Ai, Experience, and Nature

The Suitcase

A few years ago, an image went viral: a teacher in Africa was illustrating the Microsoft Word interface on a chalkboard to teach technology standards—without a single computer in sight. I could relate.

My teaching career has zigzagged between low-tech, no-tech environments and designing state-of-the-art digital learning spaces. Across the full spectrum of tools and technologies, one lesson continues to resonate: what matters most is not the tool, but the mindful use of it, the theory of learning behind it. 

Learning spaces change. Methodologies change. Content changes. Culture changes. The constant through it all is the need for a theory of learning that can travel—rooted in the ideas of Vygotsky, Activity Theory, Gibson’s affordances, MI Theory, New Literacies, Reggio Emilia, Making Learning Visible, James, Dewey, Montessori, Barthes, Papert, Varela, Barsalau, Freire, Benjamin…

Not long after large language models exploded into the edtech space, I moved to a small community jungle school in Panama—with next to no tech. The problem was wicked clear - What good is Ai out here?

The Inventory

Viewing simulations of Mars habitats helped prepare for Manglares Discovery School—a cluster of modest local constructions and shipping containers from the Panama Canal converted into classrooms. It’s tempting to call this a no-tech environment, but there is internet.

The first challenge wasn’t digital—it was spatial: designing for the human behaviors around learning. I installed a projector and designated a central wall space as our “campfire,” a shared focal point for storytelling, discussion, and collaboration.  I drilled holes in the walls to grid the students’ personal wipe boards making learning visible, he adjacent possible on display. 

Students arrive with the essentials: notebooks, pencils, sketchbooks, and assorted colored pencils. We have a limited set of textbooks for reading instruction, an eclectic library of reference materials, and just enough fiction to support a healthy cycle of self-selected reading.

If you keep a glass-half-full mindset, it’s a school pared down to the essentials—stripped of all unnecessary excess, and dare I say, BS from the capital city corporate international schools where I have spent most of my career. My own hardware toolkit includes a MacBook, iPad, iPhone, and projector, all in constant rotation throughout the day for interactive instruction.

Framing the Problem

On my first visit, the half-empty part of me recalled a book once issued to every Peace Corps volunteer—Where There Is No Doctor, a DIY manual for health care in resource-limited settings. That first year felt much the same: one challenge after another, each ending in the phrase where there is no ____.

• learning support staff
• science lab
• librarian
• grade-level team
• counselor
• resource room

For a teacher used to sprawling international schools, abundant budgets, and a chain of command with a “yes—make it so” attitude, it was a humbling experience.

But then the shift to half-full.

The children’s minds—like the classrooms—were uncluttered. Most had never known any other kind of school. Their implied learning tanks were full, their senses tuned to the rhythms of the natural world. They build, invent play, rope swing, forage, point at monkeys in the trees—and at recess, even in second grade, asked the most unprompted, profound questions:

Are the patterns in nature infinite?

In this space, learning about learning was everywhere. One live hypothesis emerged and stuck:

How might AI connect students more deeply to the natural world?

Could their rich sensory knowledge of their environment be made visible and explicit—honoring Dewey’s view of learning as experience reflected upon? 

Could iterations of sensory experiences become new lenses of inquiry—refining Heisenberg’s insight: “We have to remember that what we observe is not nature herself, but nature exposed to our method of questioning”?

How might inquiries—rooted in student observation—become the raw data for co-creating curriculum?

The danger of the live hypothesis is that it begets more questions than there is the time or resource to answer.  It is a good problem to have.

Empathy and Understanding 

Luckily, I had the students as guides.  They taught me which fuzzy, multicolored poisonous caterpillars to avoid. They stopped me from stepping near scorpions and tarantulas, and delivered a dead-eyed stare as they explained just how dangerous the X-snake could be. (Adults, especially those like me, new to the region, are in awe of the biodiversity and often share it out on the community chat)

Froebel brought gifts to kindergartners as tools for inquiry. Here, the natural world offered its own gifts—a constant, daily broadcast across multiple sensory channels—and the students bring them: snails from the waterfall, tadpoles, tree frogs, and bugs. 

“We are rich with bugs,” one student declared—the same one who later helped organize a funeral for a dead worm.  (These wildlings have other curious rituals, like burying dried-out whiteboard markers in the field with full ceremony and eulogy.)

So the first couple of months passed in this total dichotomy: inside, we built literacy routines and practiced math; outside, the students kept filling their sensory tanks with tacit knowledge.

Until the katydid appeared on the rim of Mrs. Aya’s coffee thermos.

