Academic work (including with AI tools)

Tips from our pros

Our Internal link opens in the same window:Center for Teaching and Learning (ZLL) accompanies our students on their journey at our university, from preparing to study, to taking exams. For example, the ZLL organises regular workshops on topics such as time management, learning methods, etc. 

The ZLL also supports lecturers in the development, implementation and further development of contemporary and innovative teaching models.

So that students can also benefit from the broad knowledge of the ZLL, our experts provide you with useful tips on all aspects of learning!

Do's and don'ts when using AI tools in academic work

In this article, we offer a practice-oriented overview of how to get from an initial topic idea to a written paper using AI tools. For a more in-depth insight, we recommend the numerous textbooks on “Academic Writing”. Many of these can be found in the HFU libraries.

Although the following steps of academic writing generally build on each other, it is completely normal to have to go back in between − for example, if you adjust the research question after you have researched the initial literature. In addition, the effort required for the individual steps can vary, depending on whether it is a small term paper (e.g. ten pages) or a much more extensive thesis (which is also “rewarded” with many more ECTS).

Basic rules:

  • Ask your supervisor if you are allowed to use AI tools. At HFU, students are advised by their supervisors as to whether AI tools can be used.
  • If yes, you should clarify how the use must be declared.

Our recommendation

Many of the examples mentioned here can be found in the hochschuldidaktik akademie self-study course:
https://hochschuldidaktik-akademie.de/hausarbeiten-und-essays-mit-ki-so-schreibst-du-bessere-arbeiten/

A helpful overview of AI tools for academic work can be found in the book by Bucher et al. (2024) from page 38. Full access for HFU members via the HFU Library.

First I'd like to introduce you to Anna

Anna is studying “Marketing & Sales” in her third semester. For her upcoming term paper, she has chosen the topic “The influence of social media on purchasing behavior: Opportunities and challenges for modern marketing”. The scope of her planned paper is 20 pages. 

Finding a topic and corresponding research question

Anna uses ChatGP to improve her research question. She gives ChatGPT a specific role, namely that of a competent but also strictly evaluating professor, and describes her request very specifically. She then asks ChatGPT for possible questions for her planned online survey and for an outline. 
To the prompt and ChatGPT's answers:

https://chatgpt.com/share/6745c879-ee54-8002-8851-41e289bca398

Anna has now further specified her research question with the help of ChatGPT as follows:
 „Wie beeinflussen Social-Media-Beiträge über den Klimawandel und Werbung das Reiseverhalten von Touristen in Bezug auf die Wahl umweltfreundlicher Reiseziele?“ (“How do social media posts about climate change and advertising influence tourists' travel behavior in terms of choosing eco-friendly destinations?”)

In the next step, Anna deals with the further methodological procedure for answering her research question. 

Methodological approach

Anna is already clear about how she wants to go about implementing her 20-page paper: first, she wants to present what has already been written on social media posts about climate change in relation to the travel behavior of tourists. She also wants to conduct an online survey, evaluate it and discuss the results.

Scientific papers can be purely theoretical or also contain a practical or empirical part. The requirements for the scope of the literature research and the required number of sources may vary slightly depending on the subject.

Tip: Clarify the expectations regarding the scope of the literature research and the number of sources required.

Literature research in databases and with the help of AI tools

A systematic literature search is a central component of every scientific paper (Bucher et al., 2024, p. 14 f.). ChatGPT describes this as follows: “Systematic in this context means that the search is structured, targeted and comprehensible. This is crucial in order to achieve similar results to other researchers and to ensure scientific quality.”

Anna's research question

Anna wants to investigate how social media advertising and sustainability posts influence the travel behaviour of tourists. To do this, she has to explain in her literature section whether there are already studies that deal with this issue. To make her research transparent and comprehensible − for herself and for her professor − Anna records the following information in a table:

  • What search criteria and keywords did I use?
  • Which databases did I search in?
  • Why did I choose certain sources?

Search strings

Searching with search strings is a very effective method. Search strings are search terms in conjunction with so-called operators. Such search strings can be generated and improved with the help of AI. Anna proceeds as described in this External link opens in a new window:HelpCard , where she also finds possible databases in which she can search.

Another research option is google scholar (the advantages and disadvantages of google scholar are listed on the website of the External link opens in a new window:Zentralbibliothek Zürich).

