Towards LLM-Based Autograding for Short Textual Answers

Johannes Schneider

, Bernd Schenk

and Christina Niklaus

Department of Computer Science and Information Systems, University of Liechtenstein, Vaduz, Liechtenstein

School of Computer Science, University of St.Gallen, St.Gallen, Switzerland

Keywords: Grading Support, Autograding, Large Language Models, Trust.

Abstract: Grading exams is an important, labor-intensive, subjective, repetitive, and frequently challenging task. The

feasibility of autograding textual responses has greatly increased thanks to the availability of large language

models (LLMs) such as ChatGPT and because of the substantial influx of data brought about by digitalization.

However, entrusting AI models with decision-making roles raises ethical considerations, mainly stemming

from potential biases and issues related to generating false information. Thus, in this manuscript we provide

an evaluation of a large language model for the purpose of autograding, while also highlighting how LLMs

can support educators in validating their grading procedures. Our evaluation is targeted towards automatic

short textual answers grading (ASAG), spanning various languages and examinations from two distinct

courses. Our findings suggest that while “out-of-the-box” LLMs provide a valuable tool to provide a

complementary perspective, their readiness for independent automated grading remains a work in progress,

necessitating human oversight.

1 INTRODUCTION

Large language models like ChatGPT are said to be

“foundational” for many tasks having led to a

widespread impact across both industry and academia

(Schneider, Meske, et al., 2024). However, artificial

intelligence, including LLMs, is hard to understand

(Longo et al., 2023; Meske et al., 2022) and suffers

from security concerns (Schneider & Apruzzese,

2023). Experts also perceive substantial risks

associated with LLMs and have advocated for a

development moratorium on such technologies

(Future of Life, 2023). In academia, LLMs are used as

a tool by researchers and students to such an extent

that researchers themselves have called on journals to

clarify the allowable extent of AI-generated content in

scholarly papers (Tang, 2023), leading to the

publication of guidelines for incorporating AI in the

paper-writing process(Aczel & Wagenmakers, 2023).

Ethical concerns have also been raised for education

(Yan et al., 2023) and children (Schneider, Kruse, et

al., 2024). LLMs like ChatGPT have been commonly

compared against students in various disciplines –

especially with respect to their capability to pass

exams. While some reports have indicated inferior

performance than a master’s graduate in mathematics

(Frieder et al., 2023), other instances showcase a

successful completion of an introductory physics

course (Kortemeyer, 2023), as well as the passing of

numerous law school exams (Choi et al., 2023).

However, it is important to acknowledge the existence

of limitations in the LLMs. These models can exhibit

biases, discrimination, and factual inaccuracies (Borji,

2023). Consequently, there arises doubt regarding

their suitability in education. In particular, the

necessity for human verification has been emphasized

as a pressing research priority (van Dis et al., 2023)

and the topic of human agency is also debated on a

regulatory level (EU, 2020). Especially, high stakes

decisions require careful analysis before AI can be

utilized. Grading of exams is a high-stakes situation,

as errors in grading can cause students to fail an entire

class, possibly causing a year-long delay in their

education, separation of peers, etc. This, in turn, can

lead to both financial and psychological strain.

As such it seems natural and even necessary to

assess the suitability of LLMs for supporting exam

grading and to reflect upon adequate ways to include

them in the grading process while mitigating their

risks. To this end, we seek to contribute to two

intertwined research questions: (a) How can LLMs

support educators in the grading process of exams?

(b) What are issues and concerns when using LLMs

to support grading?

280

Schneider, J., Schenk, B. and Niklaus, C.

Towards LLM-Based Autograding for Short Textual Answers.

DOI: 10.5220/0012552200003693

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 1, pages 280-288

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

Our focus is on Automatic Short Answer Grading

(ASAG), i.e., student replies are (short) textual

answers (i.e., one or a few paragraphs). We use an

LLM, i.e., ChatGPT to assess the instructor’s answer,

a student’s answer in general as well as a student’s

answer with respect to the instructor’s answer as

illustrated in Figure 1. In our experimental evaluation,

we used two exams from two educators.

Figure 1: The three questions an LLM answers to support

the grading(left) and the information used by the LLM

(right) to answer them.

While we implicitly assess the possibility of

automatic grading, our target is (i) to improve the

grading process rather than simply automating it and

(ii) to uncover shortcomings (and possible

mitigations) of LLMs for this purpose. We seek to

employ LLMs as a second opinion that might

pinpoint obvious flaws in the grading, i.e., due to

sloppiness, in the grading as well as provide a more

general view on possible answers in order to avoid

bias like accepting only correct answers that have

been discussed in the course.