Classrooms Without Walls

For decades, the edu-corporate complex has maintained tight control over curriculum and content. I grew up in Galveston, an island where major state universities run branches in marine biology. They study all the cool stuff right there in the water, implicitly - porpoises surfing waves, cuddle fish and octopus tanks, turtle breeding labs, and seining for blowfish in the bay.  Environmental organizations have established bird sanctuaries. 

But for elementary school curriculum, instead of studying the stingrays, crabs, and jellyfish they came into contact with at the beach—or the local redfish, flounder, and shrimp industries—schoolkids got state-issued textbooks void of any embodied experience. The water was all around us, but some kids didn’t know how to swim. 

In Panama, by contrast, part of the mission of a micro-school nestled in a riparian forest is to do the opposite: to connect the implicit sensory world to explicit learning experiences.

In other words:

How does a katydid crawling across Mrs. Aya’s coffee mug become the launchpad for a unit of study?

The Bug on the Mug

Moving sensory experience into learning begins with framing—with the lenses we offer students.  From my own hikes near Gamboa, the sensory experience disorients, the rainforest overloads cognition.  I started developing themes to the hikes, exploring different parts of systems, and I immediately started seeing nuances of the rainforest.  

Just like Reggio Emilia teachers do, I started by noticing what the students are noticing, and documenting by filming short video clips. You can hear the students huddled around the katydid on Mrs. Aya’s mug, one voice infecting the next with curiosity. They noticed its long, oscillating antennae. One observed that it might be attracted to the heat of the coffee in the mug.

Once something is framed, everything they say becomes critical.

The next day, another katydid appeared—a juvenile, also with massive antennae and leafy body. The students immediately began asking whether it was a grasshopper or something else. Their observational reasoning was still informal, and I thumbed it into my notes, letting NGSS serve as frame. 

In class, we projected close-up images of both katydids onto the wall. I introduced the See–Think–Wonder thinking routine to help them think about their thinking, to just focus on observing, trying to ignore that chatty interpretive, know-it-all part of the mind. We took time to observe, describe, and analyze every detail. As they spoke, I documented everything—either by typing as they dictated, writing it by hand, or recording the discussion to transcribe later using the Whisper app.

But the wondering—that was the most important. We gave it space. We let the questions emerge organically from silence, and we sat with them for a full ten to fifteen minutes—an eternity in second-grade time.  Rich inquiry often develops more slowly by creating atmosphere, allowing particles in the mind to charge, for the lightning of a live hypotheses to strike. 

Sensory becomes inquiry becomes science.

Leveraging the Tools

While the students observed and commented during two recess sessions, I worked quietly in the background—leveraging AI and Google Scholar. ChatGPT helped identify the species of katydid. In later inquiries, I began triangulating identifications using ChatGPT, iNaturalist (Ai with human verification), and Google Reverse Image Search.

Faced with the usual riddle—what do you do when there’s no second-grade-appropriate text on katydids in Panama?—my glass-half-empty instinct kicked in. The default urge was to move up the taxonomy ladder, to use the generic insect books in our library and work down from there.

But then I remembered: we’re rich with bugs.

And we’re in Panama.

And Panama, like Galveston, hosts deep research, Smithsonian Tropical Research Institute and countless field research projects conducted by institutions from around the world.

So I searched Google Scholar—just to see.

Dozens of research articles came up, on our specific species of katydid.

Knowledge, like food, is not evenly distributed.

The emergence of large language models over the past two years is starting to solve that problem.  All second graders should know their immediate natural world with all that humanity has discovered, just like famine in Sudan or anywhere in the world is a gross absurdity.

As I reviewed the research, I realized their questions—about camouflage, communication, anatomy, and environment—aligned beautifully with Next Generation Science Standards and Common Core nonfiction literacy goals.  It always works this way, where provocation is rich, students will, by the law of Kundera’s “many people few ideas”, engage in the same logical thinking used by scientists and designers of learning standards.

I pulled ten research articles and scanned the research and results sections. There was more than enough data to work with. I downloaded the PDFs and uploaded them into a custom GPT within ChatGPT.

Then I gave this GPT a personality—an anthropomorphized katydid who could talk to the class. I prompted it to adjust its tone and vocabulary for second graders, preserving scientific terminology but adding playful narration and lots of emojis.

After running it through some of the students’ original questions, it worked beautifully.

We had built something new:

A kind of interactive Socratic textbook—a very conversational, and curious katydid ready to co-create text from student inquiry.

Establishing Cycles

In a typical nonfiction reading session, we’d begin by exploring text features—giving students space to connect to prior knowledge, raise questions, and build curiosity.

Here, we took a different route.  We started with our own data: Images captured during recess, notes from our class discussions, and lists of student-generated observations.

