A Comparative Study on The Performance of ChatGPT-4 and

Claude 3 in Translating Engineering Specialized Course Textbooks

Yiqin Jiang

Institute of International and Comparative Education, East China Normal University, Shanghai, 200062, China

Keywords: AIGC, Machine Learning, Translation Applications.

Abstract: This study examines the advantages and disadvantages of Claude 3 and ChatGPT-4 for translating engineering

texts. Both AI technologies analyze scientific texts from four engineering textbooks for accuracy, fluency,

audience appropriateness and logicality using the CO-STAR framework. The translation outcomes of AI tools

were contrasted and examined with those of rule-based machine translation and human translation of the four

paragraphs. Overall, Claude 3 performs better than ChatGPT-4, according to the results, and ChatGPT-4 has

to be enhanced in terms of audience adaption and fluency. Nevertheless, the overall quality of the translations

is inferior to that of human translators, and both AI systems have representational flaws. This study

emphasizes the need for improvement in the distribution of educational resources across languages, as well

as the potential of AI technologies for translating engineering textbooks.

1 INTRODUCTION

The sharing of educational resources across

languages has become a significant force to promote

education and has drawn the attention of many

nations as a result of the internationalization of

education and the ongoing advancement of

communication technologies (Ahmadova, 2024). The

process of teaching knowledge and concepts for the

practice of engineering professions is known as

engineering education (Kadhim & Hussein, 2024).

One of the key issues facing engineering education is

globalization, which requires engineers to possess a

variety of cultural and communication abilities

(Ahrenberg, 2017). Thus, there is an urgent need for

learning materials for engineering education to be

shared and distributed throughout the world.

Nonetheless, there are numerous issues with the way

specialized engineering course textbooks are now

translated, and when translating scientific literature,

it's critical to consider linguistic and cultural

variations (Balas et al,2024). Engineering textbooks'

precise terminology and theoretical explanations

make it possible for manual translation or rule-based

machine translation to highlight its shortcomings,

such as inefficiency and low accuracy (Simpson &

Weiner,1989; Duderstadt, 2007). This makes it

difficult for engineering education learning resources

to be disseminated internationally. As generative

artificial intelligence advances, it opens up new

avenues for translating engineering textbooks and

offers a revolutionary solution to the aforementioned

issues.

It has been discovered that Artificial Intelligence

Generated Content (AIGC) technology is extensively

utilized in the translation industry in addition to being

utilized in the domains of cross-linguistic

communication and cultural exchange (Ferrag &

Bentounsi, 2024). In addition to having higher

translation quality and efficiency than human

translation, Artificial Intelligence (AI) translation is

superior to machine translation (Haq et al, 2024; Li et

al, 2024). To encourage global scientific and

technical communication, AIGC technology can raise

the translation quality in the interim (Li, 2024). AI

translators do have some limitations, though: first,

they are not as good at handling logical expressions

and textual connotations (Li, 2024); second, they are

not as good at adapting to different fields of

specialization (Li,2022), and third, while the quality

of translations is on par with intermediate human

translators, they still fall short of advanced translators

(Lommel et al, 2024). Numerous research have also

been conducted on the efficacy of AI translation

applications in several specialized domains. Artificial

intelligence (AI) does well in the literary domain

when it comes to word choice and narrative, but it

struggles to handle cultural and subtle emotional

distinctions (Mohsen, 2024). AI translations are

454

Jiang, Y.

A Comparative Study on the Performance of ChatGPT-4 and Claude 3 in Translating Engineering Specialized Course Textbooks.

DOI: 10.5220/0013999300004912

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Innovative Education and Social Development (IESD 2025), pages 454-460

ISBN: 978-989-758-779-5

typically less accurate than human translators in the

legal domain (Quigley et al,2011). Studies in the

medical field have also assessed and contrasted the

performance of various AI tools in translating

educational materials for neurological patients and

ophthalmic terminology, with ChatGPT performing

better in the latter case and Claude performing better

in the former (Shokoohifar, 2024; Wang, 2023).

