Ergonomic LLM or LLM for Ergonomics? Prompt engineering insights for an interventional case study

Ergonomic LLM or LLM for Ergonomics? Prompt engineering insights for an interventional case study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Case Report Ergonomic LLM or LLM for Ergonomics? Prompt engineering insights for an interventional case study Alireza Mortezapour This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4304633/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background With the arrival of new technologies associated with the fourth industrial revolution (IR 4.0), the way humans interact with technology has undergone fundamental changes. In the last two years, a new generation of technology, large language models (LLMs), and with the leader position of ChatGPT from OpenAI has gained a lot of attention. Objective In the current study, prompt engineering technique usefulness regards to Human-Chat GPT interaction is discussed. Methods Three types of interaction with Chat GPT including zero-shot, little-shot and fine-tune prompting are considered. Results Our small case study implies that Human-Chat GPT interaction can be influenced under the proper usage of prompt engineering. Conclusion I implied that, prompt engineering can be included in future Human Factors and Ergonomics courses in academia (for ergonomists) or in industries (for employees or employers). Theoretical Computer Science Psychology Occupational Medicine large language model human-technology interaction industry 4.0 industry 5.0 human-computer interaction 1. Introduction 1.1. Trends in human-technology interaction The rapid advancement of technology over the past decade has led to a significant shift in how humans interact with it ( 1 ). Following the onset of the fourth industrial revolution, the technologies associated with this revolution have progressively transitioned into marketable products ( 2 ). Particularly in recent years, the emergence of various social robots, self-driving cars, and similar innovations has ushered in a new era of human-technology interaction. These advancements are primarily rooted in artificial intelligence, giving rise to what is now termed "human-AI interaction" ( 3 ). It could be noted that users were becoming accustomed to modern high-tech products. However, over the past two years, OpenAI has introduced a new generation of technologies that are part of the large language models (LLMs) family ( 4 ). These products have offered people a fresh wave of innovative solutions. 1.2. LLMs as new technological products In contrast to the previous years, during which users were limited to communicating with computers and machines solely through programming languages like Python, recent advancements have introduced a new era where computers, in a general sense, can be interacted with by humans without requiring knowledge of programming languages. Now, individuals can interact with these machines using conversational language, facilitating seamless communication ( 5 ). These machines operate on the concept of large language models, enabling them to understand human conversations, primarily text-based, but also with access to other modalities. They possess the capability to engage in human-like conversations, as well as to predict future outcomes based on the context of the conversations provided ( 6 ). Various language models have been introduced to date, each with its own set of strengths and weaknesses. However, the key factor lies in the vast amount of information these machines have been trained on, enabling them to effectively address user needs. For instance, the creator of one of the most renowned models, ChatGPT, asserts that its 3.5 version, currently accessible to users at no cost, has been trained on a staggering 175 billion parameters. Apart from this widely utilized LLM, numerous others have been developed by tech giants such as Google. The important thing that is used in this study is the ability to communicate these language models with not expert computer users through simple and accessible interfaces. 1.3. LLMs: Challenges in the human-technology interaction (continuing the way of AI products) Like all artificial intelligence-based products, past studies have also examined challenges associated with LLMs, including: ethical challenges ( 7 ), information security ( 8 ), over-trust of users ( 9 ), etc. 1.4. LLMs: New opportunities for augmenting human intelligence Among all the challenges of utilizing these language models, it's noteworthy that today their use extends beyond computer science departments and academic settings. These products are now utilized by numerous users on a daily basis, reflecting their widespread adoption across various domains. In some recent studies, the authors discussed the role of LLMs in augmenting human ( 10 ) and also they provided some insights regards to LLMs role in increasing the productivity ( 11 ). 