International Journal of Artificial Intelligence in Education

https://doi.org/10.1007/s40593-023-00372-z

ARTICLE

Can ChatGPT Pass High School Exams on English Language

Comprehension?

Joost C. F. de Winter1

Accepted: 4 September 2023

© The Author(s) 2023

Abstract
Launched in late November 2022, ChatGPT, a large language model chatbot, has
garnered considerable attention. However, ongoing questions remain regarding its
capabilities. In this study, ChatGPT was used to complete national high school
exams in the Netherlands on the topic of English reading comprehension. In late
December 2022, we submitted the exam questions through the ChatGPT web inter-
face (GPT-3.5). According to official norms, ChatGPT achieved a mean grade of 7.3
on the Dutch scale of 1 to 10—comparable to the mean grade of all students who
took the exam in the Netherlands, 6.99. However, ChatGPT occasionally required
re-prompting to arrive at an explicit answer; without these nudges, the overall grade
was 6.5. In March 2023, API access was made available, and a new version of
ChatGPT, GPT-4, was released. We submitted the same exams to the API, and GPT-
4 achieved a score of 8.3 without a need for re-prompting. Additionally, employing
a bootstrapping method that incorporated randomness through ChatGPT’s ‘tempera-
ture’ parameter proved effective in self-identifying potentially incorrect answers.
Finally, a re-assessment conducted with the GPT-4 model updated as of June 2023
showed no substantial change in the overall score. The present findings highlight
significant opportunities but also raise concerns about the impact of ChatGPT and
similar large language models on educational assessment.

Keywords GPT-3.5 · GPT-4 · Large language model · Educational assessment ·

Reading comprehension

Joost C. F. de Winter
[email protected]

1
Cognitive Robotics Department, Delft University of Technology, Delft, Netherlands

13
 International Journal of Artificial Intelligence in Education

Introduction

ChatGPT, developed by OpenAI, is a chatbot designed to engage in conversation

with users and generate responses that are human-like and meaningful. Launched
on November 30, 2022, it gained widespread attention due to its ability to generate
coherent output across a wide range of topics, including creative writing (King, 2023;
Kirmani, 2023; Zhai, 2022), computer coding and bug fixing (Davis et al., 2023;
Sobania et al., 2023), teaching innovation (Rudolph et al., 2023; Šlapeta, 2023), sen-
timent analysis (Tabone & De Winter, 2023; Zhong et al.,2023a), and various annota-
tion tasks, such as genre identification (Kuzman et al., 2023) and the identification
of language explicitness (Huang et al., 2023; Rospocher & Eksir, 2023). However,
it has also been argued that ChatGPT output tends to contain incorrect statements,
cliché-like writing, and faulty references to scientific sources (Han et al., 2023; Kim
et al., 2022; Lovin, 2022; Vincent, 2022; Whitford, 2022).
Performing benchmark tasks on large language models like ChatGPT is crucial to
understand their strengths and weaknesses, as well as to identify potential areas for
improvement. A recent study by Gilson et al. (2022) revealed that ChatGPT demon-
strates a reasonable performance on medical licensing exams, with results compa-
rable to third-year medical students in one of the four exams assessed. Kung et al.
(2023) corroborated this finding by demonstrating that the model achieved scores
near the passing threshold for the United States Medical Licensing Exam (USMLE).
However, ChatGPT’s performance was found to be below average in physics (Korte-
meyer, 2023) and in various mathematical domains, such as Olympiad problem-solv-
ing, functional analysis, and symbolic integration (Frieder et al., 2023). Similarly,
Newton and Xiromeriti (2023) reported that ChatGPT lagged behind the average
student in multiple-choice tests across diverse fields, including ophthalmology, law,
economics, and physics. Therefore, further research is necessary to evaluate Chat-
GPT’s performance across different types of exams.
The aforementioned evidence pertains to GPT-3.5. On March 14, 2023, a sub-
sequent version, GPT-4, was introduced, accompanied by API (Application Pro-
gramming Interface) access. OpenAI (2023) claimed that GPT-4 exhibits superior
performance compared to GPT-3.5, particularly in the realm of intricate and nuanced
communication. Bordt and Von Luxburg (2023) found that, in an undergraduate com-
puter science exam, GPT-4 obtained 17% more points than GPT-3.5, approaching
average student performance. Savelka et al. (2023) investigated the proficiency of
GPT-4 in postsecondary Python programming assessments and reported a substantial
improvement over previous generations, where GPT-4 can now pass courses inde-
pendently. However, limitations remained in handling complex coding exercises.
Bommarito and Katz (2022) found that GPT-3.5 achieved an average score of 45.1%
on the Uniform Bar Examination, encompassing subjects such as civil procedure,
constitutional law, and contracts, while in their most recent update, GPT-4 attained a
remarkable score of 75.7% (Katz et al., 2023). A similar analysis conducted by Ope-
nAI (2023), comparing the performance of both models on the Uniform Bar Exami-
nation, revealed scores of 53% and 75% for GPT-3.5 and GPT-4, respectively. These
results correspond to the performance of the lowest 10% of human test-takers for
GPT-3.5 and the highest 10% for GPT-4. OpenAI (2023) conducted several other