2 METHODOLOGY

We use an LLM, i.e., ChatGPT (GPT 3.5, June/July

2023 versions), to assess (i) answers by the educator,

(ii) answers of students to exam questions in general,

and (iii) answers of students compared to the

instructor’s answer (see Figure 1).

That is, to assess an instructor’s answer, an

instructor must be able to define an answer for each

exam question that constitutes the optimal response

from her/his perspective. We shall elaborate on lifting

this requirement and allow for multiple possible

answers per question in the discussion section.

Our assessment is both qualitative and

quantitative. That is, we highlight a few prompts that

surfaced surprising issues (such as lack of robustness,

i.e., sensitivity to minor variations in prompts), but

we also quantify how much the LLM deviates from

the educator across all graded answers. To this end,

we ask the LLM to categorize its assessment, i.e.,

each LLM response should contain a category such as

“Good”, ”Ok.”, or “Bad” and an explanation of the

chosen category. In turn, we also categorize the

educator’s responses. This allows us to compare the

categorization of the LLM and the educator, e.g., to

assess if both rate a student answer as “good”, which

we elaborate in the last subsection. We experimented

with a few different prompts but we report only the

ones used in the end.

2.1 Assessing the Educator’s Answer

An educator should have a good sense of different

answers to exam questions. However, it is often

difficult to anticipate all arguments and answers. For

example, an educator might be primed to answers that

are based on the material taught in class but students

might include elements with different but nonetheless

valid viewpoints. We utilize the LLM to provide a

more general view on the question without explicit

knowledge on the lecture material. We also ask the

LLM for feedback on the developed educator’s

solution. The prompt has the following structure for a

question Q and the educator’s answer A.

Here is a question: Q . Here is an answer: A. How

good is the answer to the question? Start the reply

with one of the following: Extremely good., Very

good., Good., Ok., Bad., Very bad., Extremely bad.

Explain the choice. Explain also what is missing.

Example Prompt (1) to LLM: Here is a question:

What is the difference between single linkage and

average linkage (hierarchical) clustering? . Here is

an answer: The two differ in distance metric used to

cluster. Single linkage: Merge two clusters based on

minimum distance between any two points; Tendency

to form long chains; Average linkage: merge two

clusters based on average distance between any two

points; tendency to “ball” like clusters;. How good is

the answer to the question? Start the reply with one

of the following: Extremely good., Very good., Good.,

Ok., Bad., Very bad., Extremely bad. Explain the

choice. Explain also what is missing.

Response of LLM: Good. The answer provides a

clear and concise explanation of the difference

between single linkage and average linkage

clustering. It accurately states the different distance

metrics used and describes the tendencies of each

clustering method. One potential improvement is to

provide a more detailed explanation of the

advantages and disadvantages of each method.

Additionally, it would be helpful to include an

example or illustration to further clarify the concepts.

2.2 Assessing the Student’s Answer

We let the LLM assess the student’s answer to a

question. The prompt has the following structure for

Towards LLM-Based Autograding for Short Textual Answers

281

a question Q and a student answer A.

Here is a question: Q . Here is an answer: A. How

good is the answer to the question? Start the reply

with one of the following: Extremely good., Very

good., Good., Ok., Bad., Very bad., Extremely bad.

Explain the choice.

Example Prompt to LLM: Here is a question:

What is the difference between single linkage and

average linkage (hierarchical) clustering? . Here is

an answer: In single linkage, we compare the two

closest data points (the ones with minimal distance)

from two separate clusters. In average linkage, we

compare all the data points from a cluster with all the

datapoints from another cluster and get an average

distance. .How good is the answer to the question?

Start the reply with one of the following: Extremely

good., Very good., Good., Ok., Bad., Very bad.,

Extremely bad. Explain the choice.

Response of LLM: Good. The answer provides a

clear and concise explanation of the difference

between single linkage and average linkage

clustering. It accurately describes how single linkage

compares the closest data points from two separate

clusters, while average linkage compares all data

points from a cluster with all data points from another

cluster and calculates the average distance. The

answer effectively addresses the question and

provides a solid understanding of the topic.

2.3 Comparing Student’s and

Educator’s Answer

We compare the student’s answer with the educator’s

answer. The LLM assesses the similarity between the

student answer and the educator’s answer, i.e., the

optimal answer from the perspective of the educator.

The LLM only obtains the educator’s and the

student’s answer as input but not the question (see

also Figure 1). We found that when the question is

added, the assessment of ChatGPT sometimes tended

to neglect the optimal answer (although the prompt

clearly stated that the task was to compare the student

reply and the educator’s answer).