Anna gets an insight into generating search strings with ChatGPT and using them in “Semantic Scholar” with the help of the video in Lesson 14: “Literature research with Semantic Scholar” of the External link opens in a new window:hochschuldidaktik akademie self-study course.

The snowball system

Another option for literature research is the so-called “snowball system”: When Anna reads an article, she can find further literature relevant to her research question in the bibliography.

AI tools

AI tools can also support Anna in her literature research, but Anna is fully aware that with all the AI tools she uses, it is important to note that...

  • AI tools can generate sources that do not exist.
  • The tools only find online and freely accessible articles and therefore there is a risk that the “relevant and recognized works” will be overlooked.
  • Registration is usually required. Anna thinks very carefully about what data traces she wants to leave behind. She is careful never to enter personal data of other people (e.g. name, telephone number)!

(see also video “Lesson 14: Step 4 - Systematic literature research with AI tools” from the External link opens in a new window:hochschuldidaktik akademie self-study course).

Anna has the limitations of generative AI in mind during the entire process! (to the "Internal link opens in the same window:Dealing with ChatGPT" Insight entry) and she only ever uses AI research tools in addition to ‘real’ database research!

With all the examples mentioned below, it should be noted that developments are very fast-moving: Some tools are no longer on the market at some point, new ones are added. AI search engines such as External link opens in a new window:https://ki-suche.io/ or https://theresanaiforthat.com are therefore helpful (not only for text and research AI tools).

Examples of AI tools in the literature search

Anna has started researching on External link opens in a new window:elicit.com and External link opens in a new window:consensus.app:

 

elicit.com

Registration required. After entering the research question, the free version of elicit.com searches for suitable articles and displays a brief summary. In addition, various criteria of the articles found can be compared in a table, for example the method or the “main findings”.

consensus.app

SciSpace

External link opens in a new window:typeset.io

The video in Lesson 14: “Literature research with Scispace”, in the External link opens in a new window:hochschuldidaktik akademie self-study course offers an insight

solidpoint.ai

Examples of AI research tools with visualisation

The special feature of the following AI literature research tools is that these tools visualise networks based on a paper: How similar are they? How often are they cited? They also show a list of previous and subsequent works.

 

Researchrabbit.ai

Free tool for literature research (and literature management with connection to zotero). 
Information on Research Rabbit of the RHET AI Center (Center for Rhetorical Science Communication Research on Artificial Intelligence)

https://rhet.ai/2023/08/15/ki-tools-im-test-researchrabbit/

connectedpapers

External link opens in a new window:Connected papers (some graphs per month are free)

Tutorial: External link opens in a new window:“First Time User's Guide to Connected Papers for Research“  (English, approx. 11 min, you will be redirected to YouTube)

litmaps.com

inciteful.xyz

You can use External link opens in a new window:inciteful to build a network or a map.

You can find answers to many of your questions in the External link opens in a new window:inciteful help section.

Reading and understanding texts

Anna has now found helpful literature for her research question. She has researched the HFU BOSS portal as well as various AI tools. Her paper should be 20 pages in total. As a general rule of thumb: one to two sources per page. Anna's term paper is not a Bachelor's thesis in terms of scope, so she will not be able to read 30 sources in depth. However, she plans to study around ten sources intensively. Thanks to the above-mentioned quick-ball system, Anna finds further relevant literature by reading individual articles and the literature cited in them.

(see also the video “Lesson 14: Step 4 - Systematic literature research with AI tools” from the External link opens in a new window:hochschuldidaktik akademie self-study course)

Anna remembers that she once heard that there are various reading techniques for efficient reading, such as “skimming” or the “SQ3R method”. Anna refreshes her knowledge of reading techniques with the following video and also learns how AI can help with text reading comprehension: 
“Lesson 15: “Step 5 - Reading techniques” of the External link opens in a new window:hochschuldidaktik akademie self-study course)

After Anna has worked intensively with the sources she has found, she moves on to the next step: creating an outline.