13
International Journal of Artificial Intelligence in Education

benchmark evaluations to compare GPT-4 with GPT-3.5 and demonstrated improved

performance for GPT-4 in nearly all instances. Of note, while GPT-4 excelled in
the GRE verbal section (99th percentile), it scored modestly in GRE writing (~ 54th
percentile).
When we conducted initial explorations of GPT-3.5’s and GPT-4’s performance
on various online tests, we observed that ChatGPT was proficient at verbal analogies
commonly found in IQ tests but faced difficulties with solving logical puzzles that
require variable tracking, such as determining which guest was staying in a particular
hotel room (available at https://www.123test.com/verbal-reasoning-test). This pre-
liminary observation is in line with recent more comprehensive analyses (Arora &
Singh, 2023; Bubeck et al., 2023; Zhong et al., 2023b), which explicate that ChatGPT
exhibits certain limitations, specifically its inability to globally plan, engage in an
‘inner dialogue’, or self-correct its errors. While ChatGPT demonstrates remarkable
emergent properties such as sophisticated language production and basic arithmetic
(Kosinski, 2023; Wei et al., 2022), it essentially operates as an autoregressive model,
which means that it predicts each next word by conditioning on the words that have
already been produced, thus generating text one word at a time in a sequential man-
ner. In light of these characteristics, it becomes compelling to investigate whether
ChatGPT can successfully execute a high school reading comprehension exam—a
task that, upon initial observation, one might expect necessitates intricate reasoning.
In this study, we employed ChatGPT versions 3.5 and 4 to analyze their perfor-
mance on Dutch national exams, focusing on English reading comprehension. In
their technical report, OpenAI (2023) recognized the potential risk of contamination
in their benchmark evaluations, where the model might inadvertently have access
to the test questions and corresponding answers. OpenAI reported only a marginal
difference between the performance of ‘contaminated’ and ‘non-contaminated’ ques-
tions. However, their approach to screening contaminated data could be susceptible
to misses and false positives. To mitigate the contamination issue, the current study
used exams from 2022 (i.e., after the ChatGPT knowledge cut-off date), ensuring that
ChatGPT’s training datasets did not contain the exams or answer models being exam-
ined. A recent study by Chen et al. (2023) highlighted a marked decline in ChatGPT’s
(GPT-4) performance on selected subtasks between March and June 2023. Conse-
quently, the criticality of evaluating the capabilities of ChatGPT becomes increas-
ingly apparent. Therefore, in the present study, we assessed two specific versions of
GPT-4, those of March and June, to better understand possible shifts in performance.

Methods

We applied ChatGPT (version: December 15, 2022) to national exams of the VWO
program (Preparatory Scientific Education) in the Netherlands that tested English
reading comprehension. These high-stakes exams are administered by the Dutch
organization CITO (CITO, 2023), and are mandatory for all VWO students. The
VWO program is considered the most academically rigorous high-school education
in the Netherlands, designed for students who intend to pursue university-level stud-
ies. In the VWO English exam, students are tasked with reading a variety of passages,

13
 International Journal of Artificial Intelligence in Education

such as newspaper items, and answering associated questions. Although the specific
texts and questions change with each exam, the nature of the questions, such as iden-
tifying the main idea, making inferences, and interpreting vocabulary within context,
remains largely consistent. In this study, only exams from 2022 were used because
the database of ChatGPT has a knowledge cut-off in 2021.
The three available exams incorporated a total of 43, 40, and 41 questions, respec-
tively. Each exam comprised an accompanying textbook with 11 textual passages. Of
the three exams, there were 31, 32, and 29 multiple-choice questions, respectively.
These questions typically had four response options [A, B, C, and D]. However, some
questions offered three (A–C: 13 questions), five (A–E: 7 questions), or six (A–F: 1
question) response options. Every multiple-choice question carried the value of one
point. Further, the exams incorporated a number of open questions also worth one
point each. These questions either required succinct answers, called for a sequence of
statements to be arranged in a specific order, or asked for a response to a maximum
of two sub-questions in the form of a ‘yes’ or ‘no’. Examples of such questions were
as follows (translated from Dutch to English):

● “8. ‘why should he lose his perk?’ (paragraph 4). What was this ‘perk’? Please
respond in Dutch.” (Exam 1).
● “20. The text divides into critical and non-critical segments. In which paragraph
does the critical part commence? Please indicate the paragraph number.” (Exam
1).
● “2. ‘European ruling’ (title). Please determine whether Mike Short identifies each
of the following points as an issue with the introduction. Write “yes” or “no”
alongside each number on the answer sheet.

1. The implementation is impeded by the current web infrastructure.

2. The interpretation of the ‘right to be forgotten’ is contingent upon one’s cultural
background.” (Exam 3).

The remaining questions (3 to 6 per exam), worth either 2 or 4 points, were multi-part
items. More specifically, these involved scenarios where 4 to 8 statements needed a
‘yes’ or ‘no’ response, or items where multiple themes or locations in the text had to
be identified.
The text (e.g., a news item or another text fragment that was part of the exam) and
the corresponding questions were manually submitted one by one to the ChatGPT
web interface (GPT-3.5), as part of the same chat session for each exam. After the
text and before each question, a prompt was included, e.g., “Based on ‘Text 5’ above,
please choose the correct response option between A, B, C, and D for the question
below (note that the number in front of each paragraph indicates the number of the
paragraph)”.
In 15% of cases, ChatGPT appeared to misconstrue the question, leading to invalid
responses. These included not selecting any options in a multiple-choice question,
generating an entirely new question, or asserting that the correct answer could not
be determined due to insufficient information. When such a scenario occurred, the
researcher would either reiterate the question or provide further clarification. This