Thus, it is yet unclear and worthwhile to investigate

if these two AI techniques can translate engineering

texts. It is evident that the majority of current research

is concentrated on AI translation in the fields of law,

medical, and literature, whereas engineering research

is at a lower level. Although the scientific text is

primarily based on objective facts, it is still influenced

by cultural differences in terms of expression,

according to the author, who believes that the use of

AI tools to translate engineering textbooks will face

many challenges. First, the text is extremely

specialized, and arbitrary additions and omissions

could cause it to lose its original meaning; second,

even though the scientific text is primarily grounded

in objective facts, cultural differences in expression

still have an impact on it (Wang,2023). In order to

provide a higher-quality, more accurate translation,

this research will investigate the benefits and

drawbacks of AI technologies in engineering

textbook translation.

Among the various AI tools used in this study,

ChatGPT-4 and Claude 3 were selected for two

reasons: first, they are the most popular in the

translation field when compared to other AI tools like

Kimi; second, ChatGPT-4 does well when translating

scientific texts (Yan et al, 2024), and Claude does

better than ChatGPT when translating into English

(Yan & Zhao, 2023). As a result, both are equally

capable of translating technical texts. In this study,

the translation skills and outcomes of these two AI

systems for specialized engineering course texts will

be evaluated and compared for similarities and

differences. This study is innovative in two ways:

first, it closes the knowledge gap regarding the

potential and constraints of AI tools for translating

specialized engineering science texts; second, it

optimizes the research methodology by using the

Multidimensional Quality Metrics (MQM) evaluation

framework as a foundation for screening relevant

evaluation indicators. The significance of this study

is threefold: 1. To encourage the exchange of

educational resources across languages and to

compare the strengths and shortcomings of Claude 3

and ChatGPT-4 in translating engineering textbooks

in order to enhance their quality. 2. Research findings

have the potential to foster creativity and

collaboration in international education. 3. This study

can also offer fresh perspectives on the investigation

of engineering talent development in Science,

Technology, Engineering, Mathematics (STEM)

education within an interdisciplinary setting.

2 RESEARCH METHOD

This study compares the translation capabilities of

Claude 3 and ChatGPT-4 for engineering texts. For

this study, two textbooks that are fairly representative

of the Chinese-US Mechanical Design and

Manufacturing and Automation joint engineering

curriculum at Shanghai Normal University were

selected. These two texts are Fluid Mechanics

Fundamentals and Applications (2nd edition) and

Thermodynamics: An Engineering Approach (7th

edition). The author chose two scientific texts from

each engineering textbook that addressed specialized

terminology and the background of specialized

domain knowledge for additional AI translations in

order to assess whether these two AI tools could

successfully negotiate the numerous challenges

associated with translating engineering textbooks. To

assess the quality of the AI translations, the author

included the translation results in a questionnaire that

was given to Shanghai Normal University

engineering majors, ranging from freshmen to

seniors, along with their professors. The author then

contrasted and examined the rule-based machine

translation, artificial intelligence translation, and

human-translated materials after using the DeepL

program and hiring a qualified English translator to

translate the textbook simultaneously.

2.1 Textbooks Chosen for the Analysis

Thermodynamics: An Engineering Approach (7th

edition), written by Yunus A. Cengel and Michael A.

Boles, is the first textbook. Important topics like

energy, the second law of thermodynamics, entropy,

the gas power cycle, chemistry, and phase

equilibrium are covered in this book, which presents

the fundamental ideas and laws of engineering

thermodynamics and their applications. The book

also includes numerous instances that are explored

together with technical applications. Undergraduate

and graduate students studying energy and power

engineering, architecture, machinery, and other

relevant fields at universities can use this book as a

textbook.

Fluid Mechanics Fundamentals and Applications

(2nd edition), written by Yunus A. Cengel and John

A Comparative Study on the Performance of ChatGPT-4 and Claude 3 in Translating Engineering Specialized Course Textbooks

455

M. Cimbala, is the second book. This book provides

many examples of engineering while introducing the

fundamental ideas and formulas of fluid mechanics.

It discusses crucial topics such fluid kinematics,

pressure and hydrostatics, Bernoulli equations, and

energy equations. Undergraduate and graduate

students studying physical engineering, chemical

engineering, aerospace, and other relevant fields can

use this book as a university textbook.

These two volumes, which are standard textbooks

for Chinese-foreign cooperative education in

engineering, cover a lot of information in several

engineering disciplines and are appropriate for use in

this study.

