Teaching AI to K-12 Learners: Lessons, Issues, and Guidance

Independent author / education research practice (SIGCSE contribution)

United States (author affiliation Austin, TX)

Researcher–practitioner synthesis for CS education community

• In-school (K–12)

Scholarly position paper referencing diverse curricula, tools, and studies

N/A

N/A — addresses national/international K-12 AI education efforts

Examples include Teachable Machine, block-based ML extensions, unplugged games, HS coding pathways for training

• Field evolves faster than multi-year landscape reviews can capture
• Parallel non-CS AI literacy progressions risk weak alignment to core CS concepts
• LLMs and copilots shift which programming competencies are emphasized
• Teacher preparation remains a central bottleneck

K-12 across grade bands with different depth expectations

Increasing informal exposure via consumer AI and school tools

Varies by grade; secondary paths may include block or text programming for deeper ML

• Clarify what K-12 learners should know about AI/ML techniques and limitations
• Promote ethics, bias examination, and socially relevant ML investigations across ages
• Leverage AI4K12 progressions while minding integration with CS standards

• Advance data agency alongside traditional CT where ML replaces rule-driven control flow
• Broaden participation through culturally sustaining and relevant pedagogies

• Not an exhaustive systematic review of all K-12 AI education literature
• Not focused on using LLMs as instructional assistants inside courses
• Not a substitute for dedicated non-technical AI literacy programs in other subjects

Pedagogical survey across tools (unplugged kits, web ML trainers, block environments, APIs)

• Tutor
• Co-creator
• Automation tool

Primarily English-language curriculum ecosystem referenced

• Start from embodied/unplugged intuitions about data and training
• Progress toward using pretrained models/APIs and selectively unpacking internals at secondary level
• Embed critique of real deployed systems (e.g., drawing games, justice-related datasets)

• Center harms to marginalized communities when teaching bias and fairness
• Use level-of-abstraction scaffolds rather than all-or-nothing formalism for ML mathematics
• Maintain intellectual honesty about what students can verify versus trust

• Motivate K-12 AI/ML teaching amid industry acceleration and civic stakes
• Summarize goal typologies (literacy vs technical depth vs ethics-focused pathways)
• Review pedagogical strategies from unplugged through code-first ML lifecycles
• Close with teacher preparation, CS–AI integration, and adaptive ethics-forward design

• Teachers curate progression and ethical framing; students practice responsible creation and critique
• Tool vendors do not determine learning goals—standards and classroom ethics do

• Subgoals, Parsons problems, and staged code disclosure for complex algorithms
• Multiple engagement levels for high school ML depth (e.g., decision trees to GAN components)

• Tension between rapid tool innovation and stable, assessable curricula
• Risk of fragmented AI literacy strands disconnected from CS conceptual infrastructure
• Persistent gaps in evidence on critical empowerment outcomes from making activities

• Explicitly narrows scope to teaching AI/ML while noting broader AI literacy obligations elsewhere
• Imports durable lessons from decades of K-12 CS education research

• Documents vibrant early ecosystem of modules, courses, and ethics-forward projects
• Highlights AI4K12 Five Big Ideas as a widely adopted organizing framework

• Shows plural, evidence-informed pedagogies rather than a single best approach
• Emphasizes data-centric epistemologies as the new backbone alongside classical CT

The paper functions as guidance to keep K-12 AI education learner-centered, teacher-supported, ethically grounded, and deliberately coordinated with CS education as the field matures.

• Classroom use of datasets and APIs should follow student data minimization and consent norms
• Teaching with real biased systems requires care to avoid retraumatization or shallow 'gotcha' moments
• Generative AI shifts academic integrity and skill emphasis—policies must evolve with pedagogy
• Equity requires access to quality tools and teacher PD, not only open resources online

• Activity documentation
• Practitioner observation

• Design studies comparing integration models: standalone AI course vs embedded CS vs cross-curricular
• Measure long-term retention and transfer from unplugged ML intuitions to formal modeling

• K-12 AI and ML education
• Computing education research
• Ethics and fairness in computing pedagogy

• Completed

K-12

Formal schooling (US-led discourse, global relevance)

Technical ML education + ethics integration

Plural (unplugged, PBL, embodied, code-first tiers)

Medium

Medium (datasets, APIs, student work)