13
International Journal of Artificial Intelligence in Education

could involve prompting ChatGPT to choose from the response alternatives or

explaining that there was a missing word in the text that required completion. This
method occasionally facilitated a response. For example, Question 9 of Exam 3 asked
“Based on ‘Text 5’ above, please choose the correct response option between A, B,
C, D, and E for the question below (note that the number in front of each paragraph
indicates the number of the paragraph). Which of the following fits the gap (indicated
with “…”) in paragraph 1?”, together with five response alternatives (A–E). Note
that the phrase “(indicated with “…”)” was our addition to aid ChatGPT. The initial
response of ChatGPT was an elaborate general reflection: “According to data col-
lected by social scientists, there is little evidence that the typical terrorist is poor or
poorly educated. … Instead, other factors such as political and religious ideology
may be more influential in determining who becomes a terrorist.” This is why the
prompt was repeated, after which ChatGPT offered a response: “B: little to lose”. Out
of a total of 124 questions posed, 18 required re-prompting once or twice, resulting
in an eventual response in 15 of these instances. The model’s input and output during
this process were documented and can be found in the supplementary material.
The three available ‘VWO English 2022’ exams were completed by ChatGPT.
ChatGPT’s answers were assessed by an independent experimenter and checked by
the author of this work. The assessment was done using official scoring booklets con-
taining the correct answers (College voor Toetsen en Examens, 2022). These book-
lets also contained clear rules for assessment. One of these rules is: “If more than one
example, reason, elaboration, quote or other type of answer is asked for, only the first
given answers are assessed, up to the maximum number requested”, and another rule
is “The correct answer to a multiple choice question is the capital letter that belongs
to the correct choice. If the answer is given in a different way, but it is unequivocally
established that it is correct, then this answer should also be counted as correct.”. It
is noted that although the experimenter knew this was an assessment of ChatGPT, the
exams were designed in such a way that the answers left very little room for doubt
since the correct answers were mostly numbers, keywords, letters (e.g., A, B, C, D),
or yes/no statements. The assessor and author had no disagreements about whether
particular answers should be scored as correct or incorrect. Finally, the number of
points obtained on the exams were converted to a grade mark on a scale from 1.0 to
10.0 using formal conversion tables (College voor Toetsen en Examens, 2022).
On March 17, 2023, the above analysis was repeated using GPT-4 (model GPT-4-
0314). Rather than manually inputting the comprehension texts and questions to the
ChatGPT web interface, the API was used, and the procedure was fully automated
in MATLAB (version R2021b). Specifically, a script was written (see supplementary
material) that read the PDF files of the Text booklet and the Question booklet per
exam, grouped the questions by text based on their headers, and fed the questions to
the OpenAI API. When reading the text and questions, line breaks were removed, but
no additional processing was done. For example, page numbers, points per question
and page headers and footers were not removed.
The following prompt was used in MATLAB to feed the questions to the OpenAI
API: “Answer all questions, including multiple-choice questions. Only provide the
question number and the answer, nothing else:”. The prompt subsequently included
the word ‘TEXT:’ in a new line. This was followed by the entire text and the word

13
 International Journal of Artificial Intelligence in Education

‘QUESTIONS:’, which preceded all the questions for that text. This prompt design
was used for all questions. It was adopted to ensure that the API only outputs the
answers without additional clarifying text, which speeds up the response by Chat-
GPT. In all instances, GPT-4 provided concise answers, typically a letter, number,
or keyword as requested. This enabled a straightforward manual evaluation of the
examinations.
An example of a complete prompt—in this case, Text 10 from Exam 2 and three
corresponding questions (Questions 36–38)—is presented in Fig. 1. It can be seen
that the text has not been preprocessed; the prompt represents how the PDF files of
the text booklet and the exam were automatically read in, which is accompanied by
unnecessary spaces, headers and footers, and information for the candidate about
how many points the question is worth. Attempts were also made to submit cleaned-
up text to ChatGPT instead of texts in their raw form. However, this did not appear
to lead to improvement in the accuracy of the answers provided by GPT-4. We sub-
mitted raw text data, as depicted in Fig. 1, to provide an evaluation of ChatGPT free
from human intervention.
More recently, in June 2023, an update of GPT-4 was released. An evaluation
by Chen et al. (2023) reported that this version performed highly differently on
certain tasks, such as answering sensitive questions or generating code, compared
to the March version, potentially having far-reaching consequences for users and
applications. However, apart from the work of Chen et al. (2023), there is currently
limited information in the literature regarding any disparities in output between the
March and June versions. Hence, we conducted a re-analysis using the June version
(GPT-4-0613).
While the web interface of ChatGPT has an element of stochasticity in its responses,
the API offers the ability to modulate this randomness via a ‘temperature’ parameter
that is adjustable within a continuum from 0 to 2. With a temperature setting at 0,
the output is highly reproducible. In contrast, a temperature setting at 2 introduces a
significant degree of randomness or creativity in the output. For our study, the tem-

Fig. 1 One of the prompts submitted to GPT-4 (Questions 36–38 of Exam 2)

13
International Journal of Artificial Intelligence in Education

perature parameter was set to 0. The entire analysis, from automatic reading of the
texts and questions to letting GPT-4 produce the responses, took about 40 s per exam.