2.2 Procedures

The author intends to start by providing ChatGPT-4

with the same CO-STAR command framework as

Claude 3 (with “C” for Context, “O” for Objective,

“S” for Style, “T” for Tone, “A” for Audience, and

“R” for Response). These two AI systems ought to be

able to more precisely and expertly translate technical

materials from English into Chinese with the help of

this framework. Two factors led to the selection of the

CO-STAR education framework: 1. It is thorough,

producing translated text that is accurate and

readable. 2. It is focused, making translations in

specialized fields easier.

Second, upon the completion of the translation,

the writers will ask professionals in the field who

have undergone these specialized courses to assess

the outcomes. A framework for assessing the quality

of translations produced by humans, machines, and

artificial intelligence is called MQM (Multi-

dimensional Quality Metrics) (Zhang & Li, 2009).

Although there are seven assessment indicators in the

MQM system, not all of them can be used when

translating scientific materials in engineering

textbooks. Therefore, three variables from the MQM

framework — accuracy, fluency, and audience

adaptability—were selected for this study in order to

reduce redundancy and increase evaluation

efficiency. In addition, the inclusion of the indication

and logic makes it easier to assess how well the

scientific text's reasoning and argumentation process

is supported in the translation. Table 1 displays the

four indicators' definitions and evaluation goals. As

indicated in Table 2, a Likert scale was used to assess

each indicator from 1 to 5. In this study, a five-level

scale questionnaire with four assessment indicators

was designed using a quantitative-qualitative and

qualitative-mixed research methodology. This allowed

for both a qualitative analysis of the differences

Table 1: Explanation of indicators.

Indicator Definition Objectives

Accuracy The translation faithfully captures the original

text's content, message, and intent.

Evaluate whether the translation faithfully captures the

original text's meaning and make sure that terminology

is not mistranslated when translatin

scientific texts.

Fluency The translation is adequately fluid, exhibiting

natural expression, a range of language and

phrase structures, and cohesion between

sentences.

Evaluate the translation's vocabulary, syntax, and

grammar to see if it is readable, fluid, and natural.

Audience

Appropriateness

The translation is suitable for the intended

audience's reading level, comprehension, and

interests.

Evaluate whether the translation has taken into

consideration the target audience's expected level of

knowledge, which in this case was undergraduate

students.

Logicality The translation's procedures for reasoning and

arguments are sound, understandable, and

consistent.

Evaluate whether the translation adheres to the scientific

text's rigor and steers clear of ambiguous or nonsensical

reasoning.

Table 2: Likert scale for all assessments.

Metric 1 2 3 4 5

Accuracy Very inaccurate Inaccurate Moderately accurate Accurate Very accurate

Fluency Not fluent Not very fluent Moderately fluent Fluent Very fluent

Audience

Appropriateness

Very inappropriate Inappropriate Moderately appropriate Appropriate

Very

appropriate

Logicality Very illogical Illogical Moderately logical Logical Very logical

IESD 2025 - International Conference on Innovative Education and Social Development

456

between the AI translation results and those of the

human and rule-based machine translations, as well as

a quantitative analysis and interpretation of the

research data. In order to provide a higher-quality,

more accurate translation of engineering textbooks,

this study ultimately analyzes the benefits and

drawbacks of ChatGPT-4 and Claude 3.

2.3 Data Collection

In all, 71 questionnaires were gathered for this study,

with 32 male and 39 female participants. The samples

comprised professors and freshman to senior

engineering students at Shanghai Normal University,

including 13 teachers, with almost 60% of the

samples being “fourth-year university students”.

3 RESULTS & DISCUSSION

3.1 Quantitative Research Results

The scale sample in the questionnaire is dependable

and trustworthy since the Cronbach alpha coefficient,

which is larger than 0.9, was 0.972 when the gathered

questionnaire was examined for validity and

reliability in this study. The study data is very

acceptable for information extraction since the KMO

value was 0.858, which is more than 0.8. The research

sample has strong validity and reliability.

Since the discrete degree of evaluation of Claude

3 and ChatGPT-4 is essentially the same, students and

teachers of different engineering majors are in the

same divergence of evaluation of these two AI tools,

and there is never a situation in which the evaluation

of one AI tool is more controversial. According to the

Table 3: A table comparing the advertising language ability of ChatGPT-4o and Kimi in the translation beauty industry.

