AI Tools workshop

• Informatiemanagement
• HBO
• Studiejaar 4

Research 3.3 AI workshop

Objectives

• Determine possibilities, limitations and risks of using AI tools 
• Understand ethical aspects of working with AI (plagiarism, honesty, originality of your work) 
• Share ideas on how to use AI effectively and responsibly 
• Learn how AI may be used within this module and how to indicate how and where you have used AI

Do you already use AI in your studies?

Often

Sometimes

Rarely

Never

Please describe your experience using AI in your studies so far.

ChatGPT in short

ChatGPT is an auto-generative chat that extracts data from textbooks, newspapers, websites and different articles​.

Although not the first language based model, GPT enables coherent responses (natural language).

SOURCE: https://inholland.sharepoint.com/:p:/r/teams/DocentenBachelorofInformaticaInformationTechnology/_layouts/15/Doc.aspx?sourcedoc=%7B0C3BA08F-A317-46CE-9D0E-6A2A0B9533E6%7D&file=AI%20lecture%20development%20(002).pptx&action=edit&mobileredirect=true

Note down as many potential ethical dilemmas you can think of (related to using AI tools in your studies)

KEEP IN MIND:

• chatGPT does not own the material it generates/it is not the original source (it should not in itself be cited according to APA without the original source included)
• there are potential privacy concerns and copyright issues 
• see InHolland student statement
• AI as a tool for students

file:///C:/Users/Ashley.FoyTervoort/OneDrive%20-%20Hogeschool%20Inholland/Downloads/AI%20as%20a%20tool%20for%20students_ENG_studentstatement_INH%20(1).pdf

GROUP EXERCISE:

In a group of 3-4 students, think of as many benefits and risks you can (bullet points) when it comes to using AI tools (think first about your research and broaden only if you're having trouble thinking of examples). Make sure one student has all points written down!

Using ChatGPT or similar AI-powered tools in university studies comes with both risks and benefits. It's important to carefully consider these factors before incorporating AI tools into your academic work. Here are some main risks and benefits:

Benefits:

Enhanced Productivity: ChatGPT can help you quickly generate ideas, outline papers, and provide initial drafts. It can significantly speed up the brainstorming and writing process.

Idea Generation: When you're stuck or need inspiration, ChatGPT can provide creative insights and suggestions for your assignments or projects.

Learning Aid: Interacting with ChatGPT can lead to a deeper understanding of topics. It can help explain concepts you're struggling with in a different way, potentially enhancing your comprehension.

Language Improvement: Regularly using AI for writing can help improve your grammar, vocabulary, and overall writing skills as you observe the suggestions and corrections it offers.

Time Management: AI tools can help you manage your time better by assisting with repetitive tasks, allowing you to focus on more complex aspects of your studies.

Risks:

Plagiarism: Overreliance on AI-generated content without proper citation can lead to unintentional plagiarism. It's important to understand how to use AI-generated content ethically and responsibly.

Lack of Originality: If you rely heavily on AI tools, your work might lack your personal voice and originality, as it's influenced by the AI's suggestions.

Inaccurate Information: AI-generated content might not always provide accurate information. It's crucial to verify facts and information from reliable sources before including them in your work.

Dependency: Continuous reliance on AI tools might hinder your critical thinking and problem-solving skills. It's important to balance AI assistance with your own cognitive efforts.

Privacy and Data Security: If you're using an online AI tool, there could be privacy concerns regarding the data you input. Ensure that the platform you're using adheres to appropriate data protection standards.

Limited Contextual Understanding: While AI has improved, it might still misinterpret context and provide incorrect or irrelevant suggestions. Human judgment is necessary to validate the quality of suggestions.

Best Practices:

Supplement, Don't Replace: Use AI tools as supplements to your learning process. Let them assist you, but don't replace the core of your own critical thinking and learning.

Cite Appropriately: If you use AI-generated content, make sure to properly attribute it. Treat AI-generated content like you would any other external source.

Double-Check Information: Always fact-check and verify information from reliable sources. Don't blindly trust AI-generated content for accuracy.

Maintain Originality: Don't solely rely on AI for your creative work. Ensure your personal voice and ideas shine through in your assignments.

Develop Strong Writing Skills: While AI can help with grammar and vocabulary, focus on developing your writing skills through practice, feedback, and learning from your mistakes.

In summary, using ChatGPT in your university studies can be beneficial for enhancing productivity and learning, but it's essential to be aware of the risks, especially in terms of plagiarism, accuracy, and originality. Finding the right balance between AI assistance and your own efforts is key to making the most of these tools.

What are the benefits of chatGPT/AI tools?

Language understanding: ​

Models comprehend and interpret human language effectively.​

They produce coherent and contextually relevant text.​

Scale and knowledge: They have access to vast knowledge bases for information retrieval.​

Continuous learning: Models can be updated with new data for improved performance.​

Efficiency and accessibility: They process text rapidly and are accessible through APIs.​

Assistive capabilities: Models aid in translation, grammar correction, and content generation.​

Personalization: Models can be tailored to individual preferences and writing styles.​

Innovation and creativity: They inspire new ideas and offer alternative perspectives.

Potential benefits:

Enhanced Productivity: ChatGPT can help you quickly generate ideas, outline papers, and provide initial drafts. It can significantly speed up the brainstorming and writing process.

Idea Generation: When you're stuck or need inspiration, ChatGPT can provide creative insights and suggestions for your assignments or projects.

Learning Aid: Interacting with ChatGPT can lead to a deeper understanding of topics. It can help explain concepts you're struggling with in a different way, potentially enhancing your comprehension.

Language Improvement: Regularly using AI for writing can help improve your grammar, vocabulary, and overall writing skills as you observe the suggestions and corrections it offers.

Time Management: AI tools can help you manage your time better by assisting with repetitive tasks, allowing you to focus on more complex aspects of your studies.

What risks are associated with using chatGPT/AI tools?

Accuracy​:

• Inaccurate outputs: Models may occasionally generate incorrect or misleading information.​
• Biased responses: There is a risk of models reflecting biases present in the training data.​
• Lack of commonsense reasoning: Models may struggle with understanding context or commonsense knowledge.​
• Limited real-world understanding: They may not have practical knowledge beyond what is present in their training data.​

Ethical concerns: ​

• Models can be used maliciously to spread misinformation or generate harmful content.​
• Not everyone has equal access to the technology or benefits provided by language-based AI models.​

Critical Thinking:​

• Overreliance on AI as a primary source of information, diminishing critical thinking skills​
• Lack of opportunities for students to independently verify or question AI-generated responses​

Potential risks:

Plagiarism: Overreliance on AI-generated content without proper citation can lead to unintentional plagiarism. It's important to understand how to use AI-generated content ethically and responsibly. Remember: ChatGPT in itself is not an original source!

Lack of Originality: If you rely heavily on AI tools, your work might lack your personal voice and originality, as it's influenced by the AI's suggestions.

Inaccurate Information: AI-generated content might not always provide accurate information. It's crucial to verify facts and information from reliable sources before including them in your work.

Dependency: Continuous reliance on AI tools might hinder your critical thinking and problem-solving skills. It's important to balance AI assistance with your own cognitive efforts.

Privacy and Data Security: If you're using an online AI tool, there could be privacy concerns regarding the data you input. Ensure that the platform you're using adheres to appropriate data protection standards.

Limited Contextual Understanding: While AI has improved, it might still misinterpret context and provide incorrect or irrelevant suggestions. Human judgment is necessary to validate the quality of suggestions.

Working with AI

Let's try.....

Open chatGPT and ask it to write a 500 word autobiography about yourself or a friend. Tell it only the following things about yourself- your date and city of birth, your current study program, your favorite singer, and your best friend's name. In the Teams collaborative space, share your thoughts on the overall accuracy/quality. List all/some of the inaccuracies or flaws in the autobiography it generated.

Have students use their initial prompt to start a conversation with ChatGPT. 

Ask them to analyze the responses generated and assess whether the information generated is relevant and useful. 

Ask them to note any problems or limitations in the model's responses.

Example

How can we improve our prompts for a better autobiography? What can we ask instead/add/specify?

Here are some tips to help you craft better prompts:

Be Clear and Specific:

Clearly state your question or request.

Specify any context or background information that is relevant to your query.

Use Complete Sentences:

Write your prompts in well-structured, complete sentences to make it easier for the model to understand your intent.

Start with a Greeting:

Begin your prompt with a polite greeting or salutation to set a friendly tone.

Provide Details:

If your question or request involves specific details, provide those details in your prompt.

Ask One Thing at a Time:

Avoid asking multiple questions in a single prompt. Keep it focused on one topic or request to get a more precise response.

Use User Messages:

You can use user messages (system and user) to guide the conversation and instruct the model. For example, you can use a system message to set the context or instruct the model to speak like a particular character.

Experiment and Iterate:

Don't be afraid to experiment with different prompts and variations to see which one produces the best results. It may take a few tries to get the response you want.

Adjust Temperature and Max Tokens:

Temperature controls the randomness of responses, with higher values (e.g., 0.8) making responses more creative but potentially less focused. Lower values (e.g., 0.2) make responses more deterministic and coherent.

Max tokens can be set to limit the length of the response. Be careful not to set it too low, as it might cut off the response prematurely.

Provide Feedback:

If you receive a response that's not what you expected or is inaccurate, you can provide feedback through the user interface. This helps improve the model over time.

Keep an Eye on Ethical Usage:

Be mindful of the content generated and ensure it adheres to ethical guidelines. Avoid promoting hate speech, misinformation, or any harmful content.

Use Prompt Engineering:

For more complex tasks, consider using techniques like "prompt engineering," which involves designing prompts that guide the model to follow specific steps or think in a particular way to achieve your desired outcome.

Stay Patient:

Sometimes, it may take a few interactions to get the exact response you want. Be patient and willing to refine your prompts as needed.

Remember that ChatGPT is a powerful language model, but it's not infallible. Crafting effective prompts requires some trial and error, and understanding its limitations will help you get the best results.

How to cite ChatGPT

We ask students to follow the examples provided by APA given that is the style guide we use. While you should never copy and paste directly from ChatGPT without editing and fact-checking, you should copy and paste the conversation of longer texts retrieved from AI in your appendix. Please see the following slides for examples of how to cite AI in-text and in the reference list. 

Example

https://apastyle.apa.org/blog/how-to-cite-chatgpt

Shorter text in-text citation

For shorter pieces of information (less than 40 words), you can cite AI as a direct quote, though you should substantiate the information with other academic sources.

Be sure to cite AI as you would cite any academic source. And always fact check the information!

APPENDIX

For our research modules and thesis, you will be required to include a brief summary of how you used AI in the appendix. You should also include how you used AI in your methodology sections of your report if you did use AI to answer research questions.

Examples:

What role did AI play in answering your main and sub questions?

How did you fact check the information you gained by using AI?

Did you use AI to write interview scripts or surveys? 

Explain clearly what role AI played and how you fact checked it all.

Two paragraph example on next slide.

Example

This is an AI generated example. It puts AI in a very positive light and is obvious it was written by AI, as was noted directly by my colleagues reviewing these slides. 

Please write these paragraphs explaining your AI use yourself, without using AI. You're not graded on it and the more honest you are, the easier it is to assess your work honestly and fairly.

In my research report, I harnessed the power of ChatGPT in several critical ways, fundamentally transforming the way I approached various research tasks. To begin with, I utilized ChatGPT to generate interview questions that were not only comprehensive but also tailored to the specific nuances of my research topic and interviewee. For instance, when investigating the impact of remote work on employee well-being, I input key keywords and phrases related to this topic into ChatGPT. The model generated questions such as, "How has the transition to remote work affected your work-life balance?" and "Can you share any specific challenges you've faced in maintaining your mental health while working remotely?" These questions not only guided my interviews effectively but also elicited rich, context-specific responses from participants. Some questions generated inspired me to write related follow up ones like "Has your employer noticed any decline in your work due to these challenges?" and "What solutions were offered to help you as a result?" resulting in a more in-depth interview.

Furthermore, when crafting survey questions, ChatGPT played a pivotal role in optimizing the survey instrument. For instance, in a study assessing customer satisfaction in the hospitality industry, I provided ChatGPT with information about the key factors I wanted to measure, such as cleanliness, staff friendliness, and food quality. The AI generated questions like, "On a scale of 1 to 5, how satisfied are you with the cleanliness of our facilities?" and "Please rate the friendliness of our staff from 'very unfriendly' to 'very friendly.'" These questions resulted in a structured and comprehensive survey that efficiently collected the necessary data. What I learned from this process was the immense time-saving potential of ChatGPT, allowing me to allocate more time to data analysis and interpretation. Moreover, the model's adaptability to various research topics and its ability to generate context-specific questions enriched the quality of my research, ultimately leading to more nuanced findings and a more comprehensive research report.

My understanding of ethical & responsible use of AI in my research is now:

Very clear

Somewhat clear

Unclear

Very unclear

Study responsibly!