Results

Table 1 presents the results of the exams completed by ChatGPT. In the Netherlands,
a mean grade of 5.50 or higher across all courses would imply a pass of the high
school diploma. It can be seen that GPT-3.5 would pass each of the English exams,
with an overall mean grade across the exams of 7.3, whereas GPT-4-0314 and GPT-
4-0613 had overall mean grades of 8.3 and 8.1, respectively.
Upon categorizing the test items into (1) multiple-choice questions, (2) open one-
point questions, and (3) open questions valued at more than one point, it became
evident that GPT-3.5 faced some challenges with the second category (79%, 68%,
and 77% of the points earned). Compared to GPT-3.5, GPT-4 showed improvement
in the multiple-choice questions (GPT-4-0314: 92%, 84%, 80%; GPT-4-0613: 92%,
63%, 83%).
It should be acknowledged that GPT-3.5 had an inherent advantage, as we occa-
sionally provided a repeated prompt through the web interface to procure an explicit
response (see Methods section for details). Upon limiting our evaluation to only the
first instance of its (non-)response, GPT-3.5’s mean score for the three exams was 6.5,
as shown in Table 1. In contrast, the interaction with GPT-4 was fully automated via
the API, without any re-prompting. We also made an attempt to use GPT-3.5 via the
API (model GPT-3.5-turbo-0301), but the performance was deemed unsatisfactory.
The grades for Exams 1, 2, and 3 were only 5.7, 6.7, and 7.0 respectively. The score
for the first exam was particularly low, as GPT-3.5 did not answer 7 of the 43 ques-
tions. Considering these outcomes, we chose not to further investigate this approach.

Bootstrapping Self-Consistency Analysis

As shown above, GPT-3.5 performed comparably to the average Dutch high school
student in their final year of Preparatory Scientific Education, while GPT-4 outper-
formed the average student. However, GPT-4 was not flawless and made several mis-
takes on each exam. To further investigate, we explored if GPT-4 could self-identify
the questions it failed. Initial attempts using specific prompts, such as ‘Please rate the
difficulty of the question’ did not yield meaningful insights.
Previous research showed that implementing a self-consistency strategy may prove
beneficial in directing large language models towards accurate outputs (Wang et al.,
2023; Zheng et al., 2023). This entails generating a number of candidate outputs,
with the most frequent or internally consistent output being selected. In the present
study, we attempted to further this idea by incorporating an element of stochasticity
into the output, allowing us to gauge the level of confidence exhibited by GPT-4 with
respect to its own outputs. Specifically, we discovered that employing the ‘tempera-
ture’ parameter in conjunction with multiple repetitions yielded valuable insights.
In our self-consistency analysis, we used a temperature parameter of 1. This deci-
sion was informed by presenting a single prompt a large number of times under dif-

13
13
Table 1 Performance of ChatGPT on Dutch national exams on the topic of ‘English’
Number of points Corresponding grade (1 to 10)
GPT-3.5 (without GPT-4-0314 GPT-4-0613 Maximum GPT-3.5 (without GPT-4-0314 GPT-
re-prompting) number re-prompting) 4-
of points 0613
attainable
Exam 1 (May 2022, Period 1) 41 (36) 46 44 49 7.9 (7.0) 8.9 8.5
Exam 2 (June 2022, Period 2) 35 (30) 40 40 46 7.0 (6.1) 8.0 8.0
Exam 3 (July 2022, Period 3) 33 (30) 39 38 46 7.0 (6.4) 8.1 7.9
Total 109 (96) 125 122 141
Average 7.3 (6.5) 8.3 8.1
Note. The prompting of GPT-3.5 was done through the web interface, while the prompting of GPT-4 was done using the API with a temperature setting equal to 0
International Journal of Artificial Intelligence in Education
International Journal of Artificial Intelligence in Education

ferent temperature settings. The particular prompt used (see Fig. 1) required ChatGPT
to answer three questions, with the correct answers being C, B, and B, respectively.
Figure 2 shows how the correctness of answers varied with different temperature
settings, from the minimum value of 0 to the maximum possible value of 2.0, with
increments of 0.1. It can be seen that with the temperature set at 0, Questions 36 and
38 were answered correctly, while Question 37 was answered incorrectly. As the
temperature increased, Questions 36 and 38 consistently received correct answers,
suggesting that ChatGPT had confidence in its answers. Nonetheless, at very high
temperature settings, ChatGPT produced incorrect responses. Upon further investi-
gation, we noted that these faulty answers were not inaccurate responses to the mul-
tiple-choice questions, but rather ‘hallucinations’ from ChatGPT, where it generated
nonsensical text instead of an answer to the questions. In contrast, Question 37 elic-
ited a level of uncertainty from ChatGPT, where, as the temperature rose, the correct
answer surfaced approximately 20% of the time. Based on these observations, we
made the decision to conduct a bootstrapping analysis for assessing self-consistency
using a temperature setting of 1.0. This value ensures a degree of output variation, yet
it restricts excessive variation that might cause ChatGPT to generate arbitrary texts.
After deciding upon the temperature setting of 1.0, we submitted each of the three
exams to GPT-4, 50 times each. The number 50 is a trade-off, in which too few repeti-
tions carry the risk that GPT-4 coincidently produces the same output multiple times in a
row, appearing consistent when it actually is not. On the other hand, too many repetitions
involve unnecessary use of computational resources and can give the false suggestion that
GPT-4 is not consistent if it only very rarely produces an alternative output. The 50 repeti-
tions were accomplished by setting the ‘n’ parameter in the API to 50.
The performance of GPT-4 for each exam question was then manually scored as
above, and classified into three categories:

● Questions for which GPT-4 answered correctly in all 50 attempts, indicating high
consistency.
● Questions for which GPT-4 provided the same incorrect response in all 50
attempts, indicating high consistency but a wrong answer. In the case of a ques-
tion worth more than one point, not achieving the full points was also considered
an incorrect response for the respective question.
● Questions for which GPT-4 provided at least two different responses over the 50
attempts, indicating inconsistency.