* p<0.05 ** p<0.01

Passage Metric AI Mean Standard Deviation

Mean

Difference

t p

1 Accuracy ChatGPT-4 3.92 0.65 -0.07 -0.672 0.504

Claude 3 3.99 0.80

Fluency ChatGPT-4 3.90 0.74 -0.21 -2.200 0.031*

Claude 3 4.11 0.67

Audience

Appropriateness

ChatGPT-4 3.92 0.71 -0.18 -2.333 0.023*

Claude 3 4.10 0.70

Logicality ChatGPT-4 3.97 0.68 -0.04 -0.445 0.658

Claude 3 4.01 0.80

2 Accuracy ChatGPT-4 3.94 0.81 -0.07 -0.698 0.488

Claude 3 4.01 0.80

Fluency ChatGPT-4 3.89 0.73 -0.15 -1.742 0.086

Claude 3 4.04 0.78

Audience

Appropriateness

ChatGPT-4 3.90 0.74 -0.14 -1.598 0.114

Claude 3 4.04 0.76

Logicality ChatGPT-4 3.96 0.73 -0.11 -1.526 0.132

Claude 3 4.07 0.76

3 Accuracy ChatGPT-4 4.00 0.74 0.00 0.000 1.000

Claude 3 4.00 0.72

Fluency ChatGPT-4 3.90 0.78 -0.17 -2.044 0.045*

Claude 3 4.07 0.72

Audience

Appropriateness

ChatGPT-4 3.93 0.74 -0.08 -0.973 0.334

Claude 3 4.01 0.73

Logicality ChatGPT-4 3.96 0.75 -0.14 -1.926 0.058

Claude 3 4.10 0.68

4 Accuracy ChatGPT-4 3.97 0.76 0.00 0.000 1.000

Claude 3 3.97 0.77

Fluency ChatGPT-4 3.99 0.80 0.04 0.445 0.658

Claude 3 3.94 0.75

Audience

Appropriateness

ChatGPT-4 3.87 0.77 -0.21 -2.154 0.035*

Claude 3 4.08 0.71

Logicality ChatGPT-4 4.03 0.74 -0.10 -1.355 0.180

Claude 3 4.13 0.65

A Comparative Study on the Performance of ChatGPT-4 and Claude 3 in Translating Engineering Specialized Course Textbooks

457

results of the questionnaire data survey, Claude 3 has

a slightly higher average value of each index than

ChatGPT-4, which is more advantageous in user

evaluation. In contrast, the assessments of the four

indicators do not change (p>0.05) when ChatGPT-4

translates four sets of distinct engineering textbook

sections, and the translation capability remains

constant. The stability of Claude 3's translation of the

engineering textbook can still be improved, though,

as one of the four groups of distinct passages in the

textbook exhibits a significance at the 0.05 level of

fluency with the other two passages (p is 0.038 vs.

0.022, respectively).

Furthermore, Table 3 displays the survey results

of the comparison analysis of the translations of

Claude 3 and ChatGPT-4. The table shows that the

first translated passage, Claude 3, has better audience

adaptation and fluency than ChatGPT-4; the second

translated passage's results show no difference

(p>0.05); the third translated passage, Claude 3, has

better fluency than ChatGPT-4; and the fourth

translated passage, Claude 3, has better audience

adaptation than ChatGPT-4.

In summary, the research indicates that both

ChatGPT-4 and Claude 3 perform well when

translating engineering textbooks; both exhibit good

accuracy and logic, but ChatGPT-4 performs worse

than Claude 3 in terms of fluency and audience

adaptability. Claude 3's translation stability still

requires work.

3.2 Comparative Result Analysis of

Four Translation Methods

After comparing and evaluating four selections

translated with rule-based machine translation,

human translation, ChatGPT-4 translation and Claude

3, the author came to the following conclusions.

1. Rule-based machine translation is able to

recognize and translate most specialized vocabulary,

but omissions and mistranslations are more

problematic. For instance, “Fluid statics is used to

determine the forces acting on floating or submerged

bodies and the forces developed by devices like

hydraulic presses and car jacks.” Rule-based machine

translation can identify and translate the majority of

specialized vocabulary. The forces generated by

machinery such as vehicle jacks and hydraulic

presses, as well as the forces acting on submerged or

floating things, are determined using fluid statics.

While “floating or submerged bodies” was accurately

translated but just a portion of the text was translated,

leaving out important details. because the sentence's

use of “floating or submerged bodies” is incorrectly

translated, leaving out important details. In addition,

it suffers from certain translation accents, such as

slightly unnatural expressions like “the so-called” and

“become very important”, which make it less fluent

to read.

2. Certain specialist terms, such “isentropic” and

“isentropic efficiencies,” are appropriately translated

using ChatGPT-4. The translation is consistent with

the Chinese expression pattern, yet the sentences are

also fluid, natural, and logical, such as the sentence

“Electrons at outer orbits have larger kinetic

energies.” The phrase “the force relations developed

naturally involve the gravitational acceleration” is

one example of a term that has to be better improved

since it is a little stiff. The phrase “the force relations”

is too simple to translate.

3. Since there is no relative motion between the

fluid and the solid surface, there are no shear forces

acting parallel to the surface. Claude 3's translation

guarantees accuracy, fluency, and logicalness while

also excelling at handling lengthy and challenging

sentences. Shear forces operating parallel to the

surface are absent as there is no relative motion

between the fluid and the solid surface. The sentence's

translation is clear and faithfully captures the original

text's logical flow. However, there are also some

cases where the translation is too colloquial and the

meaning is vague, such as in the sentence “the

variation of pressure is due only to the weight of the

fluid. For example, in the sentence “the variation of

pressure is due only to the weight of the fluid.”, the

translation of “is due to” lacks professionalism.

4. Compared to the other three translation

processes, human translations are the slowest and

least effective, but they are of the greatest quality and

have almost evident errors.

The aforementioned investigation results indicate

that the public is aware of ChatGPT-4 and Claude 3's

capacity to translate engineering texts. Although both

ChatGPT-4 and Claude 3 have presentation issues,

Claude 3 is superior in this area in terms of

translation. In the meanwhile, ChatGPT-4 and Claude

3 continue to struggle with difficult, informal, and

ambiguous language, and their overall translation

quality falls short of that of human translators. Thus,

according to the author, when translating engineering

textbooks, translators can use the AI tool to finish the

first translation and then combine it with the benefits

of human translation quality to add embellishments or

rewrites to further improve the translation's wording

and elaboration.

IESD 2025 - International Conference on Innovative Education and Social Development

458

4 CONCLUSION

With advancements in science and technology,

translation technology also brings about innovation as

times change. According to the author, textbooks

from all disciplines should aim to use the most

advanced translation technology, such as AI

translation, when they are translated, as the academic

community is now placing a strong emphasis on the

sharing of global educational materials across

languages. This study employed a mixed research

approach to identify the advantages and

disadvantages of ChatGPT-4 and Claude 3 in the

translation of engineering textbooks. The advantages

of these two AI tools include their ability to

effectively translate scientific texts in engineering

textbooks in a way that is accurate, fluent, acceptable,

and logical, as an alternative to rule-based machine

translation. Furthermore, in this subject, Claude 3 is

better suited for translation. The drawback is that,

although scientific texts are objective, engineering

textbooks must be presented carefully to be

appropriate for this age range in order to serve as

instructional resources for students. The presentation

of these two AI tools is still difficult, informal, and

unclear; translators who are knowledgeable about the

variations in languages, cultures, and modes of

expression among nations must manually alter and

add to them. In order to produce more accurate and

expert translations of top-notch engineering

textbooks, the author suggests that the AI tool should

be able to learn the languages and cultures of other

nations and continually enhance its algorithm.

However, this study has many drawbacks, such as a

too limited selection of engineering textbooks; more

engineering textbooks in other domains may be

included. In future research, more in-depth studies

can be conducted to continue exploring the

complementary aspects of AI translation and human

translation in textbooks for different specialties in

engineering. Through this project, it is envisaged that

translators would become proficient in integrating AI

and manual touch-ups while translating engineering

textbooks, fostering collaboration and innovation in

international education while also fostering cross-

linguistic communication.

REFERENCES

Ahmadova, S. 2024. A comparative quality assessment of

ChatGPT - 4 and human translation of scientific texts

(Master's thesis, Khazar University (Azerbaijan)).

Ahrenberg, L. 2017. Comparing machine translation and

human translation: A case study. In RANLP 2017: The

First Workshop on Human - Informed Translation and

Interpreting Technology (HiT - IT) (pp. 21 - 28). Asso-

ciation for Computational Linguistics.

Al - Romany, T. A. H., Kadhim, & M. J. 2024. Artificial

Intelligence Impact on Human Translation: Legal Texts

as a Case Study. International Journal of Linguistics,

Literature and Translation 7.5:89 - 95.

Balas, M., Kaplan, A. J., Esmail, K., Saleh, S., Sharma, R.

A., Yan, P., & Arjmand, P. 2024. Translating ophthal-

mic medical jargon with artificial intelligence: a com-

parative comprehension study. Canadian Journal of

Ophthalmology.

Dictionary, O. E. 1989. Oxford English dictionary.

Simpson, Ja & Weiner, Esc, 3.

Duderstadt, J. J. 2007. Engineering for a changing road, a

roadmap to the future of engineering practice, research,

and education.

Ferrag, F., & Bentounsi, I. 2024. A. The Use of Artificial

Intelligence in Academic Translation Tasks Case Study

of Chat GPT, Claude and Gemini.

Haq, M., Mushhood Ur Rehman, M., Derhab, M., Saeed,

R., & Kalia, J. 2024. Bridging Language Gaps in

Neurology Patient Education Through Large Language

Models: a Comparative Analysis of ChatGPT, Gemini,

and Claude. medRxiv 2024 - 09.

Li, B., Yang, P., Sun, Y., Hu, Z., & Yi, M. 2024. Advances

and challenges in artificial intelligence text generation.

Frontiers of Information Technology & Electronic

Engineering 25(1): 64 - 83.

Li, Q. 2024. Bridging Languages: The Potential and

Limitations of AI in Literary Translation: A Case Study

of the English Translation of A Pair of Peacocks

Southeast Fly. Advances in Humanities Research 8.1: 1

- 7.

Li, S. Q., 2024. The Role and Challenges of Generative

Artificial Intelligence (AIGC) for Science and

Technology Communication. China Science and

Technology Industry (05):50-52.

Lommel, A., Gladkoff, S., Melby, A., Wright, S. E.,

Strandvik, I., Gasova, K. & Nenadic, G. 2024. The

multi-range theory of translation quality measurement:

Mqm scoring models and statistical quality control.

arXiv preprint arXiv 2405.16969.

Mohsen, M. 2024. Artificial Intelligence in Academic

Translation: A Comparative Study of Large Language

Models and Google Translate. PSYCHOLINGUISTICS

35(2): 134 - 156.

Quigley, C., Oliviera, A. W., Curry, A., & Buck, G. 2011.

Issues and techniques in translating scientific terms

from English to Khmer for a university-level text in

Cambodia. Language, Culture and Curriculum 24(2):

159 - 177.

Shokoohifar, M. 2024. Artificial Intelligence-Assisted

Translation: A Study of Cognitive Load and Time

Through EEG (Doctoral dissertation, Allameh

Tabataba’i University).

A Comparative Study on the Performance of ChatGPT-4 and Claude 3 in Translating Engineering Specialized Course Textbooks

459

Wang, L. 2023. The Impacts and Challenges of Artificial

Intelligence Translation Tool on Translation

Professionals. SHS Web of Conferences 163.

Wang, Y. 2023. Artificial Intelligence Technologies in

College English Translation Teaching. Journal of

Psycholinguistic Research.

Yan, J., Yan, P., Chen, Y., Li, J., Zhu, X., & Zhang, Y.

2024. Benchmarking GPT - 4 against Human

Translators: A Comprehensive Evaluation Across

Languages, Domains, and Expertise Levels. arXiv

preprint arXiv:2411.13775.

Yan Y. Q., & Zhao Y., 2023. Research on the Application

of AIGC in Traditional Artificial Intelligence Systems.

Guangdong Communication Technology (12):14 -

17+68.

Zhang, H. B., & Li, Y. S., 2009. A study on the

development status of educational resource sharing

environment and sharing mechanism. China Electronic

Education (11):68-73.

Zhao, L. H., 2022. The Relationship between Machine

Translation and Human Translation under the Influence

of Artificial Intelligence Machine Translation. Mobile

Information Systems (1): 9121636.

IESD 2025 - International Conference on Innovative Education and Social Development

460