Creating an outline

Anna would also like to be supported by text-generating AI when creating her outlines. In addition to ChatGPT from OpenAI, the following other tools could be considered: Gemini (Google), Co-Pilot (Microsoft), Claude (Anthropic), LeChat (Mistral) (open source) and Llama (Meta).
Before Anna starts creating the outline, she watches the following video, which explains what she has to pay attention to when creating an outline, depending on the scope of the intended term paper:

Video Lesson 17: “Step 7 - Finding a structure and creating an outline” and “Creating an outline with ChatGPT” of the External link opens in a new window:hochschuldidaktik akademie self-study course).

Creating text

Now Anna has to write the text. She already has a starting point thanks to the outline created by ChatGPT. Nevertheless, she finds it difficult to get into writing. She realises that she is not really making any progress, as she has to go back a few steps to read literature or look for new ones.

This gives her the feeling that she's not really making any progress. That's where text AI comes in handy... How easy it would be if ChatGPT produced the appropriate text directly for the outline provided. Apart from the fact that this is not an independent service and therefore problematic in terms of examination law, there is a great risk that ChatGPT will also not write a text at a particularly good level.

Anna deals with the question of what risks there are in text production with AI. The following video helps her with this: „External link opens in a new window:Lektion 18 Schritt 8 – Schreiben I: Rohtext erstellen“ (“Lesson 18 Step 8 - Writing I: Creating raw text”)

Ok, having a text generated exclusively by generative AI is not an option for Anna. But what to do next? Don't worry! All these thoughts that Anna has are quite typical. Anna realises that literature research and reading are also part of the work involved in writing a paper The more time you put into reading and filtering out the relevant passages to answer the research question, the more sophisticated the paper will be and the faster you will make progress in the actual formulation later on.

Nevertheless, Anna is frustrated that she has still not put any text on paper at the end of the day. Anna seeks advice from her supervising professor and she gives her the following tip: “Just start writing at the point where it is easiest for you. An academic paper is rarely written strictly from front to back.

As a first step, summarise the relevant parts of the chapters and articles you have read. Then formulate your research question and think about the methodological approach for your planned survey. Just start writing, even if only in keywords at first. It is important to find a starting point. You can work it out, refine it and put it together later.”

This helps Anna and she actually starts summarising the central statements from her sources. She notes from which source and from which page she has taken the ideas. This prevents her from plagiarising and allows her to see for herself which text she is referring to.

And Anna realises − text writing needs to be practiced! She often sticks very close to the text of the original source during her first attempts. But Anna also realises that practice makes perfect! It is important to work “cleanly” with the sources right from the start and it was really helpful for Anna that her fellow student Tim proofread a few paragraphs of her term paper and gave her feedback on her work. Both have read several books on the subject of academic writing” and help each other.

Finalisation: Text revision

Spelling / grammar check and rewording

Anna has done it! She has written her 20 pages and is satisfied. As she uses the spell checker in WORD, she sees no need to use an AI tool (such as External link opens in a new window:scribbr.de/rechtschreibpruefung, External link opens in a new window:mentor.duden) at this point.

Tools such as External link opens in a new window:languagetool.org or https://rechtschreibpruefung24.de/ offer a spell checker as well as rewording suggestions. Important: The use of AI tools for rephrasing must also be specified in the list of resources.

Feedback from AI

Before Anna revises and optimises her text with the help of AI, she watches the following video: These videos at the end of the self-study course https://hochschuldidaktik-akademie.de/hausarbeiten-und-essays-mit-ki-so-schreibst-du-bessere-arbeiten/
“Lesson 20 Step 10 - Writing III: Revising and optimizing”

Please note: It is important to note here that Anna must document the use of AI in the list of aids.

Create a list of references

Although we list this entry at the very end, it makes sense to organise literature directly when researching in a literature management system such as Zotero, Citavi or EndNote. Anna has done this very conscientiously throughout her entire process. She can now create a bibliography from the reference management system at the touch of a button. This saves a lot of time in the end and ensures that Anna can submit her term paper on time.

Anna can only recommend AI tools to a limited extent at this point:

  • External link opens in a new window:QillBot is an option for the APA citation style: after entering an ISBN or a title, for example, the bibliography is cited in APA.
  • Anna has also tried using ChatGPT and similar AI to create a bibliography according to a specific citation style. But: Anna has noticed that the formating can be wrong and you have to check everything again.

Anna comes to the conclusion that the reference management software was the most reliable method for creating the bibliography. Happy and satisfied, she hands in her term paper to her professor and is already looking forward to the feedback.

Sources and recommended reading