A locally developed elective, a teacher fellowship, and a curriculum framework we would share with districts across British Columbia.
Our students use AI systems daily and have almost no basis for deciding when to believe them. B.C. has published careful guidance on using AI tools safely in schools, and provincial and district teams have built good resources on responsible use. We have not identified a course whose primary learning outcomes address how these systems are built, why they fail, and how a student should reason about their output.
I am asking the Board to fund an eighteen-month pilot: $148,000 to fund a part-time project lead, train and support nine teacher fellows who would write and teach the course, develop a locally authorized Grade 10–12 elective, and run it in three secondary schools with independent evaluation.
This proposal is written by one person. The team it describes is what the first phase of funding would assemble.
I have spent the last several years working closely with secondary students — at school, in tutoring, and in residential settings where I see how they actually work in the evening rather than how they present in a lesson. Since 2023 the same pattern has appeared in all of it. Students are fluent operators of AI systems and have very little basis for evaluating what those systems give them. They can produce an essay, an image, or a study guide in seconds. Asked whether a particular output should be believed, most reach for surface cues: whether it sounds confident, whether it is detailed, whether it agrees with what they expected.
Two things about that vantage point are worth stating, because they shaped this proposal. I see students across every subject rather than in one, so I can see that this is not a computer studies problem — it shows up in English essays, lab write-ups and history assignments alike. And I see them using these tools unsupervised, outside class, where the tool is not framed by a teacher and no one is watching how they decide what to believe. That is where the judgment is actually being formed, and it is currently being formed without any instruction at all.
The wider evidence points the same way. A 2025 KPMG survey found that 73% of Canadian post-secondary students had used generative AI for schoolwork, and 48% believed their own critical thinking had suffered as a result.3 Those students were in secondary classrooms two or three years ago.
I want to be careful about the diagnosis here, because it determines what we should do about it.
The difficulty is not that students use these tools. It is that they have no working model of how the tools produce what they produce, so they have no principled way to decide when to trust the result.
That is a teachable problem, and it is the outcome this proposal is organised around.
The most recent national picture comes from a survey of 756 grades 6–12 teachers across Canada, conducted between November 2025 and January 2026.1
have not been provided with training or tools to identify when students are using AI in their work1
report that their school or board has given them no guidance on how to instruct students in appropriate AI use1
say their school has no AI policy at all, for either staff or students1
Training rates vary widely by region, and British Columbia was not among the provinces identified as leading.
The same survey found independent school teachers ahead of their public school colleagues on nearly every measure, including 46.6% versus 29.2% on training to instruct students in reasonable AI use.1 That gap should concern a public board.
Teachers have been asking for this support for some time. An Alberta Teachers' Association survey found roughly 60% of members did not believe their schools were adequately prepared to use AI effectively, and the members most positive about the technology were the ones most actively requesting professional development.2 CBC's 2025 reporting found the same appetite.4
A systematic review of 43 studies on K–12 machine learning education identified a shortage of resources below high school, minimal integration outside computing courses, and an explicit shortage of teacher professional development.14 A curriculum trial with 70 middle-school pupils and 60 trainee teachers found its critical-thinking components did not land, attributing this to the complexity of the concepts and to the absence of teacher training.15 Any credible proposal here has to fund teacher preparation first, which is why roughly half of this budget does exactly that.
The Ministry of Education and Child Care has done substantial work here, and this proposal builds on it rather than criticising it.
The Ministry's framework for AI tools in B.C. K–12 schools sets out seven categories for local decision-making: ethical uses, needs and impacts, accessibility and usability, integration and compatibility, data security and privacy, teaching and learning, and inclusive learning.5 Districts have layered their own generative AI guidelines on top of it,7 Focused Education has produced 21 lessons developed with B.C. teachers,8 and the Ministry has published materials for families.6
All of this addresses how schools should adopt AI tools responsibly. It was the right first priority. It does not address what a student should understand about how these systems produce their output.
The curriculum foundations are partly in place. B.C.'s Digital Literacy Framework is built on the ISTE NETS•S standards,9 which predate the current generation of these systems. ADST Computer Studies 10 covers computational thinking, programming constructs, and ethical considerations including big data use and equality of access.10 Our review of published curriculum did not identify a Ministry-developed course whose primary learning standards address machine learning or the evaluation of AI output. We would confirm this against the provincial course registry as a first step of the development work, and we would revise the proposal if we are wrong.
