Vermont AI Literacy Initiative · UVM
Every Output
Comes From Somewhere
Vermont AI Literacy Initiative (VALI)
Hands-on, visually engaging curriculum that teaches how AI works from the inside out — building helpful intuition and healthy skepticism for all skill levels, without significant math or computational pre-requisites. Designed for and in collaboration with diverse K–12+ populations. Piloting in Vermont, beginning in 2026.
Vermont AI Literacy Initiative
Demystifying AI because everyone deserves to understand it
As large language models (LLMs) rapidly become part of everyday life (at least, for now), all students and teachers need the foundation necessary to interact with them relatively safely, effectively, and responsibly — not just those in well-resourced environments. We're an interdisciplinary team at UVM's Vermont Complex Systems Institute with strengths in computational linguistics, deep learning, and STEM outreach. Our goal is to co-create a compact, informative, adaptable curriculum that explains the inner workings of AI using metaphor, visual aids, and hands-on activities alongside stakeholders from Vermont communities. Researchers studying AI are acquiring expertise, but that expertise is relatively siloed, and often contained in idiosyncratic representations. We want to translate key lessons from that expertise to accessible language that learners of all skill levels can engage with and apply.
K–12
Target audience
Vermont
Classroom pilots, beginning Fall 2026
Free
Open access, always (in development)
Learning objectives
What students will walk away with
01
Understanding the mechanics through metaphor and hands-on activities
Practicable intuition around embeddings, transformers, and plausible token generation.
02
Recognizing pitfalls & biases
Understanding where language models currently fail and how they enact bias through social and technical mechanisms.
03
Confidence and healthy skepticism
The confidence to use generative AI tools paired with the critical lens to question what they produce and how they produce it — building toward critical engagement with AI as students encounter it in a variety of contexts.
Curriculum series
Visual, hands-on, teacher-ready.
Our first 3 modules are in active development and will be available for classroom pilots beginning in Fall 2026. Each is designed as a 1–2 hour lesson, but can be dialed up or down in terms of content depth and required time. No prior AI knowledge required, for students or teachers. We plan to create open source, virtual materials that can be accessed online or printed, and physical manipulables and activity plans that won't require classroom access to computers.
Coming Fall 2026 · VALI
How AI generates language: stochasticity, determinism, the Distributional Hypothesis, and emergence
Architectural components such as embeddings, neural nets, and plausible token prediction objective functions, explained through metaphor, visual aids, and hands-on activities.
Coming Fall 2026 · VALI
Biases, pitfalls, and AI skepticism: embodied cognition and complex sociotechnical systems
Building actionable intuition for when and why AI tools "fail" — and why that matters. Students interrogate AI outputs and develop a critical vocabulary for navigating these tools in their own lives.
Coming Fall 2026 · VALI
Cultural stewardship and production: the extended mind, offloaded cognition, the commons, and the environment
Understanding the sociotechnical landscape within which AI tools are deployed — and why that matters. Students learn about the tradeoffs that could come with AI and its potential and actual dangers.
Curriculum tool · VALI
AI Tarot Cards
A set of 10–22 cards that abstract key AI concepts at a high level — crafted to support the hands-on lessons and activities, spark discussion, and give educators a flexible toolkit for critical thinking about AI. These cards aim to translate relatively siloed AI expertise into widely accessible and intellectually productive abstractions. Designed to complement existing approaches like Stanford d.school's Machineless Machine Learning toolkit that place more emphasis on the concrete mechanisms of AI.
Click to flip
Sample activities
What students might do
Activities are designed to be adapted by educators for different class sizes, ages, and time constraints. Examples in development:
- A digital golf-like game where students add words to a prompt to shift next-token probabilities to steer a ball into a hole — visualized as a ball rolling over a changing landscape altered by the addition of tokens to the prompt, analogizing the objective function as a kind of terraformer of the latent space.
- A networked "telephone" game where students model how feedback and error propagate during model training and learn the fundamental components of neural networks.
- Layering transparent acetate sheets to show how simple learned features combine into complex output, without any single layer being able to describe the final picture. This is a close analogy to matrix decomposition (SVD/PCA), which is integral to modern computation, including AI, and, on the other hand, a cautionary metaphor for how complex output can arise from simple pieces but be misinterpreted to impute Human Theory of Mind.
