A² Learning Studio · Data Ecosystem

1/8

A²

A² Learning Studio

10 interactive slides with voice narration
Click → arrow keys → or hover for details

A²

Learning Studio

Data Ecosystem

A complete map of how every A² platform collects, processes, aggregates, and exports learning data — serving six distinct stakeholder groups across 21+ scientific research domains.

10+ Platforms6 Stakeholders7 Collection Stages21 Research Domains17+ Export Files5 Data Layers

Dr. Ali Keyvanfar & Dr. Arezou Shafaghat
Dr. Da Hu · Dr. M. Jonaidi · Dr. Simin Nasseri
Kennesaw State University · Georgia Institute of Technology

Six Stakeholder Groups

Hover cards for details

🎓

Students & Professionals

Self-assessment, module replay, dimension tracking, personalized study across all A² platforms

🎓 Students & Professionals

Across 10+ platforms: radar charts, AI narratives, score trajectories, study recommendations. Students in CM courses, professionals in proposal training, and learners across confidential domains all export JSON + Certificate.

Use: Replay modules, track growth, build portfolio

👨‍🏫

Instructors & Facilitators

Teacher Dashboard, cohort analytics, at-risk detection, grading support

👨‍🏫 Instructors & Facilitators

Every A² platform exports to a Teacher Dashboard. Import student JSONs → cohort KPIs, dimension heatmaps, band distributions, risk profiles, engagement scatter. Works for university courses and professional workshops alike.

Use: Grade, detect at-risk, compare cohorts

🔬

Researchers

21+ domains, 17+ export files, self-documenting codebook, SPSS/R-ready

🔬 Researchers

Publication-grade datasets across decision science, learning analytics, risk behavior, human-AI interaction, proposal development skill acquisition, and more. Codebook enables independent replication. Multi-platform = multi-domain evidence.

Use: Peer-reviewed papers, NSF evidence, meta-analyses

🏛️

Administrators

ABET/SACSCOC accreditation evidence, SLO mapping, completion metrics

🏛️ Program Administrators

Aggregate completion rates, band distributions, survey Likert data as direct/indirect SLO assessment evidence. All A² platforms produce ABET-aligned outcome data. Continuous improvement tracking across semesters and programs.

Use: Accreditation portfolios, curriculum reports

🏗️

Industry & Funding Partners

Certificate verification, competency profiles, grant evidence, workforce analytics

🏗️ Industry & Funding Partners

Verify certificates via unique IDs. View competency bands and role-specific performance for hiring. Platform data feeds grant evidence packages. Workforce analytics inform economic development agencies.

Use: Hiring, workforce benchmarking, grant reporting

🤖

GenAI Technology Partners

Google One AI, OpenAI, Perplexity, Claude AI, GitHub Copilot, ElevenLabs — GenAI partners powering ethical, adaptive learning

🤖 GenAI Technology Partners

A² platforms integrate with frontier GenAI services to build capacity for ethical and beneficial AI use: Google One AI for multimodal learning support, OpenAI & Perplexity for research-grounded content generation, Claude AI (Anthropic) for adaptive coaching narratives, GitHub Copilot for simulation development, and ElevenLabs for voice narration. These partners contribute GenAI capabilities that deepen learning while A² data improves AI calibration — a bidirectional ecosystem advancing responsible AI literacy.

Google One AIOpenAIPerplexityClaude AIGitHubElevenLabs

Use: Ethical GenAI capacity building, adaptive coaching, voice narration, research support

Universal 7-Stage Data Collection Pipeline

Hover stages for detail · Shared by all A² platforms

🏠

Welcome

1 · Welcome

Name, session ID, browser info, timestamp. Auto-saves to localStorage across all platforms. Fully offline — no server calls.

IdentityAuto-saveAll A² platforms

🎓

Pre-Module

2 · Foundation

Domain-specific orientation adapted per platform. Examples include infrastructure walkthroughs, challenge previews, material comparisons, and regulatory rule orientation. All track completion gates & timing.

Video gateAdaptiveDomain-specific

🎮

Power Up

3 · Power-Up Game

Blind vs guided mode. Platform-specific warm-ups: tile placement challenges, cold drafting exercises, or scenario previews. Captures intuition-expertise delta as core learning metric.

OptionalScore deltaMotivation

🎯

Simulation

4 · Core Training

Mastery-gated modules ranging from 6 to 18+ per platform. Formats include strategic decision simulations, interactive game modules, and multi-phase interdependent workflows with negotiation & resubmission. All log option chosen, timing, pre/post dimension scores.