2. New added value by presenting a case study: guiding an office worker to have an ergonomic workstation Among the array of LLMs available, ChatGPT 3.5 is specifically employed in this context. The rationale behind this selection is to showcase the capabilities of an intelligent yet freely accessible LLM within the realm of human factors and ergonomics practice. It's worth noting that the author acknowledges the existence of other large language models, as well as different versions of ChatGPT, which may elicit responses differing from those utilized in this particular example. In this instance, a prompt engineering technique was employed to leverage ChatGPT 3.5. ChatGPT prompts serve as instructions or queries inputted into its interface to elicit desired responses. These prompts typically encompass keywords and phrases intended to stimulate a reply. Users can input questions or instructions to ChatGPT, which then generates responses in a conversational manner, facilitating ongoing dialogue that can extend for hours or even days. In the current case study, the primary objective is to demonstrate the initial capability of a freely available LLM in assisting an office worker in setting up an ergonomic workstation. Subsequently, the secondary aim is to illustrate the distinctions between zero-shot, few-shot, and fine-tuned prompts in accomplishing the initial goal. In the field of human-AI interaction literature, or more specifically in computer science literature, it's recognized that providing a well-crafted prompt can guide a large language model (LLM) to deliver more precise answers. Researchers have emphasized the significance of specialized prompts tailored to different LLMs for enhancing their performance and accuracy in generating responses. 2.1. Persona Here, the focus is on assisting an office worker in finding effective solutions related to office ergonomics tips. To begin, let's introduce the persona. She is a 55-year-old woman who has been diagnosed with low back pain. Her physician has informed her that the issue stems from her posture during work hours. Determined to address this problem independently, she has heard about ChatGPT and believes she can utilize it to find a solution for her backache. 2.2. Zero-shot prompting Here, in the context of zero-shot usage, the individual utilizes her ChatGPT account with minimal effort, without employing any targeted prompts. Instead, she provides a general description of her concern, focusing on the physician's recommendation. This includes: Hello. My physician told me that I have a problem in my back and it is related to my office work. please help me to improve my office ergonomic to help my back. 2.3. Little-shot prompting Here, in the little-shot approach, additional details about her problem are provided to assist ChatGPT in producing more specific and useful information. Specifically, she provides her weight, height, and age to receive personalized recommendations regarding her user profile. “C onsider this user profile: {Age: 55} -years old {gender: woman}. Weight and height {weight: 76} kg and {height: 160} cm. In addition to previously provided information, what is the more personalized recommendations regards my situation? 2.4. Fine tune prompting Here, fine tune prompt means, she learned to provide as much as useful information regards to her situation to receive more precise recommendations. In this case, in addition to all previous information (step 1 and step 2) she provided information also related to her average physical activity per day (total walking steps in a week) and her sleep pattern (which was enough or not). I have been diagnosed with backache and I provide my user profile. Now, consider additional information related to my health profile: I usually have {average steps/day: 5000} steps walk each day and my sleep pattern is {3 out of 5}. In addition to previous information, provide more comprehensive recommendation to me. The author knows that more information can be related to musculoskeletal disorders and therefore they can also include. But due to the success of past studies in quantifying some of these parameters ( 12 ), we only converted selected items (physical activity, sleep pattern) into appropriate prompts. 2.5. Qualitative comparison of the results Upon comparing the results obtained in this study, the author, an expert in the field of ergonomics with a track record of interventions across various industries and specialized training in office ergonomics, observed that providing more detailed information through true using of prompt engineering insights, led to more accurate and applicable recommendations. Additionally, the author noted the potential for the "Eliza effect" ( 13 )– whereby users perceive a computer as more intelligent when provided with personalized information – and hypothesized that users may apply recommendations more precisely in such cases. However, testing this hypothesis falls beyond the scope of the current study. 