ChatGPT for Research Writing: Limitations and Ethical Risks
Today, some difficult and time-consuming tasks are automated and made simpler thanks to artificial intelligence (AI), which has become an essential component of our lives. Even in the field of scientific research, this is true. AI writing tools like ChatGPT have become an attractive choice for scholars looking to improve their academic and research writing.
The OpenAI tool demonstrated a “high level of concordance and insight in its explanation,” according to a medRxiv preprint that assesses ChatGPT’s performance and lists ChatGPT as an author. This shows the important role that huge language models can play in decision-making, teaching, and research writing.
In this post, we examine how ChatGPT can assist researchers in the academic writing process as the language tool receives increased attention on a global scale. We also examine the drawbacks and moral quandaries associated with using such AI tools to produce research articles in academia.
What is ChatGPT?
ChatGPT (Chat Generative Pre-trained Transformer). also known as ChatGPT, is a Natural Language Processing (NLP) model based on deep learning algorithms (Islam et al. 2023). The Transformer architecture, which was first described in the publication “Attention is All You Need” by Vaswani et al., 2017, is used in this system.
The model is trained for certain tasks like question answering, language translation, and text completion after being trained on a sizable dataset of human-generated text. ChatGPT is effective for a variety of natural language processing jobs because it can comprehend and produce text that sounds like human speech. It is accessible via OpenAI’s API and in a number of open-source frameworks, and it can be tailored to specific needs.
ChatGPT, which debuted on November 30, 2022, is enthralling the world with its remarkable capacity to comprehend crucial user inputs and provide human-like responses in the form of conversational dialog. For instance, have a look at the in-depth response it gave to the question “What is ChatGPT?” above.
In addition to ChatGPT, the San Fransisco, California-based AI company OpenAI is renowned for developing DALL-E and DALL-E 2, AI tools that became well-known for their capacity to convert text into images.
How does ChatGPT work?
Large language models are able to generate full pages of text because they are built to precisely predict the next word in a sentence. These artificial intelligence (AI) models are neural networks that have been trained on vast volumes of text that have been produced by humans. For instance, ChatGPT, which is trained on 570 gigabytes of text and has 175 billion parameters (compared to GPT-2’s 1.5 billion parameters), aids in the performance of jobs for which it was not specifically taught. ChatGPT has a distinct edge over earlier technologies due to its extra Reinforcement Learning with Human Feedback (RLHF) and the fact that it is the largest language model ever trained. Therefore, it makes sense that this is being hailed as a ground-breaking method of content creation.
Why researchers are turning to ChatGPT
In order to facilitate and improve the efficiency of their academic writing, students and scholars are turning more and more to AI technologies. OpenAI’s ChatGPT has become a potent tool for assisting researchers, particularly non-native English speakers, with the academic writing process. ChatGPT produces intelligent-sounding and well-written material that closely resembles that of a real human.
In fact, Abu Dhabi University (ADU) has released ChatGPT guidelines to help students and researchers when utilising the OpenAI-created artificial intelligence (AI) application. The news is a result of the university’s dedication to utilise cutting-edge AI technology to improve research capacities, teaching strategies, and students’ learning experiences.
Here are a few quick examples of how researchers can use ChatGPT.
1. Making an outline for a research paper
Based on user inputs, ChatGPT is an excellent research assistant that may generate an organised outline or a catchy title for your PhD thesis or research paper. Before they begin writing, researchers can review this recommended outline and make any necessary modifications.
2. Abstracting research papers
It might be challenging for researchers to successfully present their work in the form of a comprehensive abstract. In a recent study, abstracts produced by ChatGPT managed to deceive scientists and even pass important plagiarism and AI output detection checks, demonstrating yet another application for the tool.
3. Text in another language being translated by ChatGPT
ChatGPT excels in generating text translations from one language to another and other natural language processing problems. To ensure perfect accuracy, researchers might need to compare its performance to translations created by humans.
4. Rewriting challenging passages of text
Researchers can also use ChatGPT to rewrite translated content and come up with definitions or parallels for common ideas, especially if they are non-native English speakers. Researchers can use ChatGPT to transform existing text into well-written content that keeps the original meaning and is free of language and grammatical problems in addition to helping them communicate complicated ideas in a clear, concise manner.