Thus, the LLM provides a second opinion like

another grader that had also knowledge on the course

in the form of what is the best answer given the taught

material, i.e., the educator’s answer. Thus, in contrast

to the prior assessments that were seeking for a more

general perspective, here, we ultimately seek to

assess, if an educator’s grading matches that of the

LLM given knowledge of the best response. That is if

both should stick to a pre-defined “ground truth” in

the form of the educator’s answer. We shall also

discuss how to relate the grading of the LLM and the

educator quantitatively in the next section. As a

prerequisite to derive quantitative statements, we ask

for a category but also for an explanation to assess

qualitatively. More concretely, the prompt has the

following structure for a student answer A and the

(assumed to be optimal) educator’s answer E.

Here is an answer: A . Here is the optimal answer:

E. How close is the answer to the optimal answer?

Start the reply with one of the following: Very close.,

Close., Somewhat close., Somewhat distant., Distant.,

Very distant.. Explain the choice.

Example Prompt: Here is an answer: In single

linkage, we compare the two closest datapoints (the

ones with minimal distance) from two separate

clusters. In average linkage, we compare all the

datapoints from a cluster with all the datapoints from

another cluster and get an average distance. .

Here is the optimal answer: The two differ in

distance metric used to cluster. Single linkage: Merge

two clusters based on minimum distance between any

two points; Tendency to form long chains; Average

linkage: merge two clusters based on average

distance between any two points; tendency to “ball”

like clusters; How good is the answer to the question?

Start the reply with one of the following: Extremely

good., Very good., Good., Ok., Bad., Very bad.,

Extremely bad. Explain the choice. Explain also what

is missing. What is the difference between single

linkage and average linkage (hierarchical)

clustering?

How close is the answer to the optimal answer?

Start the reply with one of the following: Very close.,

Close., Somewhat close., Somewhat distant., Distant.,

Very distant.. Explain the choice.

Response of LLM: Very close. The given answer

effectively highlights the key differences between

single linkage and average linkage in hierarchical

clustering. It mentions that single linkage does not

allow linkage between different groups, while

average linkage splits the dataset into average-sized

groups. Overall, the given answer accurately

addresses the main distinction, making it very close

to the optimal answer.

2.4 Highlighting Deviations of LLM

and Educator

The prior assessments are helpful, but require a lot of

effort for the educator, i.e., the educator must

manually assess all responses of the LLM. It might be

desirable to focus more on large differences between

human grading and the reply of the LLM.

Furthermore, quantifying differences allows for a

quantitative assessment, which makes it more explicit

CSEDU 2024 - 16th International Conference on Computer Supported Education

282

how well aligned the educator’s grading and the

LLM’s grading are. We assume that an educator has

provided a score in some range [0,maxPts], where

maxPts gives the maximum points attainable by a

student. For easier interpretability, we scale the scores

to be between 0 and 1. We denote the scaled points

given by the human, i.e., educator, as p

. We also

compute a score given by the LLM denoted as p

based on the LLM’s responses. We compute

correlations to understand the alignment between

human and LLM. We sort the outcomes based on how

different the LLM’s ratings are from the educator’s

grading, which allows an educator to focus on the

largest differences. The deviations are computed as

follows. We assume that if the LLM replies for a

given student answer and the educator’s answer with

“very close” then that student should have gotten

maximal points, whereas if the response is “very

distant” the student should have gotten no points.

Similarly, if the LLM rates a student answer for a

given exam question as “extremely good” then the

student should have gotten maximal points. If it is

rated “extremely bad” then no points are given.

Inbetween, we scale accordingly, i.e., we linearly

interpolate to obtain the points an LLM would assign

for other categories. We also scale by maxPts to have

a normalized value between 0 and 1 for the LLM. We

denote the scaled points by p

. We then sort outcomes

based on the gap between the human’s and LLM’s

scores | p

- p

|. This allows to focus on responses

where the educator and LLM disagree the most, i.e.,

we sort the LLM responses from largest to lowest

differences. For easier comparison we show both the

educator’s (or student’s) response and the LLM’s

response. This enables us to read the answer of the

human, i.e., student or educator, and then the reply of

the LLM. If the two responses disagree significantly,

and indeed the LLM had a valid point, then an

educator should alter his/her grading.

An example output is:

Gap: 0.9 LLM Pts p_l: 0.1 Human Pts p_h: 1.0

Answer Human: When using single linkage in

hierarchical clustering, the clusters are made with

the shortest distance between the data point the

closest cluster. Average linkage on the other hand

takes the maximum distance and the minium distance

to each cluster, evaluates the average and then builds

the clusters upon these average distances.

