Freyja Labs

Professional Development for Bellevue SD

Customized. Hands-on. Built for your teachers and your students.

Bellevue, WA · 19,000 students

Map data © OpenStreetMap contributors

What we've been reading

From outside, what we notice

Bellevue School District sits in the Eastside tech corridor — Microsoft, T-Mobile, Meta within the community's daily orbit — and ranks among Washington's highest-performing districts. Intellectual depth matching community sophistication is the implicit standard. Washington's H-AI-H framework names the pedagogical direction: Human inquiry, AI production, Human reflection. We co-design with Bellevue teachers — custom-built to meet the Eastside tech community's standard for intellectual depth.

If any of this resonates, you might also like

Next ↓ 02 · Sample lessons

Sample lessons — artifacts, not deliverables

We don't deliver lesson plans.

We deliver change for your teachers. The lessons below are the receipts.

These two examples were built for Bellevue SD — but what we'll actually do together depends on what you tell us about where you are and where you want to go. We custom-build with your teachers around what your students, your community, and your district leadership are actually navigating: AI integration, CS/STEM integration, WA policy, the local context only your educators can read. The artifacts you'll see below are one shape that capability can take.

Tech lesson / with devices
micro:bit + AI · Edge-case stress testing

Beyond the Demo: AI Systems That Actually Work

8–10 · Science / Engineering · 90 minutes

Kit: micro:bit + sample lesson plan — "Beyond the Demo: Building AI Systems That Actually Work" (grades 8-10, science/engineering). Students build a micro:bit-based AI system, then stress-test it with edge cases — discovering where AI breaks. Embedded AI literacy: the difference between a working demo and a reliable system. Designed for the Eastside tech community.

What's new — what wouldn't have happened before this PD

Without the PD, AI demos look polished. After: students build a working AI system, then stress-test it with edge cases — and discover the difference between a working demo and a reliable system. Bellevue's engineering culture knows the difference; now students do too.

Show full lesson plan objectives · procedure · materials · assessment · teacher pack ▾

Content Objectives

  • Build a working classification or prediction system
  • Design edge-case tests
  • Identify and document failure modes

AI Literacy Objectives

  • Distinguish a working demo from a reliable system
  • Apply edge-case stress testing to AI
  • Articulate engineering-grade reliability criteria

What Students Do

Phase 1 · 30 min Build

Teams of 3+ build a simple micro:bit AI system that classifies or predicts based on sensor data. Document expected operating range.

Facilitation focus

Don't standardize sensor placement across teams. Different microclimates make Phase 2 richer. Move between teams every 5 minutes; check that students are recording observations *and* numerical readings. The qualitative notes are the wedge they'll use to challenge AI in Phase 3.

Watch for

Teams logging only numbers. Push them to write at least one observation per reading ("breeze picked up", "cloud passed over"). If the campus has visibly varied environments — shade vs. sun, paved vs. planted — push teams to spread out.

Phase 2 · 30 min Stress

Teams design edge-case test scenarios deliberately outside expected range. Run the tests and document failures.

Facilitation focus

Frame the AI tool as a teammate, not an authority. When the AI prediction is wrong, students often default to "we'll fix our data." Interrupt that — the goal is to surface where AI and ground-truth diverge, not to reconcile.

Watch for

Teams that find zero divergence. Either they're smoothing data unconsciously, or the AI is generic enough to match anything. Have them pick a single 5-minute window and compare in extreme detail.

Phase 3 · 30 min Improve

Apply the verification protocol plus an engineering improvement step. Class produces a list of how the team would harden the system — Eastside-tech-grade QA.

Facilitation focus

The class trust guidelines are the deliverable. Push for specificity: not "AI is bad at humidity" but "AI underestimates humidity in conditions like ours when [specific local condition]." Local knowledge + data = the trust criteria.

Watch for

Generic statements ("AI is sometimes wrong"). Reject these gently — every guideline must reference a specific divergence the team observed.

A four-step verification protocol your teachers will build with us

A practice students learn once and apply to any AI output, in any subject, for the rest of their lives.

1. Check the source

Where did the AI get its data? Is it the same data we used or generated?

2. Check the reasoning

How did the AI reach its conclusion? Can we follow the logic?

3. Check against reality

Does the output match what we observed with our own senses, instruments, or knowledge?

