Student and Faculty Guide for using Generative AI

According to UNESCO (United Nations Educational, Scientific and Cultural Organization), Artificial Intelligence (AI) has the potential to address significant challenges in education, revolutionize teaching and learning methods, and expedite progress that ensures inclusive and equitable quality education and promote lifelong learning opportunities for all.

When using AI chatbots and generative AI tools, never share:

• Personal information (SSN, birth date, address)
• Financial details
• Medical records
• Login credentials
• Confidential documents
• Private student/employee information

These services may store your conversations and use them for training. Treat AI interactions like public conversations - if you wouldn't post it publicly online, don't share it with an AI.

Need help determining what's safe to share? Contact the UO Libraries for guidance.

For more specialized information, please take a look at the Artificial Intelligence Resource Guide, created by UO's Teaching, Support and Innovation (TEP) Office. This resource provides instructors with suggestions and options for how to address AI use in their courses, plus links to additional resources.

Suggested Readings for Teacher Awareness

According to UNESCO's Guidance for generative AI in education and research report, at a high-level teachers should be aware of the following:

GenAI algorithms are trained on data that could primarily come from Global North data sources.

ChatGPT models are trained on data from online users which reflect the values and norms of the Global North, making them inappropriate for locally relevant AI algorithms in data-poor communities in many parts of the Global South or in more disadvantaged communities in the Global North.

How could this impact students? GenAI's outputs are limited in global perspectives. Students could only be exposed to prompt responses with Global North biases.

Dominant GenAI companies have not been required to go through rigorous independent academic reviews.

Legislation around GenAI usage is still in development. Typically, technology companies want to protect their intellectual property. Data sources are pulled from publicly available online resources like web pages, social media posts, photographs, videos, etc; however, these data sources or algorithms used by GenAI systems may not have been reviewed by academic communities. 

How could this impact students? Lack of efficacy and review of GenAI tools can lead to bots plagiarizing, sharing inaccurate information, and non-creditable sources that is presented to students as as trustworthy and valid when there is bias.

Using content leveraged as data sources may have been taken without consent from data copyright owners. There is limited transparency.

Citing sources is critical for academic integrity. GenAI technologies have typically been trained using data sources gathered and used without permission from creators or content owners. As of January 2024, there are several lawsuits happening including the New York Times suing Microsoft and OpenAI for copyright infringement. 

How could this impact students? They may be using GenAI content that was taken without credit or attribution given by a copyright holder. They could also not know where the information they're using in class has come from or strategies for how information is pieced together to make arguments and claims.

Lack of explainability

How GenAI comes up with responses to prompts is not easily explainable because information about how algorithms developed is proprietary information.

How could this impact students? How an answer was derived is important when explaining methodology and reason for answers. Without knowing reasons for answers there is obscurity that can lead to learning hindrances.

Lack of Privacy and Security

New AI models increasingly depend on vast amounts of data, including sensitive information from students and educators for specific purposes. The handling and potential commercialization of this data raise significant concerns about privacy and rights. It's imperative to uphold and respect students' data privacy in accordance with regulations such as FERPA, and other relevant data protection laws.

How could this impact students? Students using GenAI and inputting personal information for classroom assignments without understanding how the bot will use their data can cause privacy concerns that inviolate with data laws such as FERPA.

Opportunities for GenAI in Education

It's crucial to educate all students about AI as it will profoundly influence various aspects of our lives. Understanding how AI functions is essential for using these technologies responsibly in both learning and career paths. Given that we are still in the early stages of this technological disruption, marginalized students have the opportunity to lead in comprehending and utilizing AI. This could potentially lead to greater equity and economic mobility, rather than being left behind compared to more privileged peers. Therefore, incorporating digital citizenship, AI literacy, and responsible AI practices into existing curricula will be vital moving forward.

What is AI Literacy?

The ability to understand and critically evaluate the role, implications, and capabilities of artificial intelligence.

Resources for AI Literacy

• Casal-Otero, L., Catala, A., Fernández-Morante, C. et al. AI literacy in K-12: a systematic literature review. IJ STEM Ed 10, 29 (2023). https://doi.org/10.1186/s40594-023-00418-7

• ChatGPT is a GenAI language model capable of generating human-like text based on the input it receives. It uses a deep neural network trained on a massive dataset of text to generate responses to a wide variety of questions and prompts.

• DALL-E 2 generates high-quality images from text descriptions. It can also generate images based on combinations of different concepts or inputs.

• Sudowrite is a writing app that helps users create their own original writing. It's specifically an aid for writing fiction prose and is marketed as a "brainstorming tool" for writers.

• Jasper is another writing tool that helps users create content after providing prompts. It can be used to create multiple types of original content, including marketing copy, product descriptions and website text.

• QuillBot is a tool that helps users paraphrase their sentences, paragraphs, essays or other writing materials. It also has other built in resources, including a grammar checker, a citation creator, a translator and a plagiarism checker.

What are the different types of generative AI tools?

1. Text Generation:

   • Language Models: These models generate coherent and contextually relevant text. Examples include OpenAI’s GPT (Generative Pre-trained Transformer) models.

   • Chatbots: These AI tools generate text-based responses to user inputs in natural language, simulating conversation.

2. Image Generation:

   • GANs (Generative Adversarial Networks): GANs consist of a generator and a discriminator, trained in tandem to generate realistic images. They are commonly used for tasks like image-to-image translation and style transfer.

   • Conditional Image Generation: Models like BigGAN can generate images based on specific conditions or inputs.

3. Video Generation:

   • Video Prediction Models: These models generate future frames in a video sequence based on the input frames, useful in applications like video synthesis and prediction.

4. Music Generation:

   • MIDI Generation: AI models can generate musical pieces in MIDI format, imitating various styles or even composing original compositions.

5. Speech Generation:

   • Text-to-Speech (TTS) Systems: These tools convert written text into spoken words, and some advanced models can generate natural-sounding human-like voices.

6. Code Generation:

   • Code Completion Tools: These tools assist programmers by suggesting code snippets or completing partial code based on the context.

   • AutoML (Automated Machine Learning): AI tools that automate the process of machine learning model selection, hyperparameter tuning, and deployment.

7. Art and Design:

   • Style Transfer Models: These models apply artistic styles from one image to another.

   • Creative Adversarial Networks (CANs): Focus on generating novel and creative content, often used in artistic applications.

8. Data Augmentation:

   • Generative Data Augmentation: AI models that can generate new training examples to augment datasets, enhancing the performance of machine learning models.

9. Interactive Generative Tools:

   • Interactive Storytelling Models: AI tools that generate narratives or stories based on user inputs or predefined scenarios.

10. Robotics and Simulation:

    • Generative Models for Robot Control: Models that can generate control signals for robotic systems in various scenarios.