2.6. Limitations Despite the intriguing findings of this study, it's important to acknowledge its limitations. Firstly, the method was tested on only one individual with a specific profile. While the results were successful, it's advisable to conduct further studies involving individuals with diverse profiles before generalizing these findings. Additionally, since only one specific large language model was investigated, similar studies should be conducted using other language models, including more specialized ones. Another limitation pertains to the qualitative analysis of the extracted recommendations. Future studies could address this by involving multiple experts or integrating machine learning techniques for quantitative analysis. These limitations highlight the need for future research to broaden the scope and depth of investigations in this area. 3. Concerns in this case study Just like in any other use of technology, this individual also learned to observe important points throughout her interaction with the large language model. These points are not only relevant to her specific situation but are also valuable for other readers to consider when utilizing similar technology for various purposes. 3.1. Privacy In this study, she learned the importance of not disclosing sensitive clinical information about herself to ChatGPT. This precaution stemmed from the understanding that ChatGPT is not bound by restrictions when it comes to training itself with such information or potentially sharing it with others in similar cases. 3.2. Over-trust The main author, an ergonomic expert, informed her that despite the appealing results of the study, she should not overly rely on them and instead advised her to consult with a clinical specialist for personalized guidance. Furthermore, she was cautioned against searching for medical information related to her condition out of curiosity or using the study's results as a substitute for professional medical advice. 4. Conclusion This study has highlighted the successful outcomes of utilizing free large language models in ergonomics interventions. It demonstrates how individuals can attain significant results with minimal training on crafting relevant prompts. As a recommendation, it is suggested that scientific ergonomics associations consider addressing whether prompt writing should be incorporated into ergonomics training courses in the near future. This could potentially enhance users' abilities to effectively utilize large language models for various purposes. Declarations Conflict of interest: Nothing for declaration. Funding body: This study does not use any financial supports. References Gualtieri L, Fraboni F, De Marchi M, Rauch E. Development and evaluation of design guidelines for cognitive ergonomics in human-robot collaborative assembly systems. Applied Ergonomics. 2022;104:103807. Clark JR, Large DR, Shaw E, Nichele E, Galvez Trigo MJ, Fischer JE, et al. Identifying interaction types and functionality for automated vehicle virtual assistants: An exploratory study using speech acts cluster analysis. Applied Ergonomics. 2024;114:104152. Salmon PM, McLean S, Carden T, King BJ, Thompson J, Baber C, et al. When tomorrow comes: A prospective risk assessment of a future artificial general intelligence-based uncrewed combat aerial vehicle system. Applied Ergonomics. 2024;117:104245. Camilleri MA. Factors affecting performance expectancy and intentions to use ChatGPT: Using SmartPLS to advance an information technology acceptance framework. Technological Forecasting and Social Change. 2024;201:123247. Zou W, Li J, Yang Y, Tang L. Exploring the Early Adoption of Open AI among Laypeople and Technical Professionals: An Analysis of Twitter Conversations on #ChatGPT and #GPT3. International Journal of Human–Computer Interaction.1-12. Pataranutaporn P, Liu R, Finn E, Maes P. Influencing human–AI interaction by priming beliefs about AI can increase perceived trustworthiness, empathy and effectiveness. Nature Machine Intelligence. 2023;5(10):1076-86. Park J-Y. Could ChatGPT help you to write your next scientific paper?: concerns on research ethics related to usage of artificial intelligence tools. Journal of the Korean Association of Oral and Maxillofacial Surgeons. 2023;49(3):105. Wu X, Duan R, Ni J. Unveiling security, privacy, and ethical concerns of chatgpt. Journal of Information and Intelligence. 2023. Zhang B, Soh H. Large language models as zero-shot human models for human-robot interaction. arXiv preprint arXiv:230303548. 2023. Moore S, Tong R, Singh A, Liu Z, Hu X, Lu Y, et al., editors. Empowering education with llms-the next-gen interface and content generation. International Conference on Artificial Intelligence in Education; 2023: Springer. Wen H, Li Y, Liu G, Zhao S, Yu T, Li TJ-J, et al. Empowering llm to use smartphone for intelligent task automation. arXiv preprint arXiv:230815272. 2023. Kim Y, Xu X, McDuff D, Breazeal C, Park HW. Health-llm: Large language models for health prediction via wearable sensor data. arXiv preprint arXiv:240106866. 2024. Eisenmann C, Mlynář J, Turowetz J, Rawls AW. “Machine Down”: making sense of human–computer interaction—Garfinkel’s research on ELIZA and LYRIC from 1967 to 1969 and its contemporary relevance. AI & SOCIETY. 2023. Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4304633","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Case Report","associatedPublications":[],"authors":[{"id":293975543,"identity":"5cde9e4c-9e3a-4696-b1b7-cd8a205aad9f","order_by":0,"name":"Alireza Mortezapour","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABAUlEQVRIiWNgGAWjYNCDj39s5ED0gQeEVB6A0owzG9KMwQIJxGph5m04nNgAYuHTwj/t8LHPHyrq5M3Ze8w+8O5IS58fdvgh0BY7Od0G7Fokbqclzzhwhs1wZ88Z4xmSZ2xyN95OMwBqSTY2O4BdC8PtHGOGg208jBtuABkGbGm5G2cngLQcSNyGQ4s8WMs/CXuwlgS2w+mGs9M/4NViANbSYJAI1nKw7XCCvHQOflsMgX5hOHMsIXlnz7FixoYzaYYbpHMKDiQY4PaL3O3kwwwVNXW229mbNzP/qbCRl5+dvvnDhwo7OZzeh7sQzjiAwiVGi3wDEapHwSgYBaNgRAEACMJnMyc6XgcAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0001-6356-2244","institution":"University of Salerno","correspondingAuthor":true,"prefix":"","firstName":"Alireza","middleName":"","lastName":"Mortezapour","suffix":""}],"badges":[],"createdAt":"2024-04-22 09:13:08","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":true,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-4304633/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4304633/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":55155178,"identity":"10e0483e-3d10-4f19-889e-c6e77b44c7b1","added_by":"auto","created_at":"2024-04-23 11:26:33","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":283419,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4304633/v1/5d693e4e-4f92-4e91-94b7-c69523f14193.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eErgonomic LLM or LLM for Ergonomics? Prompt engineering insights for an interventional case study\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"1. Introduction","content":"\u003cdiv id=\"Sec2\" class=\"Section2\"\u003e\n \u003ch2\u003e1.1. Trends in human-technology interaction\u003c/h2\u003e\n \u003cp\u003eThe rapid advancement of technology over the past decade has led to a significant shift in how humans interact with it (\u003cspan class=\"CitationRef\"\u003e1\u003c/span\u003e). Following the onset of the fourth industrial revolution, the technologies associated with this revolution have progressively transitioned into marketable products (\u003cspan class=\"CitationRef\"\u003e2\u003c/span\u003e). Particularly in recent years, the emergence of various social robots, self-driving cars, and similar innovations has ushered in a new era of human-technology interaction. These advancements are primarily rooted in artificial intelligence, giving rise to what is now termed \u0026quot;human-AI interaction\u0026quot; (\u003cspan class=\"CitationRef\"\u003e3\u003c/span\u003e). It could be noted that users were becoming accustomed to modern high-tech products. However, over the past two years, OpenAI has introduced a new generation of technologies that are part of the large language models (LLMs) family (\u003cspan class=\"CitationRef\"\u003e4\u003c/span\u003e). These products have offered people a fresh wave of innovative solutions.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003e1.2. LLMs as new technological products\u003c/h2\u003e\n \u003cp\u003eIn contrast to the previous years, during which users were limited to communicating with computers and machines solely through programming languages like Python, recent advancements have introduced a new era where computers, in a general sense, can be interacted with by humans without requiring knowledge of programming languages. Now, individuals can interact with these machines using conversational language, facilitating seamless communication (\u003cspan class=\"CitationRef\"\u003e5\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003eThese machines operate on the concept of large language models, enabling them to understand human conversations, primarily text-based, but also with access to other modalities. They possess the capability to engage in human-like conversations, as well as to predict future outcomes based on the context of the conversations provided (\u003cspan class=\"CitationRef\"\u003e6\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003eVarious language models have been introduced to date, each with its own set of strengths and weaknesses. However, the key factor lies in the vast amount of information these machines have been trained on, enabling them to effectively address user needs. For instance, the creator of one of the most renowned models, ChatGPT, asserts that its 3.5 version, currently accessible to users at no cost, has been trained on a staggering 175\u0026nbsp;billion parameters. Apart from this widely utilized LLM, numerous others have been developed by tech giants such as Google.\u003c/p\u003e\n \u003cp\u003eThe important thing that is used in this study is the ability to communicate these language models with not expert computer users through simple and accessible interfaces.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\n \u003ch2\u003e1.3. LLMs: Challenges in the human-technology interaction (continuing the way of AI products)\u003c/h2\u003e\n \u003cp\u003eLike all artificial intelligence-based products, past studies have also examined challenges associated with LLMs, including: ethical challenges (\u003cspan class=\"CitationRef\"\u003e7\u003c/span\u003e), information security (\u003cspan class=\"CitationRef\"\u003e8\u003c/span\u003e), over-trust of users (\u003cspan class=\"CitationRef\"\u003e9\u003c/span\u003e), etc.