4. Check yourself

What might we have missed? What would we want a second opinion on?

More on the thinking behind this — the framework we built it from.

Materials

  • micro:bit with sensors (included in kit)
  • USB cables and student devices
  • Browser-based simple AI system framework
  • Edge-case design worksheet
  • Chart paper and markers

Assessment

Each team produces a one-page artifact: their findings, the AI output they evaluated, and a written verdict on when this kind of AI work is worth trusting.

Hardening list cites at least two specific edge-case failures and the proposed engineering response.

Teacher pack — everything you need to teach this

For the Facilitator

Prior Knowledge Required
  • Read and create simple data tables and bar/line graphs
  • Distinguish between an observation (what we measured) and an inference (what we conclude)
  • Familiarity with one-step variable assignment in block-based or text-based code
Exit Ticket

"Describe one moment today when your direct measurement told you something the AI missed. What did you measure, and what should the AI have done differently?"

Look for
  • Specific reference to a measurement (number + unit + location)
  • Specific reference to what the AI output said
  • A concrete claim about what the AI should have changed (input, comparison, caveat)
Anticipated Misconceptions

"If the AI says it, it must be right — it has access to all the data."

→ Show the AI a deliberately wrong dataset and have students predict the (wrong) output. Reinforce: AI confidence ≠ AI correctness. The AI processes whatever input it receives, including noise and bias.

"Our sensor data is wrong because it doesn't match the AI."

→ Have students re-measure with a second device or different location. Direct measurement is the ground truth — divergence with AI is a signal worth investigating, not an error to "fix."

"The AI is broken if it gives a different answer to the same question twice."

→ This is a feature, not a bug. Use it to discuss probabilistic vs. deterministic systems. Two valid outputs can describe the same data — students should learn to ask "what stayed the same?"

Differentiation
Differentiation 1 / 4
Below grade

Pre-load the data table with column headers and one example row. Pair with a peer for the AI comparison phase. Provide a sentence frame for the evaluate phase: "We trust the AI when ___ because ___. We don't trust it when ___ because ___."

At grade

Standard procedure as written.

Above grade

Add a fourth phase: students design a follow-up experiment that would resolve a specific AI-vs-observation disagreement they found. They write the protocol; they don't need to execute it in class.

Multilingual learners

Provide vocabulary cards in advance: sensor, prediction, observation, evidence, divergence. Allow team discussion in students' home language; final class artifacts can be authored bilingually. The verification protocol works in any language — emphasize that direct measurement and lived knowledge are the most authoritative inputs.

Slide Cues — 6 slides
Slide cues 1 / 6
Slide 1 What we are doing today
  • Collect real data on our campus
  • Compare it with what an AI predicts
  • Decide together when AI is worth trusting
Slide 2 The verification protocol
  • Source · Reasoning · Reality · Yourself
  • You don't need to know how the AI works to check its work
  • A protocol, not a checklist — adapt to the situation
Slide 3 Phase 1 — Collect
  • Teams of 3 deploy sensors at chosen spots
  • Record at intervals; vary your locations
  • Log a written observation with each reading
Slide 4 Phase 2 — Compare
  • Input your data into the AI tool
  • Find at least 3 places it agrees, 3 places it diverges
  • Don't correct your data to match the AI
Slide 5 Phase 3 — Evaluate
  • Apply structured verification practice to your divergences
  • Present one trust guideline based on what you found
  • Class builds the shared trust framework together
Slide 6 Exit ticket
  • One moment when your measurement beat the AI
  • Be specific — number, location, what AI missed
Standards Alignment — 9 frameworks
Standards 1 / 9
NGSS MS-ESS3-3