5. Summarising lengthy papers or articles
The OpenAI chatbot generates concise summaries of lengthy documents and full-text publications to assist busy researchers with their literature search and reading. These brief summaries make it easier for researchers to immediately understand concepts and locate studies that might be pertinent to their work.
6. Recommending modifications to the experimental design
Language models like ChatGPT can offer intriguing variations based on an inputted description, assist in designing experimental design, or offer a fresh take on current experimental techniques.
7. Training for problems involving natural language processing
ChatGPT is a flexible language model that may be adjusted and tailored for certain tasks such as pre-training, text production, dialogue systems, and training models for language understanding.
Limitations of using ChatGPT for research writing
Although ChatGPT is an effective tool for research writing, there are some limitations that should be taken into account.
1. ChatGPT can only produce text based on the patterns it notices in training data; it is unable to come up with novel ideas. As the AI-generated text might not contain references or citations, which are essential for research writing, this increases the risk of plagiarism.
2. This chatbot is a statistical model that creates text based on patterns in the training data. Its performance depends entirely on the prompts that are given to it. This could imply that it occasionally proposes responses that are correct or illogical despite having a plausible appearance because it is unable to comprehend the meaning of the content.
3. This chatbot is a statistical model that creates text based on patterns in the training data. Its performance depends entirely on the prompts that are given to it. This could imply that it occasionally proposes responses that are correct or illogical despite having a plausible appearance because it is unable to comprehend the meaning of the content.
4. The OpenAI chatbot might not always accurately represent the particular study domain, which could result in errors or discrepancies in the research information produced. This is a challenge for researchers who must produce language that calls for expertise in a certain topic.
5. Occasionally, ChatGPT may produce offensive or biased text if the training data exhibits such biases. When using the model for research writing, it is crucial to be aware of this limitation and take the necessary steps to overcome it.
6. Last but not least, ChatGPT has little knowledge of global and international events that took place after 2021. Keep in mind that any content produced by this AI chatbot requires extensive editing and checking to make sure it complies with major submission standards.
Yejin Choi, a computer scientist at the University of Washington and Allen Institute for Artificial Intelligence, claims that “what we have at the moment is essentially a mouth without a brain.” Researchers continue to face significant challenges in figuring out how to teach large language models moral judgment, common sense, or causal reasoning.
Concerns about using ChatGPT for academic writing from an ethical perspective
Academic writing, which is an essential step in the research process, calls for careful thought and attention to detail. The prompt-based content ChatGPT can generate in a matter of seconds can speed up productivity, but it’s crucial to think about the moral ramifications of utilising AI tools to write your research papers for you.
Possibility of writing with biases
Although additional safety measures have been implemented, ChatGPT is trained on existing datasets, and if these contain unfavourable information or biases, they may manifest themselves in the chatbot’s output. This is a challenge for researchers who must be aware of potential biases and double-check the information provided to make sure it is true and fair.
Plagiarism risk in regard to the content
AI technologies can assist researchers in putting together a rough draught of a paper rapidly, but they can also be used to swiftly copy text from sources without properly citing them. It is crucial for researchers to be aware of this potential for plagiarism and take the appropriate precautions to guarantee that all sources are correctly mentioned and given credit.
Inclusion of text that is offensive or harmful
The possibility for ChatGPT to produce offensive or dangerous text is a related worry. The OpenAI chatbot may use foul or dangerous language in the text it generates if the training dataset contains any such language. Researchers who are in charge of making sure the material is respectful and inclusive may run into difficulties because of this.
Concerns regarding the academic text’s quality
Although ChatGPT can generate citations and references and even automate academic writing, it cannot ensure accurate, high-quality research. Academics should be aware of this shortcoming and thoroughly revise the automated content to ensure that it is trustworthy, reproducible, scientifically accurate, and satisfies all other strict quality requirements necessary for submission.
In conclusion, it is important to use AI technologies such as ChatGPT professionally and ethically while writing research papers. Understanding that these tools are designed to supplement the researcher’s abilities rather than completely replace them can help researchers make the best choices for their research.
What do you think about using ChatGPT for academic writing and research, does it pose an ethical risk?
