems encourage creative problem-
solving and programming skills, allowing students to test their understanding and grow their
abilities. With many problems available, Project Euler is an excellent resource for educators to
diversify their curriculum and challenge their students. LeetCode is another valuable source for
coding challenges and problems. This platform helps students refine their programming skills and
prepares them for technical interviews. Educators can use LeetCode to create problem sets that
simulate real-world challenges and foster practical skills that students can carry into their
professional lives [8]. HackerRank provides coding challenges, tutorials, and competitions across
various domains, including algorithms, data structures, and artificial intelligence. The platform's
versatility and range of topics make it a valuable resource for educators looking to create well-
rounded problem sets. By incorporating HackerRank's coding challenges into their curriculum,

22
instructors can expose students to diverse concepts and assess their understanding in a
competitive and engaging environment [9].
For those seeking a more structured learning experience, platforms like Coursera offer online
courses covering various computer science topics. These courses often include problem sets,
assignments, and quizzes as part of the curriculum. Educators can use these resources to
supplement their classes, ensuring students receive a comprehensive understanding of the subject
matter. The high-quality materials provided by Coursera can serve as a solid foundation for
educators to build upon and adapt to their specific teaching needs [10].
Lastly, university websites often publish computer science course materials, including problem
sets, online for public access. Renowned institutions like MIT and Stanford provide a wealth of
resources for educators to create engaging and challenging problem sets. By leveraging these
materials, instructors can benefit from the expertise of leading computer science departments and
incorporate cutting-edge concepts into their classrooms. The wide array of sources available
ensures that educators can continually update and enrich their curriculum, keeping students
motivated and fostering a deep understanding of computer science principles.
4. CREATING COMPUTER SCIENCE PROBLEM SETS
The following sections will show results from ChatGPT when asking for specific learning
objectives. A specific language was not used when prompting ChatGPT to show generic
problems that can be adapted to whichever language is used in a particular class. Later,
specifying languages will be discussing in improving results. While creating examples for
demonstration, ChatGPT jailbreaks techniques such as the DAN jailbreak were not used to
modify results.
The examples demonstrated in the following sections are based on the following topics that could
be considered foundational concepts for developing applications in many programming and
scripting languages. The list is not meant to be all possible foundational concepts but instead is a
representative sample. The topics examined, in no particular order, are: if statements, loops,
arrays, functions, and pointers.
Each topic supplied is prefaced with the phrase:” Write a challenging problem for a computer
science course focused on learning…” followed by each specific topic from the list above.
4.1. If Statements
The following is the result of if statements without any modification by the author:
Title: "Temperature Alert System"
Problem Description:
You are tasked with developing a temperature alert system for an
industrial facility. The system will read temperature values in
Celsius from various sensors and categorize them into different
alert levels based on the following conditions:
1. Low (L) - temperature less than or equal to 0°C
2. Normal (N) - temperature greater than 0°C and less than or
equal to 80°C

23
3. High (H) - temperature greater than 80°C and less than or
equal to 100°C
4. Critical (C) - temperature greater than 100°C
Your program should take an integer n (number of temperature
readings) as input, followed by n temperature readings. For each
temperature reading, the program should print the corresponding
alert level.
Instructions:
1. Write a function called alert_level that takes a float
(temperature) as input and returns a string representing
the alert level.
2. Write the main function to read the temperature readings
and call the alert_level function for each reading, then
print the alert level for each temperature reading.
Constraints:
 The input integer n for the number of temperature readings
will be in the range of 1 ≤ n ≤ 100.
 The input float temperature readings will be in the range
of -100.0 ≤ temperature ≤ 200.0.
 Use if statements to categorize the temperature readings;
avoid using any loops in the alert_level function.
 The temperature values should be rounded to one decimal
place when displayed with their alert level.
Example:
Input: 3 -5 25 105
Output: L -5.0 N 25.0 C 105.0
4.2. Loops
The following is the result of loops without any modification by the author:
Title: "Fibonacci's Nested Loop Adventure"
In this challenge, you will write a program that generates a
nested Fibonacci sequence matrix using loops. The nested
Fibonacci sequence matrix is an n x n matrix, where each element
is a Fibonacci number, and each row starts with the Fibonacci
number of the index of the row. The index starts from 0.
For example, given n = 5, the output should be:
0 1 1 2 3 1 1 2 3 5 1 2 3 5 8 2 3 5 8 13 3 5 8 13 21
Instructions:

24
1. Write a function called fibonacci that takes an integer as
an input and returns the corresponding Fibonacci number.
2. Write a function called nested_fibonacci_matrix that takes
an integer n as input and returns a nested list
representing the n x n Fibonacci matrix.
3. Print the generated matrix in a well-formatted manner,
similar to the example provided.
Constraints:
 The input integer n for the nested_fibonacci_matrix
function will be in the range of 1 ≤ n ≤ 10.
 You must use loops to generate the matrix; recursion is not
allowed for this problem.
Note: The Fibonacci sequence is a series of numbers in which each
number is the sum of the two preceding ones, usually starting
with 0 and 1.
4.3. Arrays
The following is the result of arrays without any modification by the author:
Title: "Array Transformation: Multiplying Elements by their
Indices"
In this challenge, you will write a program that reads an array
of integers and performs a transformation on the array. The
transformation multiplies each element in the array by its index,
and the result is a new array. The program should then output the
transformed array.
Instructions:
1. Write a function called transform_array that takes an array
of integers, its length (int), and an empty array of the
same length as input. The function should perform the
transformation, filling the empty array with the
transformed values.
2. Write the main function to read the input array, call the
transform_array function, and print the transformed array.
Constraints:
 The number of input integers, n, will be in the range of 1
≤ n ≤ 100.
 The input integers will be in the range of -1000 ≤ integer
≤ 1000.
 The transformed array should have the same length as the
input array.
Example:

25
Input: 5 2 5 8 3 6
Output: 0 5 16 9 24
4.4. Functions
The following is the result of functions without any modification by the author:
Title: "GPA Calculator for Different Grading Systems"
In this challenge, you will write a program that calculates the
Grade Point Average (GPA) of a student for two different grading
systems. The program will take a series of letter grades as input
and convert each grade into a numeric value based on the given
grading system. It will then calculate the GPA using the
converted numeric values.
The two grading systems are as follows:
1. Standard Grading System (SGS):
 A: 4.0
 B: 3.0
 C: 2.0
 D: 1.0
 F: 0.0
2. Modified Grading System (MGS):
 A+: 4.3
 A: 4.0
 A-: 3.7
 B+: 3.3
 B: 3.0
 B-: 2.7
 C+: 2.3
 C: 2.0
 C-: 1.7
 D: 1.0
 F: 0.0
Instructions:
1. Write a function called convert_grade that takes a string
(letter grade) and a string (grading system) as input and
returns a float representing the numeric value of the grade
based on the given grading system.
2. Write a function called calculate_gpa that takes a list of
letter grades, a string (grading system), and the number of
courses (int) as input. The function should call the

26
convert_grade function for each letter grade and return the
calculated GPA.
3. Write the main function to read the input letter grades and
call the calculate_gpa function for both grading systems,
then print the GPA for each system.
Constraints:
 The number of input letter grades, n, will be in the range
of 1 ≤ n ≤ 20.
 The input letter grades will be valid according to the
grading systems provided (e.g., no invalid letter
combinations).
 The input letter grades will be in uppercase.
 Round the calculated GPA to two decimal places.
Example:
Input: 5 A B A- C+ B-
Output: SGS GPA: 2.80 MGS GPA: 2.94
4.5. Pointers
The following is the result of pointers without any modification by the author:
Title: "Pointer Manipulation: Sorting an Array of Strings"
In this challenge, you will write a C program that takes an array
of strings as input and sorts the strings alphabetically using
pointers. Your program should output the sorted array of strings.
Instructions:
1. Write a function called compare_strings that takes two
pointers to strings (char*) as input and returns an
integer. The function should return a negative value if the
first string is lexicographically less than the second
string, a positive value if the first string is
lexicographically greater than the second string, and 0 if
both strings are equal.
2. Write a function called swap_strings that takes two
pointers to strings (char**) as input and swaps the string
pointers.
3. Write a function called sort_strings that takes an array of
pointers to strings (char**) and its size (int) as input.
The function should sort the strings alphabetically using
the compare_strings and swap_strings functions.
4. Write the main function to read the input strings, call the
sort_strings function, and print the sorted array of
strings.
Constraints:

27
 The number of input strings, n, will be in the range of 1 ≤
n ≤ 50.
 The length of each input string will be in the range of 1 ≤
length ≤ 100.
 The input strings will only contain lowercase English
letters.
 Use pointers and pointer manipulation to implement the
sorting process. Avoid using built-in string manipulation
functions.
Example:
Input: 5 apple orange banana kiwi grape
Output: apple banana grape kiwi orange
5. REFINEMENTS
ChatGPT's emergence onto the Internet is still recent at the time of writing, meaning its full
potential may still need to be recognized. As technology advances, users and developers will
continue discovering new capabilities and applications for this powerful tool. One of the exciting
aspects of ChatGPT is its ability to adapt its responses based on additional information provided
by the user.
For example, if a user asks ChatGPT to "answer the question like I am five years old," the AI
model can tailor its response to the request, simplifying complex concepts and explanations.
Additionally, users can request that the response be written in a specific style, adding another
layer of customization to the generated content.
One of the critical capabilities of ChatGPT is its ability to reference previous parts of the
conversation. This feature creates opportunities for refining the content while creating problem
sets. ChatGPT can generate more relevant and accurate solutions by maintaining context and
incorporating prior information. Using the loop example from the prior section, it is possible to
refine or regenerate the response altogether. If a generated solution is too similar to another
example or the educator sees issues or errors, it can easily be regenerated by re-entering the same
question or clicking the regenerate button. This flexibility allows for continuous improvement of
the problem sets. Furthermore, continuing to enter questions in the same chat session can impact
the results. Since ChatGPT can reference previous prompts, it will adapt to the questions and
improve the generated content over time. This iterative process enables users to achieve
increasingly refined and targeted results. The difficulty is another aspect that can be modified. By
simply asking ChatGPT to increase the difficulty, a more complex problem can be created for
advanced computer science courses or for students who need to be challenged by the current
content. This adaptability ensures that problem sets remain engaging and appropriate for students
at various skill levels.
AI in higher education can produce positive results, but with the capabilities of advanced
language models, there is also the potential for negative impacts. The power of Large Language
Models (LLMs) is impressive in solving questions accurately when the user provides
considerable detail. Given enough information, ChatGPT can generate code and solutions for
various problems, highlighting its potential as a valuable tool in computer science education.