Answer LLM: Somewhat close. The answer

provides a basic understanding of the differences

between single linkage and average linkage in

hierarchical clustering. It correctly states that single

linkage considers the shortest distance between a

data point and the closest cluster, whereas average

linkage calculates the average distance to each

cluster. However, it does not fully explain how

average linkage clusters are formed based on these

average distances.

3 ASSESSMENT SETUP

We assess two exams administered by two different

educators. One exam was in English at the master

level in data science with 21 participants. It consisted

of 16 questions and had a total of 90 points. It lasted

90 minutes. The other was in German at the bachelor

level in an information systems introductory course

with 34 participants. There was a multiple choice part

that was not considered, i.e., we were left with 3

questions each worth 10 points resembling 10

minutes of processing time per question, i.e., the

replies were expected to be longer than for the

questions of the data science exam. The types of

questions covered all categories according to the

revised Bloom’s taxonomy of educational

objectives(Anderson & Krathwohl, 2001). The

taxonomy ranges from simple, concrete to complex,

abstract questions. Our exams contained some simple

concrete questions related to understanding and

remembering, such as providing definitions and

examples. More complex, abstract questions

consisted, for example, of evaluating different

approaches for a practical problem. We read through

all of ChatGPT’s responses.

4 FINDINGS

We first discuss overarching findings before

elaborating on each of the three questions in Figure 1.

The LLM Replies Generically. ChatGPT tends to

assess in a mechanistic generic manner rather than

looking at content. It might respond like “There is not

sufficient detail” rather than pointing to specific

details that are missing.

The LLM and Human Assessments Differ Strongly.

The correlation between human and LLM’s

judgments is small. Generally, the LLM’s judgments

have a strong tendency to the middle, e.g., most are

“ok” or “good” despite strong variation in the quality

of student replies.

The LLM can Help to Make Sense of Hard to

Understand Answers. The LLM provided a more

open and less negative view on responses suffering

from poor language. Thus, the assessment was

Towards LLM-Based Autograding for Short Textual Answers

283

particularly positive for students with poor (English)

language skills, as ChatGPT tended to rate them

comparatively better to an educator. That is, a human

might rate them poorly because the answer is difficult

to understand or remains unclear due to grammatical

ambiguities or poor wording. We also found that it

was sometimes easier to make sense of a student reply

after reading ChatGPT’s assessment. Furthermore,

commonly specific concepts tied to a set of keywords

are accepted or looked for. If students do not provide

any of these but rather a lengthy and verbose reply,

there is a higher risk that possibly correct though

convoluted arguments are overlooked. We found that

ChatGPT’s assessment can be helpful, since it can

transform hard to grasp answers into a more concise

phrasing and its responses follow an expected

structure, which is fast to process for an educator.

The LLM can Drastically Change its Assessment

Due to Minor Changes in Answers. Additional

content that is strikingly wrong though not related to

the question (or answer) can lead to dramatic changes

in judgements by the LLM. For illustration, we

appended to the answer of the student used in the

example prompt (1) either of the following three

options:

(i) 3*5=7,

(ii) the cat sits on the mattress,

(iii) 3*5=7, the cat sits on the mattress;

ChatGPT judged two of them equivalently as the

original prompt (1), i.e. as good. For (ii) and (iii) it

would mention that the answers contain irrelevant

information, but (ii) is still judged as good by the

LLM, while the LLM judged the response (iii) as

“very bad”.

The LLM Favors Vague Content and Fails to

Recognize Contradictions. Generally, replies with

vaguely related content, which might be deemed

irrelevant or even incorrect by a human grader, is

rated more favorably by the LLM than by human

graders. We also found that ChatGPT can fail to

distinguish contradicting statements. We appended to

prompt (1) either of the following:

(i) Complete linkage uses the minimum distance

between any two points in clusters. Density based

clustering relies on computing the number of points,

possibly organizing them in a search tree or a list.

(ii) Complete linkage uses the maximum distance

between any two points in clusters. Density based

clustering relies on computing point densities, e.g.

points for a fixed volume, for the volume for a fixed

set of points.

Note, the words minimum and maximum are

switched in (i) and (ii). The LLM judged (i) and (ii)

equally, although they obviously contain

contradicting statements and information not being

asked for.

The LLM Misunderstands Questions. ChatGPT can

suggest to provide information that can obviously be

ruled out as being asked for. For the question “What

are advantages of a decision tree?” (and a student’s

answer) the LLM’s reply included “However, what is

missing from the answer is a mention of some

potential disadvantages or limitations of decision

trees.”