Specific module formats, delivery modes, and sequencing are in development. More details coming as the curriculum takes shape. Have ideas or want to be involved? Get in touch →
The team
Scientists, educators, artists, and STEM communicators
Based at UVM's Vermont Complex Systems Institute, with strengths in demystifying LLMs via computational linguistics and deep learning, building interactive and visually appealing computational stories, and engaging the broader community on STEM topics.
Dr. Julia Witte Zimmerman
Postdoctoral Associate in Artificial Intelligence and Computational Social Science · Co-lead
Calla Beauregard
PhD Candidate · Co-lead
Bryn Loftness
PhD Candidate · Co-lead
Kristine Harootunian
Educator, SBSD
Dr. Juniper Lovato
VCSI
Alexa Woodward
VCSI
Outreach, community connections, and prior experience
Where we've been working
We're presenting and developing related work through community and school partnerships across Vermont and beyond. We also view this project as the natural outgrowth of some of our previous projects and affiliations, all of which provided invaluable inspiration.
We're always looking for more school, community, and organizational partners; we welcome diverse opinions and backgrounds! Fill out the interest form →
Acknowledgements:
Project partners
Organizations supporting and hosting this work. All curriculum materials will be open source.
Initiative timeline
Where we are and where we're headed
Fall 2026
Vermont classroom pilots
First VALI modules piloted in Vermont K–12 classrooms. Collecting teacher and student feedback to inform the open-source release.
Summer 2026
Educator and student research studies
Working with teachers and students to understand current AI intuition, identify gaps, and refine the curriculum before classroom pilots. Finalizing prototypes for piloting.
May 2026
Project secures additional funding
"Advancing the VCSI AI Literacy Initiative: Translational Fellowship" selected as an awardee for a VCSI Complexity Award.
Spring 2026
Curriculum development begins
Development of prototype interactive scrollytelling modules, hands-on activities, and educator materials — in collaboration with additional researchers and outreach partners.
February 2026
Initiative launches
Proposal for the Vermont AI Literacy Initiative and VALI curriculum awarded a UVM AI Innovation grant.
More milestones, events, partnerships, and grant updates will be added here as the initiative grows.
Get involved
Connect with VALI
Whether you're a teacher interested in piloting, an administrator thinking about school-wide adoption, a researcher or organization looking to collaborate, or simply want to stay updated — we'd love to hear from you.
- Pilot the curriculum in your classroom or program
- Receive educator materials and training before public release
- Contribute feedback that shapes the final curriculum
- Partner as a community organization, district, or institution
- Collaborate on research or curriculum development
- Stay updated as materials are released
Interest form
Takes about 2 minutes. Responses go directly to our team for future follow up.
The form asks for your name, email, school or organization, role, grade levels you work with, and what kind of involvement you're interested in.
Opens in a new tab. Your info is used only to follow up about this program and generally gauge interest.
Prefer to reach us directly?
aiforclassrooms.vt@gmail.comResources & related work
This section currently under construction!
A growing library of related work, background reading, tools, and our own outputs. We'll be adding to each of these as the initiative develops.
Related work
AI4K12 — Five Big Ideas in AI
AI4K12's Five Big Ideas framework is one of several national efforts shaping K–12 AI education. We'll be linking to parallel projects, proposals, and teams working in this space.
Our work · Scrollytelling
Tokenization — an interactive story
An in-progress scrollytelling story being developed by the VCSI team exploring how LLMs tokenize language — one of the core concepts in the VALI curriculum.
Background reading & related research
Reading list
Background reading and our own work, growing over time. To start: Too poor to science (McClain, 2025); A blind spot for LLMs (2023); Tokens, the oft-overlooked appetizer (2024). More coming soon.
Tools & Demos
Interactive AI tools
Tools and demos we recommend for exploring AI concepts. Coming soon.
Coming Summer / Fall 2026
Curriculum materials
Slide decks, activity guides, Tarot Cards, and scrollytelling modules posted here as they're developed. Sign up to be notified.