These two terms are often used interchangeably and should not be. Getting the relationship right determines what we teach and in what order.
AI literacy is the goal. We take it to mean a student's capacity to recognise where AI systems are operating, to reason about how they are likely behaving, to judge when their output should and should not be relied upon, and to hold a view about where their use is appropriate. The UNESCO frameworks describe this across AI concepts, AI ethics, underlying technologies and applications, and AI system design, at the levels of understanding, applying and creating.13 The draft Canadian AI Literacy Framework covers similar ground for elementary and secondary settings.19
Machine learning is the foundation we teach it through. Nearly every AI system a student will encounter — recommendation, search ranking, image generation, chatbots, automated assessment, hiring and admissions screening — is a statistical model fitted to a body of training data. That shared architecture is why machine learning is the productive entry point:
The reverse is not true: a student can understand machine learning quite well and still have poor judgment about AI systems in the world. So machine learning is the means here, and the outcome we assess is the judgment.
By the end of the course, a student should be able to encounter an unfamiliar AI system, form a reasoned account of how it likely works and what data it was likely built from, identify the conditions under which it would probably fail, and state with justification how much confidence its output deserves.
We call this epistemic calibration, and Section 12 sets out how it would be measured. It is a higher bar than conceptual recall, and it is the reason the course culminates in generative AI rather than beginning there.
Six units, 12–15 instructional hours each. The progression moves from simple models a student can build and inspect, through the ways models fail, to the generative systems students already use daily.
| Unit | Learning objectives | Principal activity | Assessment | AI literacy connection |
|---|---|---|---|---|
| 01 Data and representation | Explain what a dataset is; identify features, labels, sampling and collection method; recognise what a dataset omits | Students collect a small dataset in the school and document its collection method | Dataset audit: what this data can and cannot support | Systems inherit the limits of their data |
| 02 Classification | Build and explain a decision tree and a k-nearest-neighbour classifier; describe a decision boundary; distinguish accuracy from correctness | Hand-drawn trees, then Activity 2.4 below | Build a classifier and explain its errors in writing | Output is a model's best guess, not a fact |
| 03 How models fail | Describe systematic label error, unrepresentative training data and distributional bias; predict failure modes from a data description | Controlled data corruption: students introduce known faults and document effects | Failure report on their own model | Failures are structural and predictable |
| 04 Prediction and uncertainty | Distinguish classification from prediction; interpret confidence scores; explain why a confident model can be wrong | Build a simple predictor and quantify its uncertainty | Prediction task with a written uncertainty statement | Confidence is not accuracy |
| 05 Generation | Distinguish classification and prediction from generation; describe next-token probabilistic generation in plain language; explain why hallucination is expected behaviour rather than malfunction; describe what an AI agent adds | Structured investigation of a language model's failures on verifiable questions | Hallucination investigation with documented evidence | Generative systems are the ones students actually use |
| 06 Judgment | Evaluate an unfamiliar AI system; reason about appropriate deployment; discuss consent, data ownership and governance | Capstone: students evaluate a system chosen from a set they have not studied | Capstone evaluation and defence | The transfer task the course is built toward |
Students work the algorithm on paper before any code or vocabulary is introduced.
Students train a working model on data they collected themselves.
Students introduce known faults into their own data and document what changes.
Students apply the same reasoning to a system they did not build and cannot inspect.
Place training points by clicking, and the model redraws its decision boundary. The corruption control introduces systematic label errors drawn from one region of the input space, which is what happens when training data comes from a single unrepresentative source.
The design draws on published work with students of comparable age. Decision tree and k-nearest-neighbour materials have been tested with 12–13 year olds, who applied the algorithms to real classification problems.11 A design-based study had 34 middle-school students co-design and build working machine learning applications.12 Both suggest the build-and-inspect approach is developmentally sound well below the grade band we are proposing to pilot.
The pilot is for Grades 10–12. An earlier version of this proposal said Grades 8–12, and that was a mistake worth correcting explicitly.