Apply scientific principles to design monitoring method for human impact

NGSS MS-ETS1-3

Analyze data from tests to determine best design solutions

CCSS Math 6.SP

Statistics & Probability — Grade 6

CCSS Math MP.4

Model with mathematics — Math Practice

CSTA 2-DA-08

Collect data using computational tools; transform for reliability

CSTA 2-DA-09

Refine computational models based on generated data

CSTA 2-IC-21

Discuss bias and accessibility in technology design

ISTE ISTE-3b

Evaluate accuracy, credibility, relevance of information

ISTE ISTE-5b

Collect data and identify data sets; use tools to analyze data

Family / Guardian Letter — copy & paste, edit to fit

Dear families, This week your student is learning a skill that will matter for the rest of their lives: how to decide when to trust an AI system. In this lesson, students used real sensors to measure conditions around our school and compared what they measured with what an AI predicted. The point is not that AI is bad — the point is that AI works best when paired with someone who knows the real situation. Your student is learning to be that someone. We call the protocol the verification protocol. It has four steps: check the source the AI used, check the reasoning, check the result against reality, and check yourself for what you might have missed. You can use this with your student at home — every time an AI assistant gives you an answer, ask: "How would we check this?" Questions? hello@freyjalabs.com — Freyja Labs (working with Bellevue SD)

↑ Top of Tech lesson ↑ Top of Lessons Next: Unplugged Lesson →
Unplugged lesson / no screens
No screens · The edge-case challenge

Break It: The Edge Case Challenge

8–10 · Science / Engineering · 60 minutes

Break It: The Edge Case Challenge — Teams identify scenarios where a well-designed AI system would produce incorrect or harmful outputs. They design test cases to expose the failure. Students discover that finding failures is a design skill. Connected to Bellevue's QA/engineering culture.

What's new — what wouldn't have happened before this PD

Without the PD, finding AI failures is a debugging story. After: students identify scenarios where a well-designed AI system would produce incorrect or harmful outputs and design test cases to expose the failure — the same QA discipline the surrounding tech community runs on.

Show full lesson plan objectives · procedure · materials · assessment · teacher pack ▾

Content Objectives

  • Read system specifications critically
  • Design edge-case test cases
  • Articulate failure-finding as a discipline

AI Literacy Objectives

  • Apply QA thinking to AI systems
  • Identify failure modes systematically
  • Articulate the engineering value of failure discovery

What Students Do — No Screens, No Devices

Phase 1 · 15 min Examine

Teams of 3+ read printed specifications for a well-designed AI system. Note the stated capabilities and assumed operating range.

Facilitation focus

Print the artifact packets in color so detail is preserved. Don't tell students which AI claims are "right" — let them notice divergence on their own. Their lived knowledge of the topic IS the comparison standard. Treat it that way explicitly.

Watch for

Teams that pick a "winning" artifact immediately. Slow them down — every artifact reflects the AI's best guess given its inputs. The question is not which is right but how anyone could have known in advance.

Phase 2 · 25 min Design

Teams identify scenarios that would produce incorrect or harmful outputs and design test cases to expose each failure. Document the reasoning.

Facilitation focus

Distinguish three error types: factual (X is asserted but isn't true), framing (the description emphasizes one thing while ignoring others), absence (something important is left out entirely). Most AI artifacts fail in framing and absence, not facts.

Watch for

Teams that only catch factual errors. Push deeper — what story is the AI telling? Whose perspective is implicit? What did it not have access to?

Phase 3 · 20 min Argue

Each team explains why finding failures is itself a design skill. Class compiles patterns from the QA discipline that the surrounding tech community uses professionally.

Facilitation focus

Frame the argument as advice to a real decision-maker who will act on it. Students must commit to a recommendation AND name specifically what would change their mind.

Watch for

Hedging ("we can't really know"). True — but the decision still has to be made. Push students to commit to a recommendation AND explain what new information would flip it.

Materials

  • Printed AI system specifications (PDF at landing page)
  • Edge-case design worksheet
  • Failure-mode reference cards
  • Chart paper and markers

Assessment

Each team produces a one-page artifact: their findings, the AI output they evaluated, and a written verdict on when this kind of AI work is worth trusting.

Test-case design covers at least three distinct failure modes with the reasoning behind each.

Teacher pack — everything you need to teach this

For the Facilitator

Prior Knowledge Required
  • Read and discuss informational text in small groups
  • Cite evidence to support a claim — written or verbal
  • Familiarity with the difference between a prediction and a confirmed result
Exit Ticket

"An AI tool gives someone you care about a recommendation. What three things should they check before they accept it?"

Look for
  • At least one item references the source or input data the AI used
  • At least one item references the AI's reasoning or comparison with known facts
  • At least one item references checking with a person, lived experience, or independent source
Anticipated Misconceptions

"AI is just like a calculator — if you give it the right numbers, you get the right answer."

→ Use a worked example where two students give the same prompt and get different outputs. AI is more like a human reader making a judgment call than a calculator computing a formula.

"If we can't see the math, we just have to trust it."

→ Pivot the protocol — "Check the reasoning" — to focus on what we CAN check: source, comparison to known facts, internal consistency. You don't need the math to evaluate a claim.

"AI hallucinations only happen with chatbots."

→ Show a printed AI example that contains a confident but factually wrong statement. Hallucinations are a property of how generative models work, not a chatbot quirk.

Differentiation
Differentiation 1 / 4
Below grade

Reduce to two artifacts in the packet (instead of three or four). Provide a graphic organizer with three columns: What the AI said / What I know / What I would tell someone who asked me. Sentence frames provided for the final argument.

At grade

Standard procedure as written.

Above grade

Add a meta-question: students draft two or three rules for when the kind of AI tool in this lesson should and should not be used by adults in their lives. The rules are evidence-cited from the packet.

Multilingual learners

Encourage discussion in students' home language. The artifacts in the packet can be discussed in any language — the analysis transfers. Final class artifacts can be authored bilingually. Pair vocabulary cards if helpful: source, reasoning, reality, evidence.

Slide Cues — 6 slides
Slide cues 1 / 6
Slide 1 Real decisions, real evidence
  • AI tools are showing up in decisions that matter
  • Your job today: evaluate whether the AI's reasoning holds up
  • No screens. Just the artifacts in front of you.
Slide 2 Verification protocol — without the machine
  • Source · Reasoning · Reality · Yourself
  • Same protocol whether you're looking at a sensor reading or a printed AI claim
  • Adapt to what's in front of you
Slide 3 Phase 1 — Examine
  • Read the artifacts carefully
  • Note what the AI got right and wrong
  • Don't pick a winner yet
Slide 4 Phase 2 — Investigate
  • WHY did the AI reach this conclusion?
  • What data did it have? What did it miss?
  • Was it the model, or what the model knew?
Slide 5 Phase 3 — Argue
  • Make a defensible recommendation
  • "It depends" is allowed if followed by "specifically, on these things"
  • Cite evidence from the packet
Slide 6 Exit ticket
  • Three things to check before accepting any AI recommendation
Standards Alignment — 6 frameworks
Standards 1 / 6
CCSS ELA RI.7

Reading Informational Text — Grade 7

CSTA 2-IC-21

Discuss bias and accessibility in technology design

CSTA 2-IC-23

Describe tradeoffs between public and private information

ISTE ISTE-3b

Evaluate accuracy, credibility, relevance of information

ISTE ISTE-7b

Use collaborative tech to examine issues from multiple views

NCSS C3 D2.Civ.14.6-8

Compare historical/contemporary means of changing society

Family / Guardian Letter — copy & paste, edit to fit

Dear families, This week your student practiced something most adults haven't been formally taught: how to evaluate an AI-generated claim before accepting it. In this lesson, students worked from printed artifacts — no screens — and applied a four-part verification protocol: check the source, check the reasoning, check the result against reality, and check yourself. They learned that the right answer to "should I trust this AI?" is almost always "let me check first." At home, you can use the same protocol. The next time an AI assistant gives your family information, ask your student: "What would we need to check before we acted on this?" Questions? hello@freyjalabs.com — Freyja Labs (working with Bellevue SD)

↑ Top of Unplugged lesson ↑ Top of Lessons Next: How we'd work together →

Worth saying again: the lessons above are receipts, not the goal. The point of the engagement is change for your teachers — their confidence to design the next ten lessons themselves, for whatever Bellevue SD faces next. We don't deliver lesson plans. We deliver capability.

More on how we think about this work

Next ↓ 03 · How we'd work together

Engagement Options

How We Can Work Together

We don't sell a packaged curriculum — every engagement is shaped around what your district tells us it needs. The options below are starting shapes; the actual work gets co-designed with your team. Click any that look promising and tell us what you're thinking.

Click any option below to mark it as interesting — then use the form to send a quick note.

Next ↓ 04 · Reach out

We do not provide generic materials. We provide the empowerment and support for teachers to build lessons like these — tailored to their students, grounded in their community's experience.

Mike Borowczak, Ph.D.

Andrea C. Burrows Borowczak, Ed.D.

Where growth begins.

hello@freyjalabs.com