AI in the writing process

Examples:

• If you are allowed to choose your own topic for your text, your first step is to consider what interests you personally, what you may have already been able to gain experience of in a course, which topics you find particularly exciting that the chairs in your subject are dealing with, etc. In this initial procedure AI is less helpful, since this step is all about your personal interests.
• Once you have found an approximate topic, you need to narrow it down. An AI can suggest different criteria to narrow down your topic. Alternatively, you can provide the AI with various criteria for narrowing down the topic (e.g. choosing a particular period in time, focusing on  a particular group of people, etc.). This way, the AI helps you to decide which criteria could be combined well with each other, for example.
• The aim of a scientific paper is not to simply write down everything you know about a topic. Rather, you are required to apply existing knowledge to a specific area and answer a concrete question. A precise research question provides a clear focus and helps you not to get lost in the multitude of information and sources on a topic. An AI can support you in developing a research question: For example, you can be "interviewed" about your writing project: By asking specific questions, the AI helps you to clarify your thoughts, question assumptions and arrive at a possible research question. 
  Keep in mind that - depending on the version and training data on which the AI is based - it may only be able to provide you with limited support for very current research questions.

Literature research is relevant at various points in the writing process, e.g.

• at the beginning in order to gain basic orientation and focus in a subject area
• after narrowing down the topic and developing a research question in order to search more specific scientific sources.

You can find information on specialized AI tools for literature research on the webpage "AI for literature research".

First, please notice the following important aspect: Although copyright-protected texts (e.g. PDFs with research articles) are made available to you via the university library or in another context, you are not allowed to simply upload them anywhere (e.g. to an AI application). In addition, you should not upload any documents that contain confidential and/or data protection-relevant information. You can find out more about this, for example, in the Moodle course KI-Kompetenz an der Uni Siegen.
Consequently the only texts that remain are those that are under a CC license. Uploading them to an AI application is generally unproblematic. 
 

Examples:

• In principle, AI can help you to capture the core statements of a text, especially with regard to your own research question. In addition, AI offers you the opportunity to ask text-related questions outside the classroom and to have terms explained to you (e.g. also in foreign-language texts). The AI is therefore a kind of assistant that can support your own critical examination of the research literature and thus improve your understanding of the text. However, this requires you to actively engage with the text. You can only draw conclusions for your own work if you take a close look at how researchers approach their studies. In addition, you can only check the accuracy of AI-generated statements if you have studied the content of a text and the logic of its argumentation in depth.
• In order to engage intensively with the research literature, it is helpful to write down your own thoughts on what you have read while reading. This will help you to develop your first rough drafts, which can be good preparatory work for your later text (see also the examples in the section on writing a rough draft).
• If you are collecting data yourself as part of your writing project, it is worth considering how an AI can support you. However, as the AI does not conduct research independently, it is all the more important that you remain responsible for your writing project and that you familiarize yourself in detail with the basic methods of data collection and analysis in your subject. On this basis, you can then get targeted support from AI, e.g. to uncover patterns and correlations in data, transcribe or visualize data. It is particularly advisable to use specialised tools for data analysis. Overviews such as those at https://www.futurepedia.io or https://theresanaifor that.com help you to select the right AI tool for a specific task. Particular care must be taken with regard to data protection, especially when using AI for data analysis: For example, if you have conducted interviews and have them transcribed by an AI, the people you have interviewed must be informed accordingly about the use of AI and give their consent. In order to provide your interviewees with sufficient information, you need to find out whether the respective AI provider stores the data on its server, where it is located, how long the data is stored, whether it is used to train the AI, etc.

Examples:

• Do you find it difficult to get started when writing your text? Do you experience the proverbial 'blank page' as a challenge? An AI can help you with these challenges. For example, it can suggest how to formulate your reading notes into complete sentences and paragraphs. It is important that you see the text generated by the AI only as a starting point for your further work on the text.
• Many thoughts only take shape when they are formulated. Writing is a medium of thought and a tool of insight. With this in mind, AI should only ever be used in combination with your own writing: Sometimes only arduous struggle to find the right wording can bring you closer to what you actually want to express, make you discover gaps in your argument, help you organize your thoughts and deepen as well as structure your understanding of the topic.
• AI can produce texts that sound convincing and often makes the resulting texts appear much more finished than they actually are. It can then be all the more difficult to let go of the AI's suggestion and develop your own argumentation and individual academic writing style .
• For example, you can tell the AI to suggest one or more alternative versions of a paragraph or chapter you have written yourself. Then compare the AI-generated versions with your own text: What do you like better about your own text and what do you perhaps like better about the AI-generated texts? Pay particular attention to which formulations express your own thoughts/arguments more precisely. Based on your considerations, create a new version in which you combine your own first version with parts of the AI-generated versions and edit them further.
• Another tip for your prompts: Always tell the AI the specific research question of your text and which function the respective paragraph or chapter fulfills in the overall context of your work (e.g. giving an example, stating or justifying a thesis, etc.). This makes it more likely that you will achieve accurate results.

Examples:

• A text ready for submission is not written in one go. Instead, you gradually approach the finished text over several rough drafts and revision steps during the writing process. Through these revision steps, you can gradually find out exactly what you want to express.
• Basically, an AI can support you on various levels of revision :
  • In terms of content: For example, let the AI take on the role of your supervising teacher and uncover potential weaknesses in the text from their perspective. Some aspects (e.g. uncovering disruptive redundancies and logical contradictions) are more suitable than those that require an understanding of the text that LLMs do not have (e.g. identifying places where information is missing).
  • Structural: On a structural level, an AI can help you, for example, to check whether each sub-chapter has a clear topic, each paragraph has a main message and whether the individual paragraphs are in a meaningful order that is easy for readers to follow.
  • Linguistic: The linguistic revision of a text is not only about spelling and grammar, but also about stylistic conventions, for example, which can differ from subject to subject. For instance, the use of the first person is accepted and common in some subjects, but should be avoided in others. In the natural sciences, you will often find a more neutral style with shorter sentences, whereas in the humanities, sentence structures tend to be more complex. The best way to develop an awareness of specific scientific language conventions in your subject is to carefully read and analyse scientific texts from that subject. An AI can support you in this analysis (e.g. by identifying the choice of words and typical syntactic structures for certain sections of text from your subject field), but it does not automatically know the linguistic and stylistic conventions that apply in your subject.
• During the revision process it is important that you ensure that the text you have written (with selective AI support) actually corresponds to your intentions, accurately reflects your argumentation, etc. In order for your text to really become your own and for you to be able to take responsibility for it, it is essential that you think for yourself about why you are still dissatisfied with a rough draft at one point or another, where something is not yet conclusive, etc.