The senior grades are the right starting point for three reasons. Units 4 and 5 depend on probability and proportional reasoning that sit more comfortably after Foundations of Mathematics 10. Students in Grades 10–12 are making post-secondary and career decisions in which this material is immediately relevant. And a Grade 10–12 elective fits the credit structure without displacing required coursework.
Grades 8–9 remain in the longer-term vision, but as a distinct pathway rather than the same course delivered to younger students. Once the senior course is evaluated, we would develop a Foundations sequence of shorter modules — dataset thinking, classification by hand, and what a model can and cannot tell you — omitting the probability-dependent material and the generative AI unit in favour of concrete, unplugged work. That development would be a separate proposal with its own costing, informed by what the pilot teaches us.
Applied AI Literacy, Grades 10–12. The six-unit elective described in Section 5. Piloted in this proposal.
Foundations, Grades 8–9. Shorter adapted modules with no probability prerequisite, developed after evaluation and proposed separately.
Three things are described in this document and it is worth separating them clearly, because they have different timelines and different approval requirements.
GLASSBOX is the initiative. It comprises the curriculum framework, the teacher fellowship, the evaluation, and the eventual dissemination to other districts. It is what the Board would be funding.
The pilot elective is one course within it. A locally developed Grade 10–12 elective, taken through our district's course authorization process, taught in three schools in the pilot year.
Cross-curricular modules are a later expansion. The eventual aim is that students meet some of this material in science, mathematics, social studies and English rather than only in an elective, for the equity reasons in Section 15. That is not something the pilot delivers, and I do not want to claim otherwise. The realistic sequence is:
This ordering matters. The research literature is clear that integrating this material outside computing is difficult and under-supported,14 and attempting it before we know what works in a dedicated course would risk both.
This course has students collecting data and using AI services, which raises obligations the Board should see addressed before approving it rather than after.
The Ministry's considerations framework already identifies data security and privacy as a required category for local decision-making on AI tools.5 The following would be written into the course documentation and the fellowship training, and reviewed with district privacy staff before the pilot begins.
No personally identifiable student information is submitted to external AI or machine learning services. This is the governing rule and it constrains activity design throughout the course.
Student-collected datasets use non-identifying subjects: objects, plants, sounds, publicly published text, sensor readings. Where an activity would benefit from images of people, we use published datasets licensed for educational use rather than photographs of students.
Alongside that rule, the course operates on data minimisation — collecting the least data that supports the learning objective — with datasets stored on district-managed systems, retained only for the duration of the course, and deleted at its conclusion unless a student asks to keep their own work. Students are told at the outset what is being collected, why, where it will live, and when it will be deleted, and consent processes follow existing district practice for classroom work.
Two points are worth drawing out. Students own the datasets they create, and their work is not repurposed as programme material without their permission. And the governance questions themselves are curriculum: consent, ownership, minimisation and retention are assessed content in Unit 6, taught partly through the decisions students can see the school making about their own data.
B.C. provides a mechanism for exactly this kind of locally developed course, and it is worth being precise about what it does and does not involve.
Board/Authority Authorized courses exist so that boards can respond to local needs while giving students choice and flexibility, and are authorized by boards according to Ministry requirements.20 There is no limit on how many BAA courses may count toward the 28 elective credits required for graduation.21 Ministry requirements include alignment to the Know–Do–Understand curriculum model.22
Who does what
Developed by a teacher or team of teachers, working to the Ministry's framework requirements. In this proposal, the nine teacher fellows are the authors of record; the project lead coordinates the work rather than writing the learning standards.
Reviewed by school administration and district staff, against district process and Ministry requirements.
Authorized by the Board of Education. The Board's approval is what permits the course to be offered for credit in our schools. This is district authorization, not provincial curriculum approval, and the course does not become part of the provincial curriculum.
Submitted to the Ministry for course code registration according to Ministry requirements, so that the credit appears correctly on student transcripts.
Offered from the school year following authorization, once timetabling and staffing are in place.
The Ministry maintains a searchable BAA sharing platform with Focused Education Resource Services, where boards can volunteer their course frameworks for other districts to consider and adopt with their own board's approval.20 We would publish the framework there at no cost. It is developed with public funds and should be available to other public districts, and doing so is the most direct route to provincial impact.
