Time

The lesson is expected to take 30-60 minutes to complete.

About the lesson: 'Generative AI as Study Buddy'

This lesson provides suggestions on how you can use generative AI in connection with your studies. There are many forms of generative AI, but the following focuses on Microsoft Copilot, as KU has a licensing agreement with Microsoft.

Learning outcomes

When you have completed the lesson, we expect that you:

know how you can benefit from using generative AI models in your studies.
know what you should not use generative AI models for, in connection with your studies.
can take advantage of AI models as a sparring partner in your studies.
can critically assess the output from generative AI - and act accordingly.

Kilde

The lesson is produced by the University of Copenhagen as part of the University of Copenhagen's learning resources for digital literacy, 2023. The module is produced in collaboration with the University of Copenhagen Library and the Center for Online and Blended Learning at SUND.

Course directors:

Birgitte Munk, special consultant, University Library (KUB)
Muhamed Fajkovic, special consultant, University Library (KUB)
Søren Davidsen, special consultant, University Library (KUB)

_________________________________________________________

Before you start

Good advice and additional information

Read Guidelines and rules for using ChatGPT and similar technologies at UCPH
Read The University Library's guide on, Guidelines for Citation When Using ChatGPT
To create good prompts, you can follow the CLEAR principles, which involve being: Precise, Logical, Explicit, Adaptive, and Reflective. Read more here: Lo, Leo S. 2023. “The CLEAR path: A framework for enhancing information literacy through prompt engineering”. The Journal of Academic Librarianship 49 (4): 102720.
If you need an explanation of what generative AI is and how it works, check out this lesson.'The technology behind ChatGPT and large language models

_________________________________________________________

How has generative AI changed student life?

Listen to Dünya Tümtürk and Katrine Lindtner Andersen from Sociology, as well as Lukas Frimer Tholander from Computer Science, talk about how ChatGPT and Copilot have changed their daily lives as students. How do they use the technology to assist with their studies? How do they perceive their instructors' reception of AI? Is it clear to them what rules apply to their use of AI? Do they consider the technology's climate impact? And what about data security? [In Danish]

_________________________________________________________

How to use generative AI as a sparring partner

Billede: ChatGPT, DALL-E, 22/3-2024, promt: "AI som study buddy på universitetet"

Generative AI can be used for many different things, but below you can read more about how you can use the technology as a sparring partner in connection with your studies. The structure is an attempt to reflect where students will benefit the most from using generative AI, knowing that reality is often much more unstructured.

The examples below can be used in, for example ChatGPT, Microsoft Copilot (free when you are logged in with your KU username) and Google Gemini.

The prompt examples are in Danish, but you should be aware that in most cases you will get a better result if you prompt the AI in English.

Last but not least, you should be aware that all large language models are trained on specific data, and this can affect the answer you get. Therefore, it is always a good idea to double-check AI-generated information.

_________________________________________________________

Prompting-examples with Microsoft Copilot

NB! As a KU student, you have free access to Microsoft Copilot. You must be logged in to Microsoft with your KU username to access the latest version of Copilot and to ensure your data is protected by KU's data processing agreement with Microsoft.

_________________________________________________________

Example 1 - Try to create a study plan

Why: This can break down the course literature into smaller parts and thus help give you an overview of which literature you need to read and when.

When: You can use it when you need to structure your time.

How: Find the reading list for a given course and determine which literature you need to read from session to session. Remember that it may be necessary to include literature related to the course as a whole and not just to individual sessions.

Click here for examples by Microsoft Bing Copilot, generated 22. March 2024.

Læseplan – Q.png

Læseplan – A.png

_________________________________________________________

Example 2 - Literature search (keywords and search strings)

Why: You can use Copilot as a sparring partner in your literature search. This can be useful if you are in the very initial phase of your literature search but are having trouble coming up with good keywords. You can also use it to get feedback on your search strings when you are a little further along in your search process.

Keywords and search strings can then be used to conduct literature searches, for example, in soeg.kb.dk or in scientific databases.

Example of a search string: ("renewable energy" OR "solar panels") AND climate*

How: Start by identifying a number of keywords that encapsulate your preliminary problem area. There are no rules for how many words you should have, but around ten is a good start. Use this as the basis for your prompt.

If you are a bit further along in your search process, you need to convert your keywords into search strings. This means you need to consider how the different keywords should be weighted relative to each other. This is done using the so-called Boolean operators, namely AND, OR, and NOT. Remember that they should always be written in uppercase letters.

Once you have created a search string, you can ask Copilot for feedback on it.

When: This is especially helpful at the beginning of your literature search process. As you read more about the topic you are writing about, you will become better at assessing the strength of your keywords and the composition of your search strings.

Why: You can use Copilot as a sparring partner in your literature search. This can be useful if you are in the very initial phase of your literature search but are having trouble coming up with good keywords. You can also use it to get feedback on your search strings when you are a little further along in your search process.

Keywords and search strings can then be used to conduct literature searches, for example, in soeg.kb.dk or in scientific databases.

Example of a search string: ("renewable energy" OR "solar panels") AND climate*

How: Start by identifying a number of keywords that encapsulate your preliminary problem area. There are no rules for how many words you should have, but around ten is a good start. Use this as the basis for your prompt.

If you are a bit further along in your search process, you need to convert your keywords into search strings. This means you need to consider how the different keywords should be weighted relative to each other. This is done using the so-called Boolean operators, namely AND, OR, and NOT. Remember that they should always be written in uppercase letters.

Once you have created a search string, you can ask Copilot for feedback on it.

When: This is especially helpful at the beginning of your literature search process. As you read more about the topic you are writing about, you will become better at assessing the strength of your keywords and the composition of your search strings.

Click here for examples by Microsoft Bing Copilot, generated 27. March 2024.

_________________________________________________________

Example 3 - Language support

Why: To ensure your texts do not contain language errors, you can get a kind of "proofreading" from Copilot. This way, you avoid the impression of the text being diminished by mistakes that you, as the author, may not necessarily notice.

How: Insert the paragraph from your document that you would like feedback on. You should specifically indicate what kind of feedback you want.

You can ask to:

Be made aware of spelling or grammatical errors.
Know if the meaning of the text is clear enough.
Check if your understanding of a text in a foreign language is correct.
Get feedback on your text structure.

Do not ask Copilot to generate the text for you.

We recommend that you do not upload the entire text.

When: You can do this continuously, for example, before you send a draft of your problem statement to your supervisor, or when you get stuck with a section.

Click here for examples by Microsoft Bing Copilot, generated 27. March 2024.

sproglig sparring_1.JPG

sproglig sparring_2.JPG sproglig sparring_3.JPG

_________________________________________________________

Example 4 - Sparring on a Problem Statement

Why: It can be useful to get a second opinion on your problem statement, even from Copilot. Like any other form of sparring, having others look at it can strengthen your argumentation. You might spar with Copilot before discussing it with your supervisor or fellow students.

How: Ask Copilot to evaluate the problem statement you are considering working on. Provide it with information about the type of assignment, the requirements for the assignment (e.g., length), and what your own goals are with the assignment. Specify what you want feedback on (e.g., you can ask if the topic is relevant or too broad).

When: This can be especially helpful at the early stages of defining your problem statement, ensuring it is focused and relevant before you proceed further.

You should consider whether Copilot's advice is good enough, and you do not necessarily have to accept them. Also, be aware that supervisors will often ask you critical questions. Copilot will only do this if you explicitly ask for it.

When: You can spar with Copilot both in your initial phase and when you have progressed further and have a better grasp of the topic.

In the initial phase, you can ask for help in formulating or narrowing down the problem area.

Later in the process, you can get feedback on specific aspects of your problem statement.

Click here for examples by Microsoft Bing Copilot, generated 27. March 2024.

_________________________________________________________

Example 5 - Sparring on Theoretical Considerations

Why: AI tools can be excellent for navigating a large theoretical field or discovering theoretical approaches you are not familiar with. You can also spar with Copilot to test your own understanding of a theory or theorist.

How: Ask Copilot to explain a theoretical approach to you or to give you an overview of the key theories in an academic field. You can also ask Copilot if your interpretation of a topic is correct. Remember to specify what you need.

Always hold up any answers to alternative sources

When: You can spar with Copilot on this topic whenever you need a kick off to familiarize yourself with a theoretical field. This could be in preparation for classes, oral exams, or when considering the theoretical part of your assignment.

Click here for examples by Microsoft Bing Copilot, generated 29. March 2024.

_________________________________________________________

Example 6 - Sparring on Methodological Considerations

Why: Within each academic discipline, there are often several different methodological approaches to choose from. To ensure that you have selected the most suitable one or to avoid overlooking other relevant approaches, it can be beneficial to spar with Copilot.

How: Ask Copilot to either suggest an appropriate method or to evaluate whether the methodological approach you are considering is suitable for the topic you are working on.

When: It is a good idea to do this before a meeting with your supervisor. This way, you are prepared and can hear your supervisor's opinion on your methodological considerations.

Remember to double check any replies.

Click here for examples by Microsoft Bing Copilot, generated 29. March 2024.

_________________________________________________________

Example 7 - Sparring on a Specific Problem

Why: It can provide a draft solution to a problem. The problem could, for example, be related to coding.

If you use Copilot for programming assistance, you can get help understanding error messages or documentation.

How: Create context for Copilot and describe step-by-step what you are starting with and the desired result of the code. Do this over multiple prompts to ensure the chatbot "understands" each step correctly.

When: You can use this when you are almost able to write the code yourself or when you need answers to specific questions. It can be difficult to give the right prompts if the code needed to solve the task is too complex for you to understand. Likewise, it can be challenging to determine if the chatbot provides a correct answer.

Click here for examples by Microsoft Bing Copilot, generated 29. March 2024.

_________________________________________________________

How to Critically Approach Generative AI

Just like with everything else you do in connection with writing assignments, the use of Copilot should reflect good academic practice. Therefore, you need to be aware of the strengths and weaknesses of generative AI.

Read the following carefully before you start using AI in your studies. Be absolutely sure that you approach the technology responsibly and follow the applicable rules. Stay updated on as many platforms as possible, as information can quickly become outdated.

Source Criticism and the Technology Behind Generative AI and Large Language Models

Generative AI draws on vast amounts of data, but it is often unclear where this data comes from and what criteria were used to select it. This makes it more difficult to evaluate the reliability of AI-generated content.
Algorithms and Training Methods: Generative AI uses underlying algorithms and training methods on extensive text material that is often unknown to the public.
See, for example: The technology behind ChatGPT and large language models
Bias and Generalization: Generative AI can reproduce and amplify existing biases in the data it has been trained on. Identifying and addressing this requires special awareness of the algorithms' tendency to perpetuate potential prejudices. See, for example: When algorithms decide
Hallucination - the Ability to Fabricate Information: Generative AI can potentially fabricate information outside any existing sources. This requires an extra degree of caution and validity assessment.
Lack of Source Information: Generally, it is often hard to access or trust information about the specific sources generative AI has been trained on. This differs from conventional sources.
See, for example: Basic source criticism.
Continuous Development and Updating: Generative AI models are continually updated and improved, meaning that the results and potential for bias can also change over time. This requires ongoing attention to the technology's development.

Good academic practice

Be aware that Copilot is designed to always give you an answer, even if it is incorrect. In this way, Copilot can be considered what some have called an "overconfident peer," meaning a fellow student who occasionally has too much confidence in their own abilities.
Always check what the guidelines for using generative AI are for your specific program.
Do you have enough knowledge in an area to use Copilot in a qualified and critical manner?
What tasks are you using Copilot for? Is it, for example, to help with simple proofreading tasks, or is it for more demanding tasks such as concept clarification? The more advanced the level at which you use Copilot, the more critical you need to be.
When using Copilot for more advanced tasks, you should investigate how recent the basis of the data is. That is, how recent the data is that Copilot has been trained on.
Always compare Copilot's response with other sources. This is especially important when you are dealing with a topic you do not know much about.

May I upload texts to a large language model like ChatGPT?

Several of the large language models allow you to upload documents. Copilot does not yet have this functionality. If you want to use this feature with another AI, you should consider the following:

According to KU’s guidelines: “As a precautionary principle, one should avoid including research texts and new findings in their prompts until it is clarified whether and how users’ text input is integrated into the technology’s data basis.”
You can use AI to analyze interview transcriptions, but you should be aware that ChatGPT is not an official KU tool. Therefore, KU does not have a data processing agreement with OpenAI, the company behind ChatGPT.

____________________________________________________

Reflections

Discuss the following questions:

What is the difference between the guidance you can get from your advisor and the guidance you can get from Copilot?
Can Copilot be used in all subject areas?
Can Copilot be biased in the answers it gives?
What can you do if you are in doubt about the correctness of an answer you have received from Copilot?

____________________________________________________________

If you want to know more

“Activities for Students to Explore AI › Education Commons”. 2023. 6. april 2023. https://ec.plymouth.ac.uk/ai-activities-for-students/.

Deng, Jianyang, og Yijia Lin. 2022. “The benefits and challenges of ChatGPT: An overview”. Frontiers in Computing and Intelligent Systems 2 (2): 81–83.

Hau, Mark Friis. 2023. “ChatGPT: Den ultimative undervisningsassistent, du ikke vidste, du havde brug for”. 1. juli 2023. https://videnskab.dk/teknologi/chatgpt-den-ultimative-undervisningsassistent-du-ikke-vidste-du-havde-brug-for/.

Lo, Leo S. 2023. “The CLEAR path: A framework for enhancing information literacy through prompt engineering”. The Journal of Academic Librarianship 49 (4): 102720. https://doi.org/10.1016/j.acalib.2023.102720.

Lock, Samantha. 2022. “What Is AI Chatbot Phenomenon ChatGPT and Could It Replace Humans?” The Guardian, 5. december 2022, par. Technology. https://www.theguardian.com/technology/2022/dec/05/what-is-ai-chatbot-phenomenon-chatgpt-and-could-it-replace-humans.

“Tips on Using Generative AI”. 2023. Education - University of Kent. 16. oktober 2023. https://www.kent.ac.uk/education/using-generative-ai-at-kent/student-guidance/tips-on-using-generative-ai.

___________________________________________________________