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n \u003ch2\u003e1.4. LLMs: New opportunities for augmenting human intelligence\u003c/h2\u003e\n \u003cp\u003eAmong all the challenges of utilizing these language models, it\u0026apos;s noteworthy that today their use extends beyond computer science departments and academic settings. These products are now utilized by numerous users on a daily basis, reflecting their widespread adoption across various domains. In some recent studies, the authors discussed the role of LLMs in augmenting human (\u003cspan class=\"CitationRef\"\u003e10\u003c/span\u003e) and also they provided some insights regards to LLMs role in increasing the productivity (\u003cspan class=\"CitationRef\"\u003e11\u003c/span\u003e).\u003c/p\u003e\n\u003c/div\u003e"},{"header":"2.\tNew added value by presenting a case study: guiding an office worker to have an ergonomic workstation","content":"\u003cp\u003eAmong the array of LLMs available, ChatGPT 3.5 is specifically employed in this context. The rationale behind this selection is to showcase the capabilities of an intelligent yet freely accessible LLM within the realm of human factors and ergonomics practice. It's worth noting that the author acknowledges the existence of other large language models, as well as different versions of ChatGPT, which may elicit responses differing from those utilized in this particular example.\u003c/p\u003e\n\u003cp\u003eIn this instance, a prompt engineering technique was employed to leverage ChatGPT 3.5. ChatGPT prompts serve as instructions or queries inputted into its interface to elicit desired responses. These prompts typically encompass keywords and phrases intended to stimulate a reply. Users can input questions or instructions to ChatGPT, which then generates responses in a conversational manner, facilitating ongoing dialogue that can extend for hours or even days.\u003c/p\u003e\n\u003cp\u003eIn the current case study, the primary objective is to demonstrate the initial capability of a freely available LLM in assisting an office worker in setting up an ergonomic workstation. Subsequently, the secondary aim is to illustrate the distinctions between zero-shot, few-shot, and fine-tuned prompts in accomplishing the initial goal.\u003c/p\u003e\n\u003cp\u003eIn the field of human-AI interaction literature, or more specifically in computer science literature, it's recognized that providing a well-crafted prompt can guide a large language model (LLM) to deliver more precise answers. Researchers have emphasized the significance of specialized prompts tailored to different LLMs for enhancing their performance and accuracy in generating responses.\u003c/p\u003e\n\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n\u003ch2\u003e2.1. Persona\u003c/h2\u003e\n\u003cp\u003eHere, the focus is on assisting an office worker in finding effective solutions related to office ergonomics tips. To begin, let's introduce the persona. She is a 55-year-old woman who has been diagnosed with low back pain. Her physician has informed her that the issue stems from her posture during work hours. Determined to address this problem independently, she has heard about ChatGPT and believes she can utilize it to find a solution for her backache.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\n\u003ch2\u003e2.2. Zero-shot prompting\u003c/h2\u003e\n\u003cp\u003eHere, in the context of zero-shot usage, the individual utilizes her ChatGPT account with minimal effort, without employing any targeted prompts. Instead, she provides a general description of her concern, focusing on the physician's recommendation. This includes:\u003c/p\u003e\n\u003cdiv class=\"BlockQuote\"\u003e\n\u003cp\u003e\u003cem\u003eHello. My physician told me that I have a problem in my back and it is related to my office work. please help me to improve my office ergonomic to help my back.\u003c/em\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n\u003ch2\u003e2.3. Little-shot prompting\u003c/h2\u003e\n\u003cp\u003eHere, in the little-shot approach, additional details about her problem are provided to assist ChatGPT in producing more specific and useful information. Specifically, she provides her weight, height, and age to receive personalized recommendations regarding her user profile.\u003c/p\u003e\n\u003cp\u003e\u0026ldquo;C\u003cem\u003eonsider this user profile: {Age: 55} -years old {gender: woman}. Weight and height {weight: 76} kg and {height: 160} cm. In addition to previously provided information, what is the more personalized recommendations regards my situation?\u003c/em\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n\u003ch2\u003e2.4. Fine tune prompting\u003c/h2\u003e\n\u003cp\u003eHere, fine tune prompt means, she learned to provide as much as useful information regards to her situation to receive more precise recommendations. In this case, in addition to all previous information (step 1 and step 2) she provided information also related to her average physical activity per day (total walking steps in a week) and her sleep pattern (which was enough or not).