Our analysis showed that out of 124 questions across the three exams combined,
GPT-4-0314 displayed inconsistency in 23 cases (19%), while GPT-4-0613 displayed
inconsistency in 25 cases (20%) (see Table 2). Among these inconsistent responses,
a significant portion (10 or 43% for GPT-4-0314; 14 or 56% for GPT-4-0613) were
indeed incorrect. In contrast, of the 101 and 98 consistent responses for GPT-4-0314
and GPT-4-0613, only 4 and 4 were incorrect (“consistently incorrect”).
The final grades on a scale from 1 to 10 were calculated by averaging the results
over 50 repetitions, and then further averaging across three exams. GPT-4-0314
scored an average of 8.11 (Exams 1–3: 8.63, 7.80, 7.89), while GPT-4-0613 obtained
an average score of 8.21 (Exams 1–3: 8.64, 7.97, 8.01). These results are comparable

13
 International Journal of Artificial Intelligence in Education

Fig. 2 Percentage of correct responses by submitting the prompt shown in Fig. 1 for a total of 320 times
(model: GPT-4-0314). The procedure was repeated for 21 different temperature settings, from 0 to its
maximum value of 2.0

Table 2 Number of exam questions per category based on the consistency of 50 GPT-4 attempts (GPT-4-
0314 / GPT-4-0613)
Exam 1 Exam 2 Exam 3 Total
Consistently correct (50 times correct) 36 / 33 31 / 31 30 / 30 97 / 94
Consistently incorrect 1/2 3/1 0/2 4/5
(50 times the same incorrect response)
Inconsistent 6/8 6/8 11 / 9 23 / 25
(1–49 times the correct response)
of which GPT-4 answered incorrectly at 2/4 2/5 6/5 10 / 14
temperature = 0
of which GPT-4 answered correctly at 4/4 4/3 5/4 13 / 11
temperature = 0
Total 43 40 41 124

to those presented in Table 1, where the mean score across the three exams was 8.3
and 8.1 for GPT-4-0314 and GPT-4-0613, respectively. In summary, the exam grades
based on bootstrapping align with the original analysis from Table 1, where a tem-
perature setting of 0 was used.
Figure 3 illustrates the performance of the two ChatGPT versions on Exam 1 rela-
tive to all students who completed this exam. The students’ mean number of points
was 35.88 (SD = 6.48) out of a maximum of 49, and the mean of their grades was
6.99. This analysis was conducted only for one of the three exams, as the students’
results for the other two exams were not publicly available.

13
International Journal of Artificial Intelligence in Education

Fig. 3 Distribution of student scores on Exam 1 (n = 35,698). The performance of GPT-3.5 (corre-
sponding to the 46th and 76th percentiles), GPT-4 (97th and 91st percentiles), and bootstrapped GPT-4
(mean score of repetitions: 44.58 and 44.54) is depicted

Discussion

The present study’s results indicate that GPT-3.5 performs comparably to, while GPT-4
significantly outperforms, the average Dutch student in the domain of English language
comprehension. Although students are prohibited from using computers during conven-
tional in-person examinations, our findings suggest that ChatGPT could compromise
the integrity of computer-based exams, which have gained popularity in the wake of the
COVID-19 pandemic (Kerrigan et al., 2022; Pettit et al., 2021). Educators may presume
that online exams with minimal supervision are secure in subjects such as comprehension,
where answers are unlikely to be readily accessible online. However, this assumption
may no longer hold, given that our study demonstrates the generation of valid answers
within minutes. Concurrently, there are concerns that ChatGPT could be exploited for
cheating on assessments (Cotton et al., 2023; Mitchell, 2022), necessitating a reevalu-
ation of current methods for assessing student knowledge. Potential solutions include
increased proctoring, reduced reliance on essay-based work, and the utilization of alterna-
tive assignment formats, such as videos or presentations (Geerling et al., 2023; Graham,
2022; Rudolph et al., 2023; Susnjak, 2022).
On a positive note, ChatGPT holds the potential to foster innovation in the realm of
education. Possible applications encompass aiding the development of writing skills,
facilitating comprehension through step-by-step explanations, speeding up informa-
tion delivery via summarization of texts, and enhancing engagement through person-
alized feedback (Kasneci et al., 2023; Rudolph et al., 2023; Šlapeta, 2023). It is worth
considering whether the focus of student assessment should transition towards the
effective utilization of ChatGPT and similar language models. For instance, it may
be advisable to instruct students on identifying inaccuracies in content generated by

13
 International Journal of Artificial Intelligence in Education

ChatGPT or on integrating ChatGPT to establish a synergistic combination of human

