ActiveAI: Introducing AI literacy for Middle School Learners with Goal-based Scenario Learning

Carnegie Mellon University learning engineering / HCI project

Pennsylvania, USA

Design team (extended abstract)

• In-school (K–12)

App-based AI literacy modules using scenario-driven tasks

Modular app sessions (evaluation planned)

Middle school target (grades 7–9)

Tablet/app interactions with intelligent tutor behaviors

• Extended abstract: empirical outcomes not yet reported here
• Ethical risks of social-media–like scenarios require careful content moderation
• Maintaining motivation across abstract ML ideas
• Dependence on quality datasets for classroom-safe examples

Grades 7–9 (middle school)

Uneven prior formal AI instruction

Not required; agents expose algorithms

• Teach Five Big Ideas through authentic scenarios
• Build critical evaluation of AI-generated outputs
• Lower barriers via structured interactions (collector/slider/stepper)

• Instrument interactions for learning engineering research questions
• Embed bias and dataset imbalance lessons in tasks

• Not a completed large-scale efficacy trial in this document
• Not a full-year curriculum specification
• Not focused on teacher PD logistics

Intelligent tutoring / interactive ML scenarios within ActiveAI app

• Tutor
• Evaluator

English-first design context (CMU)

• Collectors for dataset capture/labeling
• Sliders for thresholds and training set size experiments
• Steppers for sequential feature-group exploration in sentiment tasks

• Moderate harmful social content in scenario framing
• Teach skewed training data and spurious correlations explicitly
• Privacy for any user-generated media collected via collectors
• Avoid uncritical trust in model outputs

• Present TikTok-style sentiment scenario with guided steps
• Dog image classification with imbalanced indoor/outdoor bias lesson
• Tutor hints and feedback loops tied to learner inputs
• LEP instrumentation for interaction frequency and hint effectiveness

• Learners steer data and parameters; system gives immediate feedback
• Educators supervise classroom use and debrief ethics

• Goal-based scenarios for purposeful inquiry
• Limited interaction types to reduce extraneous cognitive load

• Middle school AI exposure often limited in traditional curricula
• Concept complexity and math demands threaten engagement
• Ethical interaction risks in realistic social scenarios

• Maps design to AI4K12 big ideas explicitly
• Uses LEP model for systematic instrumentation planning

• Design emphasizes motivating real-world tasks without coding prerequisites

• Hypothesized links between interaction richness and understanding (to be tested)

Positions future analytics on hints, assistance levels, and interaction traces as core to iterative improvement.

• Scenario realism vs age-appropriate content and mental health
• Bias and fairness lessons must avoid stereotype reinforcement
• Data minimization for any camera-based collection
• Clear policies if integrating social-media analogies

• Activity documentation

• Run controlled studies comparing ActiveAI to baseline AI modules
• Validate hint policies and tutor assistance levels

• Middle school AI literacy
• Intelligent tutoring systems
• Learning engineering

• Completed

7–9 (middle school band)

Formal school (intended)

Scenario-based ML literacy

GBS + PBL + tutoring

Medium

Medium