\u003c/p\u003e\n\u003cdiv class=\"BlockQuote\"\u003e\n\u003cp\u003eI have been diagnosed with backache and I provide my user profile. Now, consider additional information related to my health profile: I usually have {average steps/day: 5000} steps walk each day and my sleep pattern is {3 out of 5}. In addition to previous information, provide more comprehensive recommendation to me.\u003c/p\u003e\n\u003c/div\u003e\n\u003cp\u003eThe author knows that more information can be related to musculoskeletal disorders and therefore they can also include. But due to the success of past studies in quantifying some of these parameters (\u003cspan class=\"CitationRef\"\u003e12\u003c/span\u003e), we only converted selected items (physical activity, sleep pattern) into appropriate prompts.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\n\u003ch2\u003e2.5. Qualitative comparison of the results\u003c/h2\u003e\n\u003cp\u003eUpon comparing the results obtained in this study, the author, an expert in the field of ergonomics with a track record of interventions across various industries and specialized training in office ergonomics, observed that providing more detailed information through true using of prompt engineering insights, led to more accurate and applicable recommendations. Additionally, the author noted the potential for the \"Eliza effect\" (\u003cspan class=\"CitationRef\"\u003e13\u003c/span\u003e)\u0026ndash; whereby users perceive a computer as more intelligent when provided with personalized information \u0026ndash; and hypothesized that users may apply recommendations more precisely in such cases. However, testing this hypothesis falls beyond the scope of the current study.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n\u003ch2\u003e2.6. Limitations\u003c/h2\u003e\n\u003cp\u003eDespite the intriguing findings of this study, it's important to acknowledge its limitations. Firstly, the method was tested on only one individual with a specific profile. While the results were successful, it's advisable to conduct further studies involving individuals with diverse profiles before generalizing these findings. Additionally, since only one specific large language model was investigated, similar studies should be conducted using other language models, including more specialized ones.\u003c/p\u003e\n\u003cp\u003eAnother limitation pertains to the qualitative analysis of the extracted recommendations. Future studies could address this by involving multiple experts or integrating machine learning techniques for quantitative analysis.\u003c/p\u003e\n\u003cp\u003eThese limitations highlight the need for future research to broaden the scope and depth of investigations in this area.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"3. Concerns in this case study","content":"\u003cdiv class=\"BlockQuote\"\u003e\n\u003cp\u003eJust like in any other use of technology, this individual also learned to observe important points throughout her interaction with the large language model. These points are not only relevant to her specific situation but are also valuable for other readers to consider when utilizing similar technology for various purposes.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n\u003ch2\u003e3.1. Privacy\u003c/h2\u003e\n\u003cp\u003eIn this study, she learned the importance of not disclosing sensitive clinical information about herself to ChatGPT. This precaution stemmed from the understanding that ChatGPT is not bound by restrictions when it comes to training itself with such information or potentially sharing it with others in similar cases.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n\u003ch2\u003e3.2. Over-trust\u003c/h2\u003e\n\u003cp\u003eThe main author, an ergonomic expert, informed her that despite the appealing results of the study, she should not overly rely on them and instead advised her to consult with a clinical specialist for personalized guidance. Furthermore, she was cautioned against searching for medical information related to her condition out of curiosity or using the study's results as a substitute for professional medical advice.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"4. Conclusion","content":"\u003cp\u003eThis study has highlighted the successful outcomes of utilizing free large language models in ergonomics interventions. It demonstrates how individuals can attain significant results with minimal training on crafting relevant prompts. As a recommendation, it is suggested that scientific ergonomics associations consider addressing whether prompt writing should be incorporated into ergonomics training courses in the near future. This could potentially enhance users' abilities to effectively utilize large language models for various purposes.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eConflict of interest:\u003c/strong\u003e Nothing for declaration.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding body:\u0026nbsp;\u003c/strong\u003eThis study does not use any financial supports.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eGualtieri L, Fraboni F, De Marchi M, Rauch E. Development and evaluation of design guidelines for cognitive ergonomics in human-robot collaborative assembly systems. Applied Ergonomics. 2022;104:103807.\u003c/li\u003e\n\u003cli\u003eClark JR, Large DR, Shaw E, Nichele E, Galvez Trigo MJ, Fischer JE, et al. Identifying interaction types and functionality for automated vehicle virtual assistants: An exploratory study using speech acts cluster analysis. Applied Ergonomics. 2024;114:104152.\u003c/li\u003e\n\u003cli\u003eSalmon PM, McLean S, Carden T, King BJ, Thompson J, Baber C, et al. When tomorrow comes: A prospective risk assessment of a future artificial general intelligence-based uncrewed combat aerial vehicle system. Applied Ergonomics. 2024;117:104245.\u003c/li\u003e\n\u003cli\u003eCamilleri MA. Factors affecting performance expectancy and intentions to use ChatGPT: Using SmartPLS to advance an information technology acceptance framework. Technological Forecasting and Social Change. 2024;201:123247.\u003c/li\u003e\n\u003cli\u003eZou W, Li J, Yang Y, Tang L. Exploring the Early Adoption of Open AI among Laypeople and Technical Professionals: An Analysis of Twitter Conversations on #ChatGPT and #GPT3. International Journal of Human\u0026ndash;Computer Interaction.1-12.\u003c/li\u003e\n\u003cli\u003ePataranutaporn P, Liu R, Finn E, Maes P. Influencing human\u0026ndash;AI interaction by priming beliefs about AI can increase perceived trustworthiness, empathy and effectiveness. Nature Machine Intelligence. 2023;5(10):1076-86.\u003c/li\u003e\n\u003cli\u003ePark J-Y. Could ChatGPT help you to write your next scientific paper?: concerns on research ethics related to usage of artificial intelligence tools. Journal of the Korean Association of Oral and Maxillofacial Surgeons. 2023;49(3):105.\u003c/li\u003e\n\u003cli\u003eWu X, Duan R, Ni J. Unveiling security, privacy, and ethical concerns of chatgpt. Journal of Information and Intelligence. 2023.\u003c/li\u003e\n\u003cli\u003eZhang B, Soh H. Large language models as zero-shot human models for human-robot interaction. arXiv preprint arXiv:230303548. 2023.\u003c/li\u003e\n\u003cli\u003eMoore S, Tong R, Singh A, Liu Z, Hu X, Lu Y, et al., editors. Empowering education with llms-the next-gen interface and content generation. International Conference on Artificial Intelligence in Education; 2023: Springer.\u003c/li\u003e\n\u003cli\u003eWen H, Li Y, Liu G, Zhao S, Yu T, Li TJ-J, et al. Empowering llm to use smartphone for intelligent task automation. arXiv preprint arXiv:230815272. 2023.\u003c/li\u003e\n\u003cli\u003eKim Y, Xu X, McDuff D, Breazeal C, Park HW. Health-llm: Large language models for health prediction via wearable sensor data. arXiv preprint arXiv:240106866. 2024.\u003c/li\u003e\n\u003cli\u003eEisenmann C, Mlyn\u0026aacute;ř J, Turowetz J, Rawls AW. \u0026ldquo;Machine Down\u0026rdquo;: making sense of human\u0026ndash;computer interaction\u0026mdash;Garfinkel\u0026rsquo;s research on ELIZA and LYRIC from 1967 to 1969 and its contemporary relevance. AI \u0026amp; SOCIETY. 2023.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"University of Salerno","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"large language model, human-technology interaction, industry 4.0, industry 5.0, human-computer interaction","lastPublishedDoi":"10.21203/rs.3.rs-4304633/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4304633/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eWith the arrival of new technologies associated with the fourth industrial revolution (IR 4.0), the way humans interact with technology has undergone fundamental changes. In the last two years, a new generation of technology, large language models (LLMs), and with the leader position of ChatGPT from OpenAI has gained a lot of attention.\u003c/p\u003e\u003ch2\u003eObjective\u003c/h2\u003e \u003cp\u003eIn the current study, prompt engineering technique usefulness regards to Human-Chat GPT interaction is discussed.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThree types of interaction with Chat GPT including zero-shot, little-shot and fine-tune prompting are considered.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eOur small case study implies that Human-Chat GPT interaction can be influenced under the proper usage of prompt engineering.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eI implied that, prompt engineering can be included in future Human Factors and Ergonomics courses in academia (for ergonomists) or in industries (for employees or employers).\u003c/p\u003e","manuscriptTitle":"Ergonomic LLM or LLM for Ergonomics? Prompt engineering insights for an interventional case study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-04-23 11:18:26","doi":"10.21203/rs.3.rs-4304633/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"dfd8a776-520f-4ae9-83fe-089744b09a7f","owner":[],"postedDate":"April 23rd, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":30982951,"name":"Theoretical Computer Science"},{"id":30982952,"name":"Psychology"},{"id":30982953,"name":"Occupational Medicine"}],"tags":[],"updatedAt":"2024-04-23T11:18:26+00:00","versionOfRecord":[],"versionCreatedAt":"2024-04-23 11:18:26","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4304633","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4304633","identity":"rs-4304633","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}