The design sessions intended to bring people from different disciplines and sectors to critically consider how different communities might be drawn to learning about AI, and creatively iterate on ideas of how to promote agency with AI through different program formats in a variety of informal learning contexts.  

The conference attendees were divided into eight groups. Each group was given 6 cards, 2 from each category: 

1. 1. 1. Two AI Literacy learning goals connected to promoting agency in using AI, interrogating AI, and designing with AI, 

      2. A particular audience, and 

      3. A program format (e.g., an exhibit or afterschool program). 

The sessions were not intended to generate full-fledged program ideas that could be implemented. Instead, the goal was to provide evocative examples of how one might approach the design of AI literacy experiences in informal settings that center humans first, rather than the technology.

What follows are summaries of the design ideas that came out of this creative process.  We hope these examples can provide a process and a way of thinking about where informal educators might go next in developing programs and provide fodder for the AI education community to think out of the box about programming.  As you read through, think about “What unique needs do these ideas fill?”, “What are the concerns remaining?”, “Will this work with our communities?”.  

Given the recent rapid development of AI technology, we are eager to share these exploratory ideas with the informal learning and the AI fields, and are seeking feedback and insight from colleagues, experts, and community members to form a community of practice to ensure the human-centered approach will play a critical role in future AI education. 

Program Idea and How it Works

This team came up with an interactive kiosk where all family members get to input their preferences for what they would want in the neighborhood where they would like to live. The AI decision-making system would be trained exclusively on buzzfeed lists to recommend places, resulting in the high likelihood of output being biased and a mismatch. This kiosk could include multimodal features that show how the AI is getting trained on the data to make this transparent. The kiosk may also include less emotionally charged examples to orient family members to notice how inputs into the system influence the decisions the AI system makes.. Facilitators can discuss the importance of AI transparency, opt-out options, discuss where AI influences decisions in daily life, and suggest actions they can take like writing Congress or going to websites to learn more (i.e., AI for All, a website version of Barrio Broker for hands-on exploration after the visit, etc.) To make this activity accessible to multiple age groups in a family, the team explored the idea of using object recognition for presenting categories to families (e.g., if they held up a dolphin, that would signify living by the ocean as an important criterion).

Program Idea and How It Works

This team’s program idea was based on a real-life experience of one of the members who had an interaction with a parent and teacher about the use of ChatGPT in school and its role in “cheating”. The team was curious to explore what the role of education is in AI, and what AI's role is in education. The program was designed for students and teachers encountering tensions around the use of generative AI in school assessments.  

In the program, students and adults would be assigned to separate affinity groups to discuss whether the use of ChatGPT is cheating so they can be openly honest without the fear of repercussions. Students will be asked to explore how, why, and when to use ChatGPT within current pedagogy at school. Teachers and other adults would be asked to discuss the larger macro and historical forces that may inhibit rapid change in applying AI in the curriculum (school bureaucracy, traditional ideologies of learning, etc.). The hope is that these separate affinity group will eventually be brought together to reflect on their experiences and perspectives, creating an intergenerational space for all stakeholders (students, teachers, admin, parents/guardians) to more critically surface the kind of learning that is possible and AI’s place in it. The long-term goal is to develop transformative learning experiences that integrate the conscientious and appropriate use of ChatGPT/AI.  The key talking points for discussion may include the following:

• Youth & adults need safe, trusting, affinity spaces to process, explore, and reflect on the use of ChatGPT in learning environments

  • Grounding in Mindfulness

  • Understand why students are using it

  • Understand teachers’ fears, worries, concerns, inadequacies, burnout, stress, and constraints

  • What’s working/not working with current classroom curriculum, pedagogy, and assessments?

  • Prioritize the whole student/teacher (feelings, agency, body response, heart, spirit)

  • Meta-reflection on ChatGPT - affordances and limitations

• Engage in honest, reflective, inquiry-centered dialogue to transform traditional schooling structures and systems.

  • Why school? What are we learning? Why? How are students demonstrating their learning?

  • Developing meta-technological and meta-educational reflexive fluency, using that fluency to change what education looks like locally

Concerns

• How to pull this off in a low-tech manner that could take place in settings where youth like to gather.  

• How to successfully balance having youth explore important issues underlying surveillance while at the same time promoting agency by having youths make choices about the data they share.