and computational capabilities.
A noteworthy observation emerged when a bootstrapping method, consisting of
50 repetitions, was used to determine if GPT-4 exhibited uncertainty in its outputs.
This approach has demonstrated its efficacy as a tool for self-assessment, where we
found that about half of the responses labeled as ‘inconsistent’ were incorrect while
only about 5% of the responses deemed ‘consistent’ were incorrect. The exploitation
of randomness and bootstrapping has the potential to be a tool for future research.
Here it is noteworthy that, although the self-consistency method in our case provided
a way for ChatGPT to make statements about the certainty of the answer it delivered,
this method did not prove to be useful for arriving at a more accurate answer, contrary
to findings by Wang et al. (2023). A possible explanation is that the English exam
questions did not involve chain-of-thought reasoning as in Wang et al. (2023), where
diverse reasoning paths could lead to the same correct answer. In our case, ChatGPT
had to directly converge on the correct answer (such as the letter ‘A’, ‘B’, ‘C’, or
‘D’). This can explain why the answer that ChatGPT found most likely (i.e., at the
temperature setting of 0) also had the highest probability of being correct, and adding
variation by choosing a higher temperature provided no additional value in terms of
the accuracy of the output.
Interestingly, the principle of self-consistency may also provide advantages in the
construction of tests and exams. For Exam 1, psychometric properties were available
online (CITO, 2022). An initial analysis revealed that, for the ‘inconsistent’ or incor-
rectly answered questions of Exam 1 (n = 7 for GPT-4-0314, n = 9 for GPT-4-0613,
as shown in Table 2), students achieved an average score of 63.0% (SD = 10.5%)
and 67.7% (SD = 12.0%), compared to 75.8% (SD = 12.7%) and 75.3% (SD = 13.2%)
for the remaining questions (n = 36, n = 34). Thus, questions that ChatGPT answered
inconsistently or incorrectly appeared to pose greater challenges to human examin-
ees. Based on this insight, organizations that design exams could apply the principle
of self-consistency to verify the level of difficulty of their exams, aiming to maintain
consistency year on year. For instance, by having ChatGPT repeatedly attempt newly
developed exams (with a temperature setting of 1) and assessing the mean consis-
tency level across all items, one could obtain an initial measure of exam difficulty.
However, the feasibility and efficacy of this approach would require further valida-
tion through additional research.
Chen et al. (2023) observed a substantial deterioration in the performance of the lat-
est version of GPT-4 compared to its predecessor from a few months earlier. Contrary
to this, our analysis found no notable difference between the two versions. Our results
revealed an average exam grade of 8.3 for the March version and 8.1 for the June ver-
sion when using a temperature setting of 0, while upon using a bootstrapping method
involving 50 repetitions with a temperature setting of 1.0, the resultant grades were 8.11
and 8.21, respectively. A plausible explanation for the absence of noticeable disparity lies
in the fact that the base model, GPT-4, remained the same, with updates, as far as is cur-
rently known, focusing exclusively on the fine-tuning layers. Chen et al. (2023) showed
a degradation in the June version of GPT-4’s capability to determine whether a given
number is prime. It is important to note here that ChatGPT, in its essence, is incapable
of running algorithms required to determine if a number is prime; nevertheless, through

13
International Journal of Artificial Intelligence in Education

a smart chain-of-prompt strategy, accurate responses can still be obtained. In a parallel

finding, Chen et al. (2023) reported that the June version of GPT-4 exhibited a decreased
willingness to respond to sensitive content questions compared to the March version. We
also found that there were differences between the GPT-4-0314 and GPT-4-0613 outputs,
even though the overall average was comparable. With a temperature setting of 0, GPT-
4-0314 answered a total of 14 questions incorrectly (worth 16 points), while GPT-4-0613
incorrectly answered 18 questions (worth 19 points). However, only 9 of these questions
were answered incorrectly by both models (see supplementary material for the complete
overviews). There was even one question where the bootstrapping analysis showed that
GPT-4-0314, in 49 out of 50 instances, gave the answer ‘C’, while GPT-4-0613, in all 50
instances, correctly gave the answer ‘B’. We have no logical explanation for this varia-
tion, except that we have noticed that very minor changes in the prompt can have a large
impact (see also Reiss, 2023), which could be traced back to a potential lack of robust-
ness in the autoregressive modeling principle (LeCun, 2023). In summary, it seems that
GPT-4’s performance in English language comprehension has not changed over time,
even though it is possible that specific subtasks are performed differently by each version.
One limitation of our study was that ChatGPT was applied to an English exam for
Dutch-speaking students, which means that the level of the exam is not equivalent
to a native language exam. The estimated level of earlier VWO English exams was
reported to be C1, with a score of 64% as the minimum required to achieve this level
(College voor Toetsen en Examens, 2020). Furthermore, even though our study was
conducted on a well-constructed high school exam, it does not guarantee that Chat-
GPT will also perform well in other types of verbal tests.
With the introduction of GPT-4, large language models have reached a point
where they can outperform average humans in certain areas. There is a high prob-
ability that ChatGPT or similar models will evolve to become more intelligent over
time. Given that current models can already do well on high school exams today, it
raises significant questions about what lies ahead in education and beyond. A wealth
of opportunities are at reach for new applications and integrations, such as incorpora-
tion into, for example, Microsoft 365 Copilot (Office Microsoft Blog, 2023). As large
language models continue to evolve, they have the potential to redefine the boundar-
ies of human-computer interaction.

Acknowledgements Dr. Dimitra Dodou’s role in scoring the output of ChatGPT according to the correc-
tion instruction is acknowledged.

Declarations The author has no relevant financial or non-financial interests to disclose.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License,
which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long
as you give appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use
is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission
directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/
licenses/by/4.0/.