On timing more broadly: Canadian curriculum authority sits with the provinces, so no federal mandate will force this. As of December 2025, three provincial ministries had published guidance on AI in K–12 education.16 Canada's AI Taskforce has called for a funded National AI Literacy Strategy to support provincial ministries,17 and researchers have made the same case publicly.18 The draft Canadian AI Literacy Framework is open for contribution now and is explicitly described as a living document.19 A district with a course running and evaluation data in hand would be positioned to contribute to that work.
Nine teachers, one academic year, with paid time and sustained support. This is the part of the budget I would defend hardest, because the evidence in Section 2 says it is where these initiatives fail.
The teacher fellows are the course authors of record. That is partly a Ministry requirement — course development sits with teachers20 — and mostly a design choice, for the reasons set out above. My role is the coordination the project needs and no one currently has time for: research and framework alignment, the authorization paperwork, the privacy review, technology procurement, managing the evaluation contract, and dissemination afterwards. That work is substantial, and it is the part that quietly does not happen when a good idea is added to a teacher's existing workload.
A note on scale: this proposal is written by one person, and references to what "we" would do describe the initiative once the fellowship exists rather than a group that currently meets. Recruiting that group is the first thing the funding buys and the first decision gate in Section 13. If nine teachers cannot be recruited, Section 15 sets out what happens instead.
Approximately 85 hours beyond contracted time across eighteen months: 30 hours of summer institute, 25 hours of curriculum authoring and revision, 18 hours of cohort sessions and coaching outside instructional time, and 12 hours of evaluation participation including pre- and post-surveys, a classroom observation debrief, and an exit interview. At $3,500 this is roughly $41 per hour, which is below the district's curriculum-development rate and is offered as recognition rather than full compensation. Fellows are told this plainly during recruitment.
The six release days exist because the alternative is asking teachers to absorb curriculum development into evenings, which is how this work usually gets abandoned in February.
Nine is three teachers from each of the three pilot schools, which gives each site a working group rather than an isolated individual — the arrangement most likely to survive one person's departure. Recruitment is voluntary and open, with selection aiming for at least four fellows from outside computer studies, drawn from mathematics, science, social studies and English.
No prior programming experience is required, and this is deliberate rather than a concession. A course that only computer studies teachers can deliver cannot later be embedded across subjects, and a fellow who has recently learned this material remembers what was confusing about it. Fellows do need comfort with being a learner in front of colleagues, and willingness to teach at least three units in the pilot year.
Canadian research on pre-service teacher education found that even a single module on AI in education measurably shifted teacher candidates' willingness to engage and their sense of the material's importance.24 A year of sustained support should do considerably more than a module.
Co-design research with K–12 educators found teachers consistently asking for support that develops their professional practice and augments their role.25 Beyond that principle, co-authorship does four concrete things: it surfaces classroom problems during writing rather than during delivery; it produces material that non-specialists can actually teach, because non-specialists wrote it; it distributes knowledge across nine people so the initiative survives any one departure; and it gives fellows a professional stake that materially improves the odds they are still teaching the course in year three. Work with secondary computing teachers already using these tools produced a six-phase integration sequence — prepare, adapt curriculum, lead, assess, synthesize, share — close to the structure the fellowship year follows.26
$148,000 over eighteen months, in Canadian dollars. This is a one-time development and pilot investment, not an annual operating cost. Recurring costs after development are set out below and are substantially smaller.
This line assumes schools supply student laptops from existing stock; no general computing devices are purchased. Implementation research on K–12 AI courses identifies immature market provision of data-collection tools and model-training platforms matched to students' cognitive level, and of affordable classroom compute,27 which is why this is budgeted rather than assumed available.
The distinction matters for how the Board reads the per-student figure. Development happens once; delivery recurs. Move the control to see how the one-time investment spreads as more schools adopt the course.
An earlier draft of this proposal projected revenue from partner districts and claimed the initiative would cover its own operating costs by year three. That framing was wrong for a public school initiative and I have removed it. We are not proposing to run a business.
The return to the district is that development costs are not repeated: the framework is reusable, the fellowship materials can train subsequent cohorts, and other districts adopting the course through the provincial platform reduce duplicated effort across the system rather than generating income for us. If, after evaluation, other districts ask us to deliver fellowship training for their staff, we would bring a cost-recovery proposal to the Board with a full costing at that time. No figure is being claimed here.