In small groups, students worked to identify the most critical questions to begin with. I kept the experience text-based at first—many parents and teachers were still wary of the AI critics that had dominated the 2023 discussion of imminent singularity, mind control, 

Even so, it was an immediate hit.  Just seeing their own questions turned into readable text, to know there are answers to all their questions, was thrilling.

Once we had generated enough content, we printed copies and used them as nonfiction reading material. As they read, students didn’t just answer comprehension questions. Instead, we used the Connect–Extend–Challenge thinking routine to help them:

• Connect new ideas to what they already knew

• Extend their understanding

• Challenge assumptions or open new lines of questioning

Critically, they used this reflection to curate inquiry—developing new questions to bring back to “Katy the Katydid” in future sessions.

Over the course of several days, a learning cycle emerged:

1. Curate inquiry

2. Hold a live discussion with Katy

3. Read newly generated content

4. Discuss, write, and observe

5. Generate further inquiry

What began with one insect became a recursive system of thinking, observing, questioning, and co-creating knowledge—rooted in their own experience, amplified by Ai.

Open Ceilings

The results far exceeded our initial learning goals.

Echoing the early 2000s lessons of Pokémon—where each creature revealed an intricate layered world of biology, evolution, and imagination—the katydid’s antennae opened a portal to deeper understanding. We explored sensilla, the tiny sensory organs that detect temperature shifts, environmental pheromones, and vibrational nuances far beyond human capacity.

Katy, having senses beyond human capabilities,  became a kind of superhero—a playful hybrid of reality and fantasy.  But the breakthrough wasn’t just in declarative knowledge.  It was in how students began to frame their world through sensory observation becoming more agentic with purpose and hybrid audience.

Soon after, a student discovered a bromeliad spider.  The green iguana, part of the riparian ecosystem, and our class mascot, became part of the study.  Another insisted we study the axolotl.

Each time, we followed the same model: sensory observation, build a custom GPT and craft interactive texts anchored in scientific research and student questions.  

And again, they noticed the smallest things:

• The drumming patterns a banana spider makes on a leaf to attract a mate

• The parietal eye of an iguana, functioning like a light sensor

• The axolotl’s extraordinary regenerative ability—and its vulnerability to pollution

With each iteration they became more attuned to the sensory specializations, a concept that interestingly came back when they studied celestial bodies by tracking the routes of different space probes, and all the cool gadgets and sensors each was equipped with.  

Lessons from LLMs

Large language models (LLMs) present a uniquely wicked problem for the edtech world. On the surface, they’re deceptively simple—chatbots anyone can use. But underneath, they are powerful sandboxes for creating entirely new tools and learning systems.

For now, I’ll defer the larger discussion of how such tools might be unpacked, the infinite ways they can be leveraged. The focus here is framed on the katydid.

This began in the 2023–2024 school year. Since then, the affordances of LLMs have continued to grow exponentially. But even in the first months of interacting with LLMs, Katy the Katydid taught six lessons about how LLMs are becoming the ultimate open-source learning toolkit.  

Six Lessons from Katy the Katydid

1. Bridging Sensory and Academic Experience

ChatGPT bridged the episodic sensory experience—students gathering around the katydid on the coffee mug—and the explicit learning experience through shared video reflections, classroom discussions, and evolving documentation.

2. Filling the “Where There Is No Text” Gap

ChatGPT responded to the challenge: how do we study this species of katydid when there’s no second-grade text?  It synthesized ten scientific studies into leveled, accessible nonfiction that retained scientific integrity.

3. Co-Creating in Socratic Dialogue

ChatGPT became a co-author, answering student questions in a living, evolving dialogue—an emergent textbook shaped by inquiry rather than pre-set chapters.

4. Developing a Transferable Ritual of Framing

A new ritual took hold: students began framing their everyday sensory observations with the expectation that these could lead to explicit constructs—scientific terms, systems, patterns.

5. Driving Interdisciplinary, Student-Centered Curriculum

Inquiry became the spine of a co-created curriculum that organically bridged Reading, Writing, and Science standards—without ever feeling like “coverage.”

6. Scaffolding for All Learners

ChatGPT created customized supports for vocabulary acquisition—short plays, character dialogues, and even raps—all grounded in students’ own questions.  A year later they still know many of the raps.

Ai Parameters 

Nonfiction writing and visual representation are forms of art—and students do need to learn how to engage with nonfiction text. AI of 2025 doesn’t replace that.

We still studied bears around the world from our textbook.

We still practiced decoding nonfiction features, learned new vocabulary, and navigated multimodal texts. That traditional skillset—reading captions, analyzing diagrams, skimming headings—is still essential.

While ChatGPT has improved at generating images from text, it remains unreliable for anything approaching scientifically accurate illustrations—especially of insects. We tried. The results sparked useful conversations about fantasy vs. reality in AI-generated visuals. In the end, the images we took of our own katydids—and those found on iNaturalist from our region of Panama—were far more meaningful.

All our AI interactions took place as whole group “campfire” events, projected on the wall. This was intentional. For young children, talking to a chatbot can be disorienting. One student, for example, was visibly let down when she realized we weren’t actually talking to a polar bear—but to a computer simulation based on the text we had read.

Just like many adults unsure where AI is operating inside their tech interfaces, onboarding second graders took time.  Know your community and let them know what you are doing with Ai.  That year I kept a running blog of our progress and it really helped demystify a lot of the hype.  

And then—there’s the question of ethics.

LLMs are, in many ways, high-tech plagiarism. The ethics of classroom use remain murky. We anchored our work in open source scientific articles and relied on our own photography and video. But from my earliest interactions with ChatGPT—back in January 2023, while analyzing children’s literature through Barthes’ five codes of communication—I was struck by the insight that emerged from just a few pages of text, whether copyrighted or not.

There’s no shortage of negative press surrounding LLMs.  But from an educator’s standpoint—especially one who values open-source tools and is constantly solving “where there is no _____” challenges—LLMs hold real promise.  LLMs do not have to mean outsourcing the heavy cognitive processing.  If used correctly they create cognitive power processing integrating the sensory observation of a whole class of students. The sheer volume of niche research on katydids would have been inaccessible for both teachers and student.  Interdisciplinary work can be daunting given the web of multiple sets of standards.  Like recent research where Ai helps Anthropic map thousands of emails to find out how much cognitive load and what level of Blooms is being offloaded by students and teachers, or how Facebook data is giving us a more complete picture of global migrations, Ai is extending zone of proximal development to volumes and lexile levels beyond our normal fluency and comprehension.  What it has not done is reduced our cognitive challenges. 

Ai in Local Culture

Katy the Katydid was our first full iteration of AI constructed through student inquiry. I tried to keep the variables minimal. But the possibilities are endless: image recognition tools with agency, voice-based dialogue between students and AI, real-time feedback on handwritten notebook work.

Most of these students aren’t “screen kids”—they’re not drawn to tech the way many of my former urban students were. All AI interaction here is shared: projected, whole-class, and conversational. But since then I have encouraged a lot more independent use of Chatgpt, particularly with middle schoolers.  

Years ago, I facilitated global classroom exchanges between students in Colombia, Canada, and Australia. They studied many of the same ecological concepts. The next step seems obvious: design GPTs based on local species from our riparian forest and exchange them with classrooms around the world.

This would be the natural evolution—from mailing packages across continents, to video conferencing, to… live classroom conversations with an AI katydid as co-host.

When there are tech surges in education it can feel threatening with all the hype and the corporate urgency that this has to happen now.  I would suggest go no-tech for a while, work on the theory of learning, really think through what tools are mindfully aligned with that theory, and proceed from there, blocking out the hype.  The Idea of Progress coupled with the corporate need for constant growth to satisfy stock holders, this is not part of the educator’s somewhat anti-capitalist agenda.  

The nice thing about this peculiar sandbox is you can tell it your theory of learning and it will work with you at your readiness and level.  As China pushes Ai into all learning spaces and the US is looking at a new draft of an executive order to integrate Ai into classrooms, now is a great time to reflect on how this mindful integration works without sacrificing experiential learning, multimodality, and cooperative learning goals. 

Reach Out

This is just one example of how we’re exploring AI in education at Manglares Discovery School.  We have many other projects with student inquiry in motion—and we’re always looking for collaborators.

If you’re interested in learning more or contributing to any of the following, feel free to reach out:

 • Differentiated Reading Interventions

Using Words Their Way patterns to generate raps, skits, and leveled support materials tailored to content areas.

 • Student Inquiry Based, Co-Created Curriculum

Interdisciplinary projects grown directly from student inquiry. 

 • Middle School Word Work Projects

Students building power glossaries, quizzes, riddles, and escape rooms from their independent reading.

 • Higher-Order Literary Thinking Prompts

Aligned with Common Core, leveled for depth and flexibility on any text.

 • Socratic Discussions & Audio Transcription Loops

Student-led book talks where recorded discussions are transcribed, refined, and revised collaboratively.

 • iNaturalist Ecosystem Project

Students approach a study of ecosystems beginning with river walks, sketching illustrations, photographing, and working with Ai to review their inquiries and propose curricular connections.

 • Ethnographic Assessment

Alternative assessment through ethnographic documentation from audio, image, and video.