13
 International Journal of Artificial Intelligence in Education

References
Arora, D., & Singh, H. G. (2023). Have LLMs advanced enough? A challenging problem solving bench-
mark for large Language Models. arXiv. https://doi.org/10.48550/arXiv.2305.15074.
Bommarito, M. J., II, & Katz, D. M. (2022). GPT takes the Bar Exam arXiv. https://arxiv.org/
abs/2212.14402.
Bordt, S., & Von Luxburg, U. (2023). ChatGPT participates in a computer science exam arXiv. https://
arxiv.org/abs/2303.09461.
Bubeck, S., Chandrasekaran, V., Eldan, R., Gehrke, J., Horvitz, E., Kamar, E., Lee, P., Lee, Y. T., Li, Y.,
Lundberg, S., Nori, H., Palangi, H., Ribeiro, M. T., & Zhang, Y. (2023). Sparks of artificial general
intelligence: Early experiments with GPT-4 arXiv. https://arxiv.org/abs/2303.12712.
Chen, L., Zaharia, M., & Zou, J. (2023). How is ChatGPT’s behavior changing over time? arXiv. https://
doi.org/10.48550/arXiv.2307.09009.
CITO (2023). CITO: toetsen, examens, volgsystemen, certificeringen en trainingen [CITO: tests, exams,
tracking systems, certifications, and trainings]. https://cito.nl.
CITO (2022). Toets en item analyse VWO Engels 2022 tijdvak 1 [Test and item analysis VWO English
2022 period 1]. https://www2.cito.nl/vo/ex2022/VW-1002-a-22-1-TIA.docx.
College voor Toetsen en Examens (2020). Syllabus centraal examen 2022 Arabisch, Duits, Engels, Frans,
Russisch, Spaans, Turks [Syllabus central exams 2022 Arabic, German, English, French, Russian,
Spanish, Turkish]. https://havovwo.nl/pics/vmvtsyl22.pdf.
College voor Toetsen en Examens. (2022). Engels VWO 2022. https://www.examenblad.nl/examen/
engels-vwo-2/2022.
Cotton, D. R. E., Cotton, P. A., & Shipway, J. R. (2023). Chatting and cheating: Ensuring academic integ-
rity in the era of ChatGPT. Innovations in Education and Teaching International. https://doi.org/10.
1080/14703297.2023.2190148.
Davis, J. C., Lu, Y. H., & Thiruvathukal, G. K. (2023). Conversations with ChatGPT about
C programming: An ongoing study. Figshare. https://figshare.com/articles/preprint/
Conversations_with_ChatGPT_about_C_Programming_An_Ongoing_Study/22257274.
Frieder, S., Pinchetti, L., Griffiths, R. R., Salvatori, T., Lukasiewicz, T., Petersen, P. C., Chevalier, A.,
& Berner, J. (2023). Mathematical capabilities of ChatGPT. arXiv. https://doi.org/10.48550/
arXiv.2301.13867.
Geerling, W., Mateer, G. D., Wooten, J., & Damodaran, N. (2023). ChatGPT has mastered the principles
of economics: Now what? SSRN. https://doi.org/10.2139/ssrn.4356034.
Gilson, A., Safranek, C., Huang, T., Socrates, V., Chi, L., Taylor, R. A., & Chartash, D. (2022). How well
does ChatGPT do when taking the medical licensing exams? The implications of large language
models for medical education and knowledge assessment. medRxiv. https://doi.org/10.1101/2022.1
2.23.22283901.
Graham, F. (2022). Daily briefing: Will ChatGPT kill the essay assignment? Nature. https://doi.
org/10.1038/d41586-022-04437-2.
Han, Z., Battaglia, F., Udaiyar, A., Fooks, A., & Terlecky, S. R. (2023). An explorative assessment of
ChatGPT as an aid in medical education: Use it with caution. medRxiv. https://doi.org/10.1101/20
23.02.13.23285879.
Huang, F., Kwak, H., & An, J. (2023). Is ChatGPT better than human annotators? Potential and limitations
of ChatGPT in explaining implicit hate speech. Companion Proceedings of the ACM Web Confer-
ence, Austin, TX, 294–297. https://doi.org/10.1145/3543873.3587368.
Kasneci, E., Seßler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., Gasser, U., Groh, G.,
Günnemann, S., Hüllermeier, E., Krusche, S., Kutyniok, G., Michaeli, T., Nerdel, C., Pfeffer, J.,
Poquet, O., Sailer, M., Schmidt, A., Seidel, T., & Kasneci, G. (2023). ChatGPT for good? On oppor-
tunities and challenges of large language models for education. Learning and Individual Differences,
103, 102274. https://doi.org/10.1016/j.lindif.2023.102274.
Katz, D. M., Bommarito, M. J., Gao, S., & Arredondo, P. (2023). GPT-4 passes the bar exam. SSRN.
https://doi.org/10.2139/ssrn.4389233.
Kerrigan, J., Cochran, G., Tabanli, S., Charnley, M., & Mulvey, S. (2022). Post-COVID changes to assess-
ment practices: A case study of undergraduate STEM recitations. Journal of Educational Technology
Systems, 51, 192–201. https://doi.org/10.1177/00472395221118392.
Kim, N., Htut, P. M., Bowman, S. R., & Petty, J. (2022). (QA)2: Question answering with questionable
assumptions. ArXiv. https://arxiv.org/abs/2212.10003.