A B.C. faculty of education partner would be contracted to evaluate three things: whether students learn what we claim, whether teachers can deliver it independently, and whether the course is implementable at reasonable cost. The third is the one initiatives like this usually skip.
| Domain | Outcome | Measurement | Success threshold |
|---|---|---|---|
| Student | Conceptual knowledge of machine learning | Pre- and post-course assessment, common instrument across sites | Mean gain of at least 40% over baseline |
| Reasoning about model error and dataset limitations | Performance task: diagnose a supplied faulty model | At least 70% of students reach "proficient" on the rubric | |
| Epistemic calibration on unfamiliar systems | Scenario assessment using AI systems not studied in class | At least 60% reach "proficient"; this is the transfer measure and the hardest bar | |
| Completion and engagement | Enrolment and completion records | At least 85% course completion | |
| Teacher | Confidence and self-efficacy | Pre- and post-fellowship survey, validated instrument | Significant positive shift across the cohort |
| Instructional competence | Two classroom observations per fellow against a shared rubric | At least 7 of 9 fellows rated competent or above by term two | |
| Independent delivery | Records of units taught without project-lead co-teaching | At least 7 of 9 fellows independently delivering 4 or more units | |
| Equity | Who enrols and who succeeds | Enrolment and completion analysis by prior computing coursework, gender, and school | Enrolment within 10 percentage points of school population on measured characteristics |
| Access barriers | Student and teacher interviews on device access, prerequisites, timetabling | Reported and addressed; no numeric threshold | |
| Implementation | Which units are actually used | Teacher logs of units taught, adapted or skipped | Reported; informs which units are candidates for cross-curricular use |
| Preparation burden | Teacher time diaries, sampled across the year | Median preparation time within 25% of a comparable elective | |
| Technical reliability | Incident log of hardware, software and network failures | Reported with cost implications for scaling |
Epistemic calibration is the measure that matters most. A student can score well on conceptual recall and still trust a chatbot for the wrong reasons. The scenario assessment presents AI systems the class has not studied and asks students to reason about likely training data, plausible failure modes, and how much confidence the output deserves. We expect this to be the weakest result in year one, and we would rather find that out than avoid measuring it.
Implementation evaluation protects the Board's later decisions. Knowing which units teachers skipped, how long preparation actually took, and what technology failed is what makes a scaling decision informed rather than hopeful. It is also what tells us which units are realistic candidates for the cross-curricular phase in Section 7.
Equity analysis carries privacy obligations. Demographic analysis uses the minimum data needed, is reported only in aggregate with small cells suppressed, and follows district practice and the privacy rules in Section 8. Where a characteristic cannot be analysed without identifying individuals in a small cohort, it is not analysed.
This is a proposed schedule, not a provincial requirement, and it would need to be matched against the receiving district's own internal calendar. The binding constraint is that authorization has to be secured with enough lead time for timetabling and staffing ahead of a September start.
A course addressing a gap we have documented. Students graduate able to reason about the systems they will meet in post-secondary study and work, assessed against thresholds stated in advance rather than asserted afterwards.
Development work that is not repeated. The framework is reusable, the fellowship materials train subsequent cohorts at a fraction of the first cost, and evaluation evidence informs both the Grades 8–9 pathway and the cross-curricular phase without starting from scratch.
A contribution to the wider system. Publishing on the provincial platform20 means other districts do not each fund this work separately, and it gives our teachers standing in the national framework conversation while that conversation is still open.19
Nine teachers with a reason to stay. Fellows finish with authorship credit on an authorized course and expertise very few teachers in B.C. currently hold. Retention of experienced mid-career teachers is a standing problem, and meaningful professional advancement is one of the few reliable levers available to us.
Release time is genuinely difficult to find, and B.C. staffing is tight.
Stipends and release-day coverage are budgeted rather than assumed, participation is voluntary, and the cohort is capped at nine. If we cannot recruit nine, we run with six and extend the timeline rather than press people into it.