28
6. FUTURE WORKS
As we explore, further implementations of computer science problem sets, it is essential to
recognize that the availability of such tools is still in its early stages. The rapid development and
evolution of the field make it necessary to continually revaluate various research areas, including
creating problem sets using AI tools like ChatGPT. By staying up to date with advancements in
the field, educators can ensure that they make the most of these emerging technologies.
With the maturation and increased adoption of AI-driven technologies, it becomes increasingly
essential to revisit and reassess the implications and applications of these tools. Periodic
assessments can help educators identify new possibilities and potential benefits that arise as the
technology evolves. This ongoing evaluation process ensures that educators can make informed
decisions on integrating AI tools into their curriculum best. Moreover, the responsible and
effective use of innovative tools is crucial for harnessing their full potential in computer science
education. As AI technologies become increasingly sophisticated, educators must balance their
opportunities with the ethical considerations accompanying their use. By maintaining this
balance, instructors can create enriching and engaging learning experiences while upholding the
principles of responsible AI use.
7. RESULTS & CONCLUSION
This work shows that ChatGPT can generate focused problem sets for educators as needed.
ChatGPT, an advanced language model based on the GPT-4 architecture, offers immense
potential for revolutionizing computer science education. Its ability to generate human-like text,
adapt to user input and reference previous conversations makes it a valuable tool for creating
diverse and engaging problem sets for programming courses. By assisting educators in
developing relevant and challenging problems that align with the dynamic landscape of computer
science, ChatGPT contributes to fostering a more enriching and interactive learning environment
for students.
However, it is essential to acknowledge the limitations and ethical considerations surrounding the
use of AI tools in education. Ensuring the accuracy of generated content, mitigating potential
biases, and maintaining student privacy are crucial aspects that educators must address when
integrating AI-driven technologies into their curriculum. Moreover, staying up-to-date with the
rapidly evolving field of AI and its potential applications in education is vital for leveraging these
tools effectively and responsibly.
Future research should focus on assessing the practical implications of ChatGPT-based problem
generation, evaluating its effectiveness, and developing best practices that educators can adopt
globally. By doing so, the academic community can harness the full potential of advanced
language models like ChatGPT, ultimately creating more engaging and stimulating learning
experiences for computer science students.
REFERENCES
[1] “ChatGPT,” OpenAI. https://chat.openai.com/chat?model=gpt-4 (accessed Mar. 01, 2023).
[2] N. M. S. Surameery and M. Y. Shakor, “Use Chat GPT to Solve Programming Bugs,” Int. J. Inf.
Technol. Comput. Eng., no. 31, pp. 17–22, Jan. 2023, doi: 10.55529/ijitc.31.17.22.
[3] S. Kelly, “ChatGPT passes exams from law and business schools,” Jan. 26, 2023. [Online].
Available: https://www.cnn.com/2023/01/26/tech/chatgpt-passes-exams/index.html

29
[4] D. Baidoo-Anu and L. Owusu Ansah, “Education in the Era of Generative Artificial Intelligence
(AI): Understanding the Potential Benefits of ChatGPT in Promoting Teaching and Learning,”
SSRN Electron. J., 2023, doi: 10.2139/ssrn.4337484.
[5] J. Rudolph, S. Tan, and S. Tan, “ChatGPT: Bullshit spewer or the end of traditional assessments in
higher education?,” J. Appl. Learn. Teach., vol. 6, no. 1, Jan. 2023, doi: 10.37074/jalt.2023.6.1.9.
[6] M. Shidiq, “The Use of Artificial Intelligence-Based ChatGPT and its Challenges for the World of
Education; From the Viewpoint of the Development of Creative Writing Skills”.
[7] “Project Euler.” https://projecteuler.net/ (accessed Mar. 01, 2023).
[8] “LeetCode.” https://leetcode.com/ (accessed Mar. 01, 2023).
[9] “HackerRank.” https://www.hackerrank.com/ (accessed Mar. 01, 2023).
[10] “Coursera.” https://www.coursera.org/ (accessed Mar. 01, 2023).
AUTHORS
Tyler Flaagan is an Assistant Professor at Dakota State University, teaching in the Cyber Operations and
Network and Security Administration degrees. Tyler holds a Ph.D. and B.S. in Cyber Operations and an
M.S. in Computer Science. Tyler is also the Director of the Deep Red Lab within Dakota State University
MadLabs. Besides degrees, Tyler holds multiple professional certifications specializing in offensive
security, such as Offensive Security Certified Professional and GIAC Exploit Researcher and Advanced
Penetration Tester. Previously, Tyler worked as a red team operator for Naval Information Warfare Center
Atlantic, supporting various projects within the Department of Defense.

ChatGPT to Generate Computer Science Problem Sets

More Related Content

Similar to ChatGPT to Generate Computer Science Problem Sets (20)

Recently uploaded (20)

ChatGPT to Generate Computer Science Problem Sets