Program Idea and How It Work

This team’s idea focuses on high school students who want to flirt and explore career opportunities. The problems they are addressing include demystifying career pathways and looking at what the job market will be like in the future (and how is AI going to impact it).  It is the antidote to the usual boring career exploration events.  This group wants to explore the larger idea of how humans and machines understand the world through entropy (the number of possibilities that can be experienced or produced). Both humans and machines learn from a lot of input but machines need a lot more entropy to understand the world. Entropy was included in this project to reflect what happens when machines don’t have enough data. 

The program will consist of a self-guided (with your date) immersive experience or a portable pop-up (it can be set up in various stations). The AI tool will have a tactile combination of human experiences as well as a digital interface and will be future thinking. 

The experience itself includes: 

1. 1. Choosing objects representing what you love and want

   2. AI takes picture of the objects and the user describes what the objects mean to them 

   3. ChatGPT generates a career narrative based on these choices and descriptions

   4. DAL-E generates images of that narrative

   5. Output shows people what that actual job would be

   6. Shows how whole life would be created by ChatGPT (prompted by a question: Is this a good life?)

Program Idea and How It Works

How can we develop prompt engineering skills among middle school students at NYSCI during a pop-up community experience? This team translated the original card “prompt engineering skills” into a game of Jeopardy.  Topics in this game would be relevant to the local community and would be done in multiple languages. 

Participants will first pick a category (e.g., “books for a hundred dollars” and the answer is “Harry Potter series”) and then find out what is the prompt that leads to this answer (e.g., participants may guess the prompts be: “Children's books adapted into 8 films”, “Who is the greatest youth wizard?”, “Which book generated the most box office revenue?”, etc.). Participants would then put these guesses into the chat to see what the actual prompt that was used (in this case, the correct answer was: “What is the greatest book of all time for middle schoolers in the United States?”). 

Questions with higher points can get more complicated than lower ones (e.g., 200-point questions are harder than 100-point questions). For example, a prompt for 400 points can be: “The Night Diary” and the potential guesses can be “What was the greatest young adult science fiction novel in the last 10 years?”, “What is a good book to explore the night sky?”, “What is a book that middle schoolers would find scary?”). And the answer, in this case, was: “The greatest book of all time where both main characters aren’t white”. 

Where to Go Next?

This design idea raises interesting questions about geneolgical studies and what AI’s role can be in that as well as how stories can serve as data.  It would be interesting to think about what families from different cultures might do with a tool like this.

Program Idea and How It Works

This team designed a program for international high school students in informal learning environments to explore why diversity in AI matters. Since AI is going to affect high school students more than any other generation, the team wanted the students to feel in control (how do they affect the outcomes rather than being affected by the outcomes of AI) and also see how their own cultural perspectives and experiences can affect what kinds of outputs and decisions AI systems provides. The core idea being explored is the need for diversity and the different layers of diversity needed for AI to work—e.g., diversity in creators and users of AI as well as how culture, race, gender, accessibility, etc. affect how data and information get represented and analyzed through AI. 

When brainstorming, the group members considered students’ interests in future jobs (what job fields will look like), and the global markets that students would be interacting with (and even familiar with from their points of origin), and how it would impact them. To keep high school students’ interest, the group members designed a model tool using Teachable Machine in which students would input drawn data, and train the teachable machine on the data.  The tool could be applied to a range of data they create on subject matter young people care about or are familiar with (e.g., their houses, video games).   The tool would be a social game in order to make it relevant/personal to the students. 

The digital tool would be something that could run on an iPad or web browser and would consist of challenges based on drawings of different people (user interface) and youth could see how it changes when users add more data (more pictures). For example, the group came up with the idea of building a digital house classifier. The model will be trained to determine “what a house is” and with more data, the definition of a house could change (this also helps students understand what the concept of exclusion in AI is).  This tool can be made into a game where your team's model is ranked and then participants have to collect more data to improve their rank and make it more accurate (shows the need for more data diversity in AI). This team likes this idea because it can show how scale and diverse inputs influence results (e.g., international students may have very different ideas of what a typical house looks like based on their culture and countries) which can show up in the data.  In the group's mock-up of this idea, the teachable machine in fact did not recognize one of the team members’ drawings of her house in Jamaica which varied from other drawings of houses included from other cultures.

Extensions of This Idea:

• Using data training tools on music or slang that shows differences among countries- slang translator, classifier that lets you look at slang similarity- how slang terms get shortened, generational changes

• Writing chatbots using different rules

• Career placement tool where students write the rules and see the outcomes