The LLM’s Grading Criteria are Language

Sensitive. We applied the same prompt patterns for

both exams, i.e., we utilized the English prompt

pattern for the German exam. While at first, this did

not seem to pose a problem, we found that ChatGPT

occasionally provides a lower rating giving as reason

that (German) texts contain grammar and spelling

issues, but this would not happen for answers in

English.

Figure 1: Ratings of educator's answers by LLM.

4.1 Findings on Assessing the

Educator’s Answer by the LLM

We read through all of ChatGPT’s responses. None

of them led to any changes of the human-crafted

responses. Most suggested improvements were

generic, e.g., related to giving more details, an

example and sometimes visualization or limitations.

ChatGPT’s responses were quite sensitive to the

phrasing (and potentially other factors). For example,

omitting the term “Explain also what is missing.”

changed the LLM’s response for one reply from “very

bad” (see Figure 2) to “good”, while still giving

mostly the same reasoning. Overall, Figure 2 suggests

that the educator provided mostly “very good”

answers and no answer was below “good” (at least

CSEDU 2024 - 16th International Conference on Computer Supported Education

284

when slightly changing the prompt as mentioned

before).

4.2 Findings on Assessing the Student’s

Answer in General and Relative to

the Educator’s Answer

Here, the LLM had more impact on the grading. That

is, we made minor adjustments after reading through

the LLM’s assessment. The adjustments were made

due to two types of replies: First, the LLM would rate

student answers (more) positively that were not part

of the lecture material and also not directly being

asked for. For example, we asked “Which two of the

factors ‘data, compute and algorithms’ are most

important for the rise of GPT-3 (around 2020) and

ChatGPT in 2022 and other generative AI models?”

Some students responded that one factor was media

coverage and accessibility due to its public release.

ChatGPT rated the responses of these students

positively, although (1) the question explicitly

restricts the factors to be discussed and (ii) the

importance of media coverage is debatable – at least

for GPT-3. That is, in the lecture, it was mentioned

for ChatGPT that its public release led to widespread

adoption and a surge in media coverage, but not so

much for GPT-3. GPT-3 was covered less in the

media, and it was less accessible, i.e., only (some)

scientists got access.

Figure 2: Ratings of students’ answers by LLM.

Second, the LLM would more positively rate

replies with poor English. That is, the LLM’s

interpretation of the answer made the student answer

more understandable. For an educator, the quality of

the answer needs to exceed a minimum level of

proficiency to be understood. Comprehensibility is

generally a factor influencing grading. Educators are

not supposed to make too many assumptions about

what a student wants to say (i.e. interpret) but they

have to stick with the incomprehensible answer and

grade accordingly.

Overall, we found that any judgement of the

grading after consulting the LLM was preceded by

considerable reflection and debates, and it was not

evident whether the differences of the LLM should

really be considered.

Interestingly, using an answer set of an exam

conducted in German, the LLM incorporated errors in

spelling and grammar in the feedback and

downgraded answers of poor language quality.

Figure 3: Comparison of students and educator’s answers

by LLM.

The LLM tended to rate most student responses as

“good” or “very good” (Figure 2), i.e., there was little

differentiation. This is in stark contrast to the rating

of the educator (Figure 3). The educator scored many

answers with maximum or minimum points but

he/she also assigned commonly points in-between the

two extremes. The extremes were mostly common for

short and easy answers with few points.

When it comes to assessing similarity between the

educator’s and the students’ answers the LLM gave

somewhat more diverse replies. However, overall

alignment was poor. The Pearson correlation between

the LLM’s similarity assessment p

and educator’s

grading p

was close to 0.

Figure 4: Distribution of frequency (y-axis) of normalized

points by educator (x-axis).

Towards LLM-Based Autograding for Short Textual Answers

285

5 DISCUSSION

We set out to assess LLMs for autograding, primarily

as a second opinion as for high stakes decision

regulation also demands human oversight. Specifics

of the course have not been provided to the LLM, e.g.,

the course material for which the exam was made for.

That is, the LLM lacked any lecture specific context.

It relied on world knowledge incorporated in its

training data. Thus, discrepancies between the

judgments of the LLM and lecturer are expected, e.g.,

as many terms are defined differently in other

contexts and fields. Such contextualization of terms

seems to be an essential part of teaching and learning

and allows students to establish different perspectives

on issues. However, for grading, we believe that the

lack of context by the LLM can be highly valuable, as

it provides a strongly complementary view that aims

to avoid strong biases of a lecturer towards the lecture

material. Still, this also hints that grading (or possibly

even exam questions) derived by an LLM given

access to the course material could provide a

welcome addition to be investigated in future work.

We also faced a number of practical issues when

using LLMs. For example, the LLM’s replies would

not always follow the given structure, i.e., ChatGPT

would reply with any of the asked for words “Very

good”, ”Good” etc. but started the reply with some

other sentence. This problem can often be mitigated

by providing a few examples of inputs and desired

outputs (in-context learning). However, doing so

means additional work for the educator, increases

response time of the LLM and also costs, i.e., longer

input prompts imply higher costs for commercial

LLMs.

Our experimentation with prompting surfaced

trade-offs. For example, when comparing the

student’s and the educator’s answer, we tested

prompts that included the question (as well as the

answers of the student and educator) and prompts that

did not. We found that without the question,

ChatGPT’s assessment sometimes tended to include

aspects that were rather unrelated to the question. If

the question was added, the assessment of ChatGPT

sometimes did not consider the educator’s answer.

It is also tempting to use LLMs for fully automatic

grading. However, from our experience this should

not be undertaken at the current point in time since

there is very strong disagreement between gradings of

educators and LLMs. That is, they perform

significantly worse in judging questions than in

providing responses. This might be improved using

techniques such as in-context learning, i.e., providing

examples on how answers should be graded, or fine-

tuning LLMs specifically towards autograding.

However, first experimentation did not yield the

hoped performance boost. In general, finding the best

prompts for grading is non-trivial and responses could

be sensitive to the slightest changes in phrasing.

Grading should be robust, fair, and consistent.

Accordingly, the achievement of competency levels

of students should be assessed as independently as

possible of individual course delivery, of lecturers

and examiners, and of the performance of other

students in an exam. ChatGPT did not (yet) meet

these requirements in our evaluation.

We assessed the idea to focus on answers where

the LLM and the human showed largest

discrepancies. However, unfortunately, ChatGPT’s

rating was not too well-aligned with that of the

educator. Furthermore, if not all answers are checked

(but only those with large differences), biases in the

LLM might further impact the grading by leading to

a bias in which answers are looked at (again) by a

human grader. Furthermore, biases also appear as

“misleading clues”. If, for example, the LLM judges

arguments identical to a human, except for an

argument A, then students using A are more likely to

show a large gap (even if aligned with the educator)

and thus being assessed by the educator.

One assessment within our work assumed that an

educator provides a single answer to a question. In

principle, a question might permit fairly different

valid answers. However, it is not hard to allow for

multiple responses, i.e., an educator could define

various answers that might even be contradictory. We

could then compare a student’s answer with all of the

educator’s responses and focus on the educator’s

response that is deemed closest by the LLM.

However, specifying answers becomes more difficult

the more open-ended a question is, i.e., the more

knowledge should be applied and transferred, as

opposed to simply replicating knowledge.

From an ethical point of view, one might also

debate whether changes due to LLMs should only

improve grades. That is, LLMs should not be allowed

to fail a student, as punishing innocent people can be

seen as worse than rewarding people not having

deserved it. Furthermore, using an LLM as “a second

opinion” might also provide a false sense of security.

6 FUTURE WORK

We might add a more explicit grading scheme that

aims to identify specific aspects in the answer, i.e., “Is

this concept in the answer?” (If not deduct x points).

CSEDU 2024 - 16th International Conference on Computer Supported Education

286

Furthermore, a fine-tuned LLM towards grading

might lead to better outcomes than relaying on

prompting. To this end, large number of graded

exams would be needed. While graded exams already

exist, sharing them is non-trivial as responses might

have to be anonymized to comply with privacy

regulations.

LLMs might also be useful for exam

development, i.e., assessing questions prior to posting

an exam. One might also provide access to the lecture

material to the LLM to assess gradings. This might

uncover more minor issues in the grading scheme, but

might not help so much in uncovering general issues.

In this study, we used LLM only in grading answers

on questions that have been formulated by lecturers.

It would be interesting to test the end-to-end support

by LLMs in designing a lecture, including the

selection of topic areas, creating the lecture material,

and preparing and assessing the exam.

7 RELATED WORK

The manual grading process involves a labor-

intensive evaluation of students’ responses, requiring

expertise and careful judgment to assign appropriate

scores. Thus, to assist educators in reducing the time

and effort spent on grading, there is a growing interest

in leveraging AI-driven correction aids (Basu et al.,

2013; Condor et al., 2021). When comparing the

conventional teacher’s judgement (“human scoring”)

to the capabilities auf automatic feedback and

assessment tools (“machine scoring”), we encounter

distinct strengths along various quality criteria

(Seufert et al., 2022), i.e., AI can support objectivity,

reliability, validity and comparative values and

standards.

The evolution of assessment methodologies is

currently exploring hybrid solutions that harness the

strengths of both mechanisms. These developments,

such as AI-based assistants for assessment and learner

feedback, hold promise for the future education,

offering more efficient and objective evaluation

processes while maintaining the depth of

understanding provided by human judgement (Saha

et al., 2018; Schneider et al., 2023). A few works have

also assessed the value of feedback through

autograding, e.g., (Vittorini et al., 2020) also assesses

the value of feedback provided by the autograder for

students. (Li et al., 2023) investigated the effects of

AI grading mistakes on learning showing that, in

particular, marking wrong answers as right had a

negative impact on learning

We concentrate on the field of Automatic Short

Answer Grading (ASAG) (Burrows et al., 2015). It

deals with grading student answers, typically ranging

from a phrase to a paragraph. ASAG also covers the

grading of open-ended answers (Baral et al., 2021).

The primary focus in ASAG is on content quality

rather than the writing style and structure emphasized

as in automatic essay scoring (AES) (Dikli, 2010).

For ASAG, prior work has mostly relied on BERT

as a large language model (Baral et al., 2021; Haller

et al., 2022; Schneider et al., 2023; Sung et al., 2019;

Zhang et al., 2022). Schneider et al. (2023)

investigated how LLMs such as BERT suffer from

trust issues that might be partially mitigated by only

automatically grading answers, if the LLM is certain

about its grading.

While LLMs can provide justifications for their

decisions without any additional work, dedicated

methods for enhancing explainability have been

evaluated in the realm of automatic grading systems

(Kumar & Boulanger, 2020). Efforts have also been

made to tackle the limitations of AI in terms of

fairness in the context of automatic grading (Madnani

et al., 2017) and, more generally, ethical issues

related to LLMs in education (Yan et al., 2023).

8 CONCLUSIONS

The integration of LLMs into academic settings have

become an undeniable reality. These models possess

remarkable linguistic capabilities, coupled with

unexpected reasoning abilities. Yet, using LLMs such

as ChatGPT “out-of-the-box” to support grading

requires great care due to a multitude of issues such

as sensitivity to minor changes in answers and lack of

concise reasoning, which is also reflected in poor

alignment with human graders. Despite these

limitations, LLMs currently offer a valuable resource

that provides a supplementary viewpoint with

minimal effort.

REFERENCES

Aczel, B., & Wagenmakers, E.-J. (2023). Transparency

Guidance for ChatGPT Usage in Scientific Writing.

PsyArXiv. https://doi.org/10.31234/osf.io/b58ex

Anderson, L. W., & Krathwohl, D. R. (2001). A taxonomy

for learning, teaching, and assessing: A revision of

Bloom’s Taxonomy of Educational Objectives.

Longman.

Baral, S., Botelho, A., Erickson, J., Benachamardi, P., &

Heffernan, N. (2021). Improving Automated Scoring of

Towards LLM-Based Autograding for Short Textual Answers

287

Student Open Responses in Mathematics. Proceedings

of the Int. Conference on Educational Data Mining.

Basu, S., Jacobs, C., & Vanderwende, L. (2013).

Powergrading: A Clustering Approach to Amplify

Human Effort for Short Answer Grading. Transactions

of the Association for Computational Linguistics, 1,

391–402. https://doi.org/10.1162/tacl_a_00236

Borji, A. (2023). A categorical archive of chatgpt failures.

arXiv Preprint arXiv:2302.03494.

Burrows, S., Gurevych, I., & Stein, B. (2015). The eras and

trends of automatic short answer grading. International

Journal of Artificial Intelligence in Education.

Choi, J. H., Hickman, K. E., Monahan, A., & Schwarcz, D.

(2023). Chatgpt goes to law school. Available at SSRN.

Condor, A., Litster, M., & Pardos, Z. (2021). Automatic

short answer grading with bert on out-of-sample

questions. Proceedings of the 14th International

Conference on Educational Data Mining.

Dikli, S. (2010). The nature of automated essay scoring

feedback. Calico Journal, 28(1), 99–134.

EU. (2020). White Paper on Artificial Intelligence: A

European Approach to Excellence and Trust.

Frieder, S., Pinchetti, L., Griffiths, R.-R., Salvatori, T.,

Lukasiewicz, T., Petersen, P. C., Chevalier, A., &

Berner, J. (2023). Mathematical capabilities of chatgpt.

arXiv Preprint arXiv:2301.13867.

Future of Life. (2023). Pause Giant AI Experiments: An

Open Letter. https://futureoflife.org/open-letter/pause-

giant-ai-experiments/

Haller, S., Aldea, A., Seifert, C., & Strisciuglio, N. (2022).

Survey on automated short answer grading with deep

learning: From word embeddings to transformers.

arXiv Preprint arXiv:2204.03503.

Kortemeyer, G. (2023). Could an Artificial-Intelligence

agent pass an introductory physics course? arXiv

Preprint arXiv:2301.12127.

Kumar, V., & Boulanger, D. (2020). Explainable automated

essay scoring: Deep learning really has pedagogical

value. Frontiers in Education, 5, 186.

Li, T. W., Hsu, S., Fowler, M., Zhang, Z., Zilles, C., &

Karahalios, K. (2023). Am I Wrong, or Is the

Autograder Wrong? Effects of AI Grading Mistakes on

Learning. Proceedings of the Conference on

International Computing Education Research

Longo, L., Brcic, M., Cabitza, F., Choi, J., Confalonieri, R.,

Del Ser, J., Guidotti, R., Hayashi, Y., Herrera, F.,

Holzinger, A., & others. (2023). Explainable artificial

intelligence (XAI) 2.0: A manifesto of open challenges

and interdisciplinary research directions. arXiv Preprint

arXiv:2310.19775.

Madnani, N., Loukina, A., Von Davier, A., Burstein, J., &

Cahill, A. (2017). Building better open-source tools to

support fairness in automated scoring. Proc. of the

Workshop on Ethics in Natural Language Processing.

Meske, C., Bunde, E., Schneider, J., & Gersch, M. (2022).

Explainable artificial intelligence: Objectives,

stakeholders, and future research opportunities.

Information Systems Management, 39(1), 53–63.

Saha, S., Dhamecha, T. I., Marvaniya, S., Sindhgatta, R., &

Sengupta, B. (2018). Sentence Level or Token Level

Features for Automatic Short Answer Grading?: Use

Both. In C. Penstein Rosé & others (Eds.), Artificial

Intelligence in Education (Vol. 10947).

Schneider, J., & Apruzzese, G. (2023). Dual adversarial

attacks: Fooling humans and classifiers. Journal of

Information Security and Applications, 75, 103502.

Schneider, J., Kruse, C., Leona, & Seeber, I. (2024).

Validity Claims in Children-AI Discourse: Experiment

with ChatGPT. Proceedings of the International

Conference on Computer Supported Education

(CSEDU 2024).

Schneider, J., Meske, C., & Kuss, P. (2024). Foundation

Models. Business Information Systems Engineering.

Schneider, J., Richner, R., & Riser, M. (2023). Towards

trustworthy autograding of short, multi-lingual, multi-

type answers. International Journal of Artificial

Intelligence in Education, 33(1), 88–118.

Seufert, S., Niklaus, C., & Handschuh, S. (2022). Review

of AI-Enabled Assessments in Higher Education.

https://www.alexandria.unisg.ch/handle/20.500.14171/

109523

Sung, C., Dhamecha, T., Saha, S., Ma, T., Reddy, V., &

Arora, R. (2019). Pre-training BERT on domain

resources for short answer grading. Proceedings of the

Conf. on Empirical Methods in Natural Language

Processing and the Int. Joint Conf. on Natural Language

Processing (EMNLP-IJCNLP).

Tang, G. (2023). Letter to editor: Academic journals should

clarify the proportion of NLP-generated content in

papers. Account. Res. https://doi.org/10.

1080/08989621.2023.2180359

van Dis, E. A., Bollen, J., Zuidema, W., van Rooij, R., &

Bockting, C. L. (2023). ChatGPT: five priorities for

research. Nature, 614(7947), 224–226.

Vittorini, P., Menini, S., & Tonelli, S. (2020). An AI-Based

System for Formative and Summative Assessment in

Data Science Courses. International Journal of

Artificial Intelligence in Education, 1–27.

Yan, L., Sha, L., Zhao, L., Li, Y., Martinez-Maldonado, R.,

Chen, G., Li, X., Jin, Y., & Gašević, D. (2023).

Practical and ethical challenges of large language

models in education: A systematic scoping review.

British Journal of Educational Technology.

Zhang, L., Huang, Y., Yang, X., Yu, S., & Zhuang, F.

(2022). An automatic short-answer grading model for

semi-open-ended questions. Interactive Learning

Environments, 30(1), 177–190.

CSEDU 2024 - 16th International Conference on Computer Supported Education

288