Elective courses in this area tend to attract students already advantaged in STEM, and the national data already shows independent schools ahead of public ones on AI training.1
No programming prerequisite, deliberate recruitment of fellows from outside computer studies, and enrolment demographics as a reported evaluation metric with a stated threshold. The cross-curricular phase described in Section 7 is the structural answer, and the pilot is designed to produce the evidence needed to get there.
The tools will change faster than any course can be revised.
Units teach concepts rather than products: training data, classification, error, uncertainty, generation, and the limits of what a model can be asked to do. A decision tree drawn on paper is unaffected by the next model release. Unit 5 is the most exposed and is written to be revised annually, with that revision costed as recurring rather than development work.
Course authorization depends on district process and Board scheduling.
Internal deadlines would need to be confirmed with the district's curriculum office in month one, and the schedule built backwards from them. If authorization slips a cycle, the fellowship and curriculum development still proceed and the pilot moves a year. The work is not wasted, but the Board should know a delay is possible.
Epistemic calibration is difficult to assess reliably, and our threshold of 60% is a judgment rather than a validated benchmark.
The external evaluator would review and, if necessary, revise the instrument and threshold before baseline data is collected, and any change would be reported to the Board with reasons. We would rather have a contested measure of the right outcome than a clean measure of the wrong one.
$148,000 as a one-time development and pilot investment over eighteen months: a 0.4 FTE reallocation of my time to lead the project, stipends and release coverage for nine teacher fellows, classroom technology, and an independent evaluation contracted outside the district.
What we would deliver is a locally authorized Grade 10–12 elective, nine teachers prepared to teach it, evaluation evidence measured against thresholds set in advance, and a framework published for every other district in the province to use.
Districts across B.C. will need this capacity within the next several years. Most will wait for someone else to develop it. This proposal is an invitation for one district to develop it first, and to carry its name on the framework the rest of the province adopts.
Eighteen months · three secondary schools · nine teacher fellows · roughly 240 students in Grades 10–12 · one authorized course, written by teachers and published for the province.
Citations in the text link to this list. Canadian and B.C. sources are marked, as are peer-reviewed studies.
MacPherson, P. (2026). Teacher survey on artificial intelligence in Canadian classrooms. Fraser Institute. Survey of 756 grades 6–12 teachers across Canada, November 2025 – January 2026.Canada
fraserinstitute.orgAlberta Teachers' Association (2023). Member survey on AI preparedness in Alberta schools.Canada
KPMG Canada (2025). Public opinion survey on generative AI use among Canadian post-secondary students.Canada
Sumanac-Johnson, D. (2025, August 4). Canadian teachers want lessons on how to ethically incorporate AI into classrooms. CBC News.Canada
cbc.ca/news/education-teachers-ai-training-1.7597806B.C. Ministry of Education and Child Care. Considerations for Using AI Tools in K-12 Schools. Covers ethical uses; needs and impacts; accessibility and usability; integration and compatibility; data security and privacy; teaching and learning; inclusive learning.B.C.
www2.gov.bc.ca — considerations-for-using-ai-tools-in-k-12-schools.pdfB.C. Ministry of Education and Child Care. Digital literacy and the use of AI in education: supports for British Columbia schools.B.C.
www2.gov.bc.ca — ai-in-educationCentral Okanagan Public Schools (SD23), rev. 2025. Guidelines for the Responsible Use of Generative Artificial Intelligence.B.C.
lts.sd23.bc.caFocused Education Resources. Artificial Intelligence resources for B.C. K-12. Twenty-one lessons for primary, intermediate and secondary, developed with B.C. teachers.B.C.
focusedresources.ca/artificial-intelligenceB.C. Ministry of Education. B.C. Digital Literacy Framework. Six characteristics identified by B.C. educational leaders, based on the ISTE National Educational Technology Standards for Students.B.C.
www2.gov.bc.ca — digital-literacyB.C. Ministry of Education. ADST 10 Computer Studies: curriculum and elaborations.B.C.
curriculum.gov.bc.ca — adst_10_computer-studies_elab.pdfInstructional design study on machine learning for early adolescents. Decision tree and k-nearest neighbour materials tested with 12–13 year olds in an afterschool AI programme.Peer-reviewed
Vartiainen, H., Toivonen, T., Jormanainen, I., Kahila, J., Tedre, M., & Valtonen, T. (2020). Machine learning for middle-schoolers: Children as designers of machine-learning apps. 2020 IEEE Frontiers in Education Conference (FIE), Uppsala, Sweden, pp. 1–9. Design-based study with 34 middle-school students in Finland.Peer-reviewed
doi.org/10.1109/FIE44824.2020.9273981UNESCO. AI competency frameworks for students and for teachers. Competencies across AI concepts, AI ethics, underlying technologies and applications, and AI system design, at the levels of understanding, applying and creating.International
Sanusi, I. T., et al. (2022). A systematic review of teaching and learning machine learning in K-12 education. Education and Information Technologies. Review of 43 conference and journal articles across curriculum, technology, pedagogy and teacher training.Peer-reviewed
Mazzucato, A., & Larghi, S. (2024). Introducing artificial intelligence and machine learning in K12 education to foster 21st century skills: From theory to practice. Proceedings of the World Conference on Research in Education, 1, 15–27. Curriculum delivered to 70 middle-school pupils, 12 elementary pupils and 60 trainee teachers.Peer-reviewed
doi.org/10.33422/worldcre.v1i1.227EdCan Network (2025, December). Artificial Intelligence in K-12 Education: Frameworks and Guidance.Canada
edcan.ca — Ai-in-K-12-Frameworks.pdfCanada's AI Taskforce (2025). AI Literacy Policy Briefing. Calls for a funded National AI Literacy Strategy supporting provincial and territorial ministries of education.Canada
Estaiteyeh, M. (2025). Canada needs a national AI literacy strategy to help students navigate AI. The Conversation.Canada
theconversation.com — 257513Digital Moment. The AI Literacy Framework for Elementary and Secondary Education in Canada (draft). A living document inviting contribution from educators, school and system leaders, policymakers, Indigenous education authorities and community partners.Canada
digitalmoment.org/ai-literacy-frameworkB.C. Ministry of Education and Child Care. Board/Authority Authorized and First Nation Authorized courses. Includes the searchable BAA sharing platform maintained with Focused Education Resource Services.B.C.
www2.gov.bc.ca — board-authority-authorized coursesB.C. Ministry of Education and Child Care. Board/Authority Authorized Courses and First Nation Authorized Courses Policy. No limit applies to the number of BAA courses used toward the 28 elective credits required for graduation.B.C.
www2.gov.bc.ca — BAA courses policySchool District No. 73 (Kamloops-Thompson). Board/Authority Authorized Courses. Ministry requirement that BAA courses align to the Know–Do–Understand curriculum model.B.C.
sd73.bc.ca — board-authority-authorized-coursesGreater Victoria School District (SD61). BAA Courses, Learning Team. One district's published internal timeline and curricular overlap requirements, cited here as an example of district process rather than as a provincial standard.B.C.
learn.sd61.bc.ca/secondary/baaBuilding AI Literacy in Pre-Service Teacher Education in Canada: A Case Study of Two Cohorts (2025). Journal of Teaching and Learning. Mixed-methods study of 108 teacher candidates in 2024 and 104 in 2025.Canada · Peer-reviewed
jtl.uwindsor.ca — article/view/10041Co-designing Large Language Model Tools for Project-Based Learning with K-12 Educators (2025). CHI 2025. Teachers advocated for support that develops professional practice and augments rather than replaces their role.Peer-reviewed
A Framework for LLM Integration in Secondary Education: Insights from Computing Teachers. Springer. Interviews and surveys with 11 secondary computing teachers producing a six-phase framework: prepare, adapt curriculum, lead, assess, synthesize, share.Peer-reviewed
East China Normal University and Shanghai AI Laboratory (2023). 《中小学人工智能课程指南》 [K-12 Artificial Intelligence Curriculum Guide]. Journal of East China Normal University (Educational Sciences). Identifies immature market provision of student-appropriate data-collection tools, model-training platforms and affordable compute.International
Prepared for the Board of Education. Figures in Canadian dollars. Budget lines are modelled and require verification against current district release-time, per-diem and teacher-on-call rates. References 2 and 3 are cited through secondary reporting and should be replaced with the primary documents before submission. The claim in Section 3 regarding existing provincial courses should be confirmed against the Ministry course registry before this proposal is tabled.