Mastery gateMulti-dimCascades

📊

Reports

5 · Reports & Feedback

Radar charts, AI narratives, trajectory graphs, formative assessment. Some platforms add specialized review layers such as simulated panel reviews and expert feedback. All platforms log hover events, duration, thumbs ratings, study preferences.

HoversRatingsStudy prefs

📝

Survey

6 · Program Survey

Standardized instrument across all platforms: Demographics, Learning Experience, Engagement, Outcomes, Your Voice. 29+ questions. Editable after submission. Feeds 21 research domains.

5 sectionsLikert + openStandard

🏆

Certificate

7 · Certificate & Export

Unique certificate ID, final scores, band, track, module list. Print/HTML/JSON export. Refreshable. JSON feeds the Teacher Dashboard. Identical structure across all A² platforms.

Cert IDTriple exportUniversal

🏠

Welcome

🎓

Pre-Module

🎮

Power Up

🎯

Core Training

📊

Reports

📝

Survey

🏆

Certificate

What Makes A² Different

AI enhances the learner — it doesn't bypass them

✗ AI-Bypassed

AI does the work. Student submits AI output. No reasoning traced. No skill built.

✓ AI-Enhanced (A²)

Student makes every decision. AI coaches after each choice — revealing risks, not giving answers.

📚

Case-Based Learning

Real industry scenarios with authentic complexity

📚 Case-Based Learning

Students face real projects — Google, AWS, Apple data centers, actual crises, genuine proposals. Cases preserve professional complexity. AI provides context, not answers.

🔗

Cascaded Decisions

Early choices compound into downstream consequences

🔗 Cascaded Decisions

Choose cheap materials → quality failure downstream. Skip assessment → regulatory delays. AI traces exactly how Decision A propagated to Outcome X. Students learn systems thinking.

🤖

AI-Scaffolded Coach

Feedback after every decision, never the answer

🤖 AI-Scaffolded Coaching

After EVERY decision: impact analysis, hidden risk reveal, expert comparison, revision prompt. AI never says "choose B" — it says "here is what happened because you chose A."

📊

Assess Reasoning

Not answers — decision quality, process, and systemic thinking

📊 What We Assess

Decision quality (trade-offs, risk appetite). Learning process (time curves, self-correction). Systemic thinking (cascade awareness, compound consequences). Every layer captured as research data.

21 Research Domains

Hover any domain · Each fed by multiple A² platforms

17+ Export Files

Hover files for details · Per-platform + cross-platform

5-Level Data Flow Architecture

Input → Process → Output · Each level produces a traceable dataset report · Hover for I/O specification

DFD 1.0 · Interaction Capture

Input: Raw user events (clicks, hovers, timing, selections) → Process: Client-side event logger with ms timestamps → Output: interaction_log.json

DFD 1.0 · Interaction Capture

Every user action is captured client-side in real time: option selections, hover durations, navigation paths, revision events, and decision timestamps. Stored in localStorage with no server dependency. This raw event stream is the atomic data unit for all downstream processing. Output: interaction_log.json — the immutable audit trail.

Event streamClient-sideZero-latency

DFD 2.0 · Learner Analytics Engine

Input: interaction_log.json → Process: Score computation, dimension mapping, band classification → Output: masterdata_[user].json

DFD 2.0 · Learner Analytics Engine

Transforms raw interactions into structured learner analytics: multi-dimension scores, competency band classification, risk appetite profiling, decision trajectory curves, and metacognition indicators. One comprehensive JSON per learner per platform — the single source of truth for all downstream consumers. Output: masterdata_[user].json.

ETLScore engineBand classification

DFD 3.0 · Cohort Aggregation

Input: N × masterdata.json files → Process: Cross-learner aggregation, heatmaps, risk profiling → Output: cohort_dashboard_report

DFD 3.0 · Cohort Aggregation

Instructor imports multiple learner JSONs via drag-drop. System auto-computes: cohort KPIs, dimension heatmaps, band distributions, at-risk flags, engagement scatter plots, and survey cross-tabulations. Supports synthetic data generation for testing. Output: cohort_dashboard_report — the instructor's decision-support dataset.

N:1 mergeAt-risk detectionHeatmaps

DFD 4.0 · Research Data Pipeline

Input: Cohort + individual data stores → Process: Flattening, normalization, codebook generation → Output: 17+ publication-ready CSV/JSON files

DFD 4.0 · Research Data Pipeline

Transforms aggregated data into publication-grade research datasets: decisions.csv, modules.csv, trajectories.csv, tradeoffs.json, survey.csv, feedback_engagement, and more — 17+ files total. Every file includes certