The Whose AI? workshop is designed for young people aged 14-20 to learn more about text-based AI technologies such as Large Language Models (LLMs), and gain perspectives on their social implications.
In a series of four exploratory activities, participants express their own views on AI, investigate the inner workings of Large Language Models, experiment with data processing, and build understandings of what an ethical AI should entail.
To use the lesson plans and workshop materials below, please credit artist and educator Carrie Sijia Wang and theWhose AI? project (2024).
Participants working on their interactive chatbots and conversing with the facilitator. Image courtesy of CultureHub.
Workshop 01:
What Are You
Make an interactive chatbot that explores your own thoughts and feelings about AI.
Workshop 02:
Chat with Me
Create an AI version of yourself and test how well it represents you.
Workshop 03:
An Experiment with Data
Build an automated text bot using the chat history we generate in the workshop.
Workshop 04:
Ethical AI Manifesto
Engage in a discussion about speculative AI tools and write a manifesto for the future of AI.
Workshop 01: What Are You
Main Activity:
Write down a conversation between a human and an AI that explores what AI is and how it can affect our lives. Make the conversation into an interactive program using a template made with p5.js and RiveScript and share with the group.
Learning Objectives:
- Rethink our relationship with AI and explore possible visions of the future beyond the profit-driven and efficiency-seeking perspective;
- Get a lightweight introduction to JavaScript coding through a playful activity.
Tools and Equipment:
- p5.js: an open-source JavaScript library with the goal of making coding accessible to artists, designers, educators, and beginners. Participants will need to create a free account on the p5.js web editor.
- RiveScript: an open-source scripting language for chatbots with a friendly, easy to learn syntax.
- Equipment Needs: Presentation monitor or projection system, laptops or computers for participants with the Chrome browser installed.
Part 01: Write a Conversation Between a Human and an AI
Write a conversation between a human and an AI where you talk through your feelings, thoughts, curiosities, wonders, hopes, and critiques of AI technology. Work in pairs, one person can play the role of the human, the other can play the role of the AI. Or you can discuss and write for both roles together.
Use this Writing Template.
Part 02: Code the Conversation into a Rule-Based Chatbot
- Duplicate the Rule-Based Chatbot Template into your p5 account.
- Use this RiveScript Formatter to format the lines you wrote for RiveScript.
- Add the formatted lines to the bottom of your copy of the Rule-Based Chatbot Template. Important: In RiveScript, triggers (user’s lines) are ALWAYS lowercased & without punctuations.
- Save your chatbot: File > Save or Command + S
- Test your chatbot: File > Share > Fullscreen
Part 03: Perform the Conversation
Share the rule-based chatbot you created with the group by performing the conversation. Type the user lines you wrote into the input field, and hit the “return” key to get the prewritten AI responses.
Workshop 02: Chat with Me
Main Activity:
Create an AI version of ourselves by giving a description of our personalities to OpenAI’s GPT model. Discuss how well the chatbot represents us.
Learning Objectives:
- Explore the possibility of customizing existing large language models to represent different voices;
- Form understandings of how Large Language Models work and how they’re biased;
- Engage in a simple coding exercise and learn a few programming concepts.
Tools and Equipment:
- p5.js: an open-source JavaScript library with the goal of making coding accessible to artists, designers, educators, and beginners. Participants will need to create a free account on the p5.js web editor.
- OpenAI’s API (Application Programming Interface) connected to its GPT (Generative Pre-trained Transformer) language model.
Note for the facilitator: In order to use this coding template, the workshop facilitator will need to have an account on OpenAI’s developer platform, put a minimum of $5 in the account (it’s usually enough for 20 people to use in a two-hour workshop), and generate an API key for the participants to use. - Equipment Needs: Presentation monitor or projection system, laptops or computers for participants with the Chrome browser installed.
Part 01: Design an AI Character Based on Yourself
Imagine an AI character that reflects parts of your own personality. Think about what traits you have and what qualities you’d like your AI to mirror. It could be an AI with your personality traits, like being funny, laid-back, driven, or detail-oriented. Or it could be an AI that mimics your tone of voice when you talk to your best friend, family, or teacher.
Summarize your design into a concise instruction for the language model to follow. Use this template for writing the instruction.
Part 02: Customize a GPT Chatbot with the Instruction
- Duplicate the Generative Chatbot Template into your p5 account.
(See on the left note for the facilitator.) - Replace the example instruction with your own (this will be sent to the AI model):
- Save your chatbot: File > Save or Command + S
- Interact with it. Test your chatbot: File > Share > Fullscreen
- Revise your instructions if needed to make it act more like the character you designed.
Part 03: Share What You Think
Share with the group what you think of the customized chatbot. Does it do a good job representing the character you designed? What worked well and what didn’t?
Workshop 03: An Experiment with Data
Main Activity:
Collectively produce a good amount of text-based data. Use the data to create a chatbot. Test the chatbot and discuss our understandings of data.
Learning Objectives:
- Explore what it means to collect data, process data, and use it to make an automated text-based AI;
- Discuss possibilities of using community-based data to create AI that represents diverse voices;
- Gain insights into the history of prediction models through hands-on experience with a Markov Chain model.
Tools and Equipment:
- p5.js: an open-source JavaScript library with the goal of making coding accessible to artists, designers, educators, and beginners. Participants will need to create a free account on the p5.js web editor.
- RiTa: an open-source library we will use to create a Markov Chain model based on our chat history.
- p5.speech: a simple p5 extension to provide Web Speech Synthesis for our program (so our bot can have a voice).
- Equipment Needs: Presentation monitor or projection system, laptops or computers for participants with the Chrome browser installed.
Part 01: Have a Text-Based Chat
Have a silent group chat for about 20-30 minutes to generate text data unique to us as a collective.
Note for the facilitator: Choose a chat platform where you can save the chat history for later use. You can use the chat feature on Zoom, or create a public Google Doc for participants to type in.
Part 02: Turn Our Chat History into an Automated Bot Using Markov Chain
- Take a few minutes to learn what a Markov Chain is: slides introducing Markov Chain
- Duplicate the Data Experiment Markov Bot Template into your p5 account.
This template can help us make a Markov Chain model from the chat history, and automatically generate new sentences based on the model. - Copy and paste the chat history into the data.txt file of the template.
- Save your chatbot: File > Save or Command + S
- Test your chatbot: File > Share > Fullscreen
Part 03: Reflect on the Process of Turning Data into a Bot
Answer these questions:
Q1: How is the result similar to or different from what you thought it would be like? Any surprises?
Q2: How does the experience of listening to the bot differ from interacting with a bot trained on an UNKNOWN dataset?
Q3: How has your understanding of data changed or not changed after this experiment?
Workshop 04: Ethical AI Manifesto
Main Activity:
Look at a selection of speculative AI tools, have a discussion about the good and the harm they could bring*. Based on the discussion, write a collective “Ethical AI Manifesto.”
*This part of the activity is inspired by the Speculative Futures Activity of the MIT AI Ethics Curriculum.
Learning Objectives:
- Learn to evaluate AI technologies from socially conscious perspectives. Practice articulating both the benefits and potential risks of AI;
- Discuss and explore the values conversational AI should encompass in order to be ethical.
Tools and Equipment:
- Equipment Needs: Presentation monitor or projection system, laptops or pen and paper for participants to note down their thoughts.
Part 01: Discuss the Pros and Cons of Speculative AI Technologies
In small groups of 3-5 people, look into a selection of speculative AI tools, discuss the potential positive and negative changes they could being to our society. Share the key takeaways from your discussion with the bigger group.
Use these slides of speculative technologies and worksheets.
Part 02: Contribute to an Ethical AI Manifesto
- Work with your group, revisit everything you have discussed and written down during the speculative exercise. Identify a few of your strongest points.
- Consider the values reflected in these key points and transform them into statements that can be included in the collective Ethical AI Manifesto.
Part 03: Review and Revise the Manifesto
- Together, read through the contributions from all groups.
- If you disagree with any points, raise your hand to express concerns, and suggest edits or additions.
- Discuss the proposed changes as a group to reach a consensus.
- Finalize the Ethical AI Manifesto!