13
International Journal of Artificial Intelligence in Education

King, M. R. (2023). The future of AI in medicine: A perspective from a chatbot. Annals of Biomedical
Engineering, 51, 291–295. https://doi.org/10.1007/s10439-022-03121-w.
Kirmani, A. R. (2023). Artificial Intelligence-enabled science poetry. ACS Energy Letters, 8, 574–576.
https://doi.org/10.1021/acsenergylett.2c02758.
Kortemeyer, G. (2023). Could an artificial-intelligence agent pass an introductory physics
course? Physical Review Physics Education Research, 19, 010132. https://doi.org/10.1103/
PhysRevPhysEducRes.19.010132.
Kosinski, M. (2023). Theory of mind may have spontaneously emerged in large language models. arXiv.
https://doi.org/10.48550/arXiv.2302.02083.
Kung, T. H., Cheatham, M., Medenilla, A., Sillos, C., De Leon, L., Elepaño, C., Madriaga, M., Agga-
bao, R., Diaz-Candido, G., Maningo, J., & Tseng, V. (2023). Performance of ChatGPT on USMLE:
Potential for AI-assisted medical education using large language models. PLOS Digital Health, 2,
e0000198. https://doi.org/10.1371/journal.pdig.0000198.
Kuzman, T., Ljubešić, N., & Mozetič, I. (2023). ChatGPT: Beginning of an end of manual annotation?
Use case of automatic genre identification. arXiv. https://arxiv.org/abs/2303.03953.
LeCun, Y. (2023). Do large language models need sensory grounding for meaning and understanding? Spoiler:
YES! [Presentation]. https://drive.google.com/file/d/1BU5bV3X5w65DwSMapKcsr0ZvrMRU_Nbi/
view.
Lovin, B. (2022, December 3). ChatGPT produces made-up nonexistent references. https://brianlovin.
com/hn/33841672.
Mitchell, A. (2022, December 26). Professor catches student cheating with ChatGPT: ‘I feel abject terror’.
https://nypost.com/2022/12/26/students-using-chatgpt-to-cheat-professor-warns.
Newton, P. M., & Xiromeriti, M. (2023). ChatGPT performance on MCQ-based exams. EdArXiv. https://
doi.org/10.35542/osf.io/sytu3.
Office Microsoft Blog (2023). Introducing Microsoft 365 Copilot – your copilot for work. https://blogs.
microsoft.com/blog/2023/03/16/introducing-microsoft-365-copilot-your-copilot-for-work.
OpenAI (2023). GPT-4 technical report. https://cdn.openai.com/papers/gpt-4.pdf.
Pettit, M., Shukla, S., Zhang, J., Sunil Kumar, K. H., & Khanduja, V. (2021). Virtual exams: Has COVID-
19 provided the impetus to change assessment methods in medicine? Bone & Joint Open, 2, 111–118.
https://doi.org/10.1302/2633-1462.22.BJO-2020-0142.R1.
Reiss, M. V. (2023). Testing the reliability of ChatGPT for text annotation and classification: A cautionary
remark. arXiv. https://doi.org/10.48550/arXiv.2304.11085.
Rospocher, M., & Eksir, S. (2023). Assessing fine-grained explicitness of song lyrics. Information, 14,
159. https://doi.org/10.3390/info14030159.
Rudolph, J., Tan, S., & Tan, S. (2023). ChatGPT: Bullshit spewer or the end of traditional assessments
in higher education? Journal of Applied Learning and Teaching, 6. https://doi.org/10.37074/
jalt.2023.6.1.9.
Savelka, J., Agarwal, A., An, M., Bogart, C., & Sakr, M. (2023). Thrilled by your progress! Large Lan-
guage Models (GPT-4) no longer struggle to pass assessments in higher education programming
courses. arXiv. https://doi.org/10.48550/arXiv.2306.10073.
Šlapeta, J. (2023). Are ChatGPT and other pretrained language models good parasitologists? Trends in
Parasitology. https://doi.org/10.1016/j.pt.2023.02.006.
Sobania, D., Briesch, M., Hanna, C., & Petke, J. (2023). An analysis of the automatic bug fixing perfor-
mance of ChatGPT. arXiv. https://doi.org/10.48550/arXiv.2301.08653.
Susnjak, T. (2022). ChatGPT: The end of online exam integrity? arXiv. https://arxiv.org/abs/2212.09292.
Tabone, W., & De Winter, J. (2023). Using ChatGPT for human–computer interaction research: A primer.
Royal Society Open Science, 10, 231053. https://doi.org/10.1098/rsos.231053
Vincent, J. (2022, December 5). AI-generated answers temporarily banned on cod-
ing Q&A site Stack Overflow. https://www.theverge.com/2022/12/5/23493932/
chatgpt-ai-generated-answers-temporarily-banned-stack-overflow-llms-dangers.
Wang, X., Wei, J., Schuurmans, D., Le, Q., Chi, E., Narang, S., Chowdhery, A., & Zhou, D. (2023).
Self-consistency improves chain of thought reasoning in language models. Proceedings of the
International Conference on Learning Representations, Kigali, Rwanda. https://doi.org/10.48550/
arXiv.2203.11171.
Wei, J., Tay, Y., Bommasani, R., Raffel, C., Zoph, B., Borgeaud, S., Yogatama, D., Bosma, M., Zhou, D.,
Metzler, D., Chi, E. H., Hashimoto, T., Vinyals, O., Liang, P., Dean, J., & Fedus, W. (2022). Emergent
abilities of large language models. arXiv. https://doi.org/10.48550/arXiv.2206.07682.

13
 International Journal of Artificial Intelligence in Education

Whitford, E. (2022, December 9). A computer can now write your college essay — Maybe better than
you can. https://www.forbes.com/sites/emmawhitford/2022/12/09/a-computer-can-now-write-your-
college-essay---maybe-better-than-you-can/?sh=35deca9ddd39.
Zhai, X. (2022). ChatGPT user experience: Implications for education. ResearchGate. https://www.
researchgate.net/publication/366463233_ChatGPT_User_Experience_Implications_for_Education.
Zheng, C., Liu, Z., Xie, E., Li, Z., & Li, Y. (2023). Progressive-hint prompting improves reasoning in large
language models. arXiv. https://doi.org/10.48550/arXiv.2304.09797.
Zhong, Q., Ding, L., Liu, J., Du, B., & Tao, D. (2023a). Can ChatGPT understand too? A comparative
study on ChatGPT and fine-tuned BERT. arXiv. https://doi.org/10.48550/arXiv.2302.10198.
Zhong, W., Cui, R., Guo, Y., Liang, Y., Lu, S., Wang, Y., Saied, A., Chen, W., & Duan, N. (2023b). AGIEval:
A human-centric benchmark for evaluating foundation models. arXiv. https://doi.org/10.48550/
arXiv.2304.06364.

Supplementary Information All inputs (prompts) and outputs of ChatGPT, as well as the MAT-
LAB scripts used to access the API and create the figures can be found here: https://doi.
org/10.4121/545f8ead-235a-4eb6-8f32-aebb030dbbad.

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.

13