How not to write an academic paper with the help of ChatGPT

How does ChatGPT actually work?

To understand the issues that can arise when writing academic papers using ChatGPT, it is important to understand how the tool works. When we look at how large language models actually work, we can think of it as a smart helper that can read and write very well. This robot has learned how to answer questions and create texts by reading a huge number of books, articles and websites. Let's zoom in a bit more.

Basic Architecture

ChatGPT is based on an architecture called Transformer, first introduced by Vaswani et al. in 2017. Transformers are deep neural networks that are particularly efficient for processing sequences of data such as text. The main components of a Transformer are:
Self-Attention Mechanism: This mechanism allows the model to pay attention to different parts of the input sequence while generating the output. This means that the model can perceive the context of the entire text when generating each word.
Encoders and Decoders:A typical Transformer consists of several blocks of encoders and decoders. Encoders process the input text and decoders generate the output text. In the case of ChatGPT, only the decoder part is used because its job is to generate text based on the input provided.

Training the model

ChatGPT, specifically its GPT-4 version, was trained on a large amount of text data from the Internet. It was trained as follows:
Data collection:The model was trained on a huge corpus of texts that included various types of texts including books, articles, web pages and other text sources.
Unsupervised learning:The model learns to predict the next word in a text based on the previous words. This process is called language modeling. The model iteratively adjusts itself to minimize the difference between its predictions and the actual words in the training data.
Fine-tuning: After initial training, the model can be fine-tuned on specific tasks or domains to improve its performance in specific applications.

Text Generation

For text generation, ChatGPT uses the following principles:
Next Word Prediction:The model predicts the next word based on the previous context. In doing so, it uses a probability distribution that determines how likely it is that a given word will follow.
Sampling: A word is randomly selected from the probability distribution to be added to the output. This process can be guided by various strategies, such as "top-k" or "top-p" sampling, which restrict the selection to the most likely words.
Coherence Control: To make the text coherent and meaningful, the model uses the context of the entire text to generate each new word. This includes complex structures such as sentences and paragraphs.

So ChatGPT or any other large language model does not use intelligence, but only probabilities when generating text. It can't produce new ideas, only predict the subsequent section of text. When using ChatGPT for academic purposes, students and researchers should carefully validate the text generated and use the model as a supplementary tool, not as a main source of information.

Dangers of using ChatGPT for academic writing

One of the main risks of using ChatGPT for academic writing is the potential for plagiarism. ChatGPT generates texts based on the huge amount of data that has been used to train it. This can lead to the resulting text containing sections that are directly taken from other sources without being properly cited. Students and academics should be particularly careful to avoid inadvertently copying texts, which can have serious consequences for their academic integrity.

There is also a risk of text copying - artificially generated texts may contain passages that are similar or identical to existing texts. Without careful checking, these texts could be considered plagiarized. Therefore, it is essential to perform thorough originality checks using plagiarism detection tools to ensure that all content is indeed original or properly cited.

In addition, there are also quality issues. ChatGPT-generated texts often lack the depth and critical analysis that is essential for academic writing. The AI may provide superficial answers that are not detailed or analytical enough. AI models such as ChatGPT can generate texts that are superficial and underdeveloped. These texts lack the in-depth analysis and critical thinking that are key to quality academic work. Critical thinking and deep analysis are essential for producing quality academic texts. Authors should actively add their own critical thinking skills to ChatGPT-generated texts, as well as perform in-depth analysis and interpretation of data and information.

Ethical and Academic Principles

There are also ethical and academic principles to consider when creating academic papers using AI tools. These principles ensure that the resulting work is not only high quality and accurate, but also ethically acceptable and original.

One of the key aspects of academic work is originality. The use of ChatGPT should not lead to direct copying of texts without proper acknowledgement of sources. Although ChatGPT generates text based on extensive data, it is essential that the author of the paper carefully checks the generated content to ensure that plagiarism is not taking place. Sources and references to original material should always be cited when used.

It is important to be transparent when using ChatGPT or other AI tools. The author should clearly state how and where these tools were used in the process of creating the paper. This transparency helps readers better understand how the work was created and ensures that the use of AI is not hidden or obfuscated.

Although ChatGPT can generate text, the author is always primarily responsible for the content of the academic work. The author should be able to explain and defend all parts of the thesis, including those created with AI. This also includes the ability to answer questions about methodology and interpretation of results.

The use of tools such as ChatGPT should be seen as an aid, not as a substitute for actual thinking and research. Students should be encouraged to learn and develop their skills, not rely on AI as their main source of information and inspiration. ChatGPT can be a useful tool, but the real educational value lies in active engagement and critical thinking.

How to use ChatGPT effectively in academic writing

How to use AI for inspiration and brainstorming

One of the first steps in writing an academic paper is generating ideas and finding inspiration. ChatGPT can be extremely useful at this stage. Students can input general topics or keywords and ChatGPT can provide a list of potential topics, research questions or directions to take. For example, if a student is considering writing about climate change, the AI can suggest specific subtopics such as "the impact of climate change on biodiversity" or "political aspects of climate change."

Using AI for language correction and revision

One of ChatGPT's greatest strengths is its ability to perform language correction and revision. After writing a text, a student can ask the AI to check for grammar, spelling and stylistics. ChatGPT can identify and correct errors, suggest better wording, and ensure that the text is clear and coherent. This is particularly useful for students who write in a foreign language or have weaker language skills.

Combining AI and Self Research

When writing an academic paper, it is essential to conduct your own research. ChatGPT can serve as a complement to this research by providing quick access to information and resources. However, it is important to remember that AI should not be the only source of information and, more importantly, that it can be a flawed source of information. Students should combine what they discover through ChatGPT with information from books, articles, and other credible sources.

Generating only short chunks

Using generative AI to create long chunks of work can be very counterproductive. For this reason, it is necessary to write the work only in very short chunks so that ChatGPT still perceives the context of the work and does not misinterpret the information given.

Examples of possible uses of AI in writing

Conclusion

Using tools like ChatGPT to write academic papers can be useful, but requires a careful and responsible approach. ChatGPT can help with idea generation, language correction, and text structure, but it cannot replace deep research, critical thinking, and original creation. Students should be cautious when using AI to avoid plagiarism and ensure that their work meets all academic and ethical standards. Transparency in the use of these tools is key, as is the combination of AI with self-directed research and critical analysis. Proper use of ChatGPT can greatly enrich the academic writing process when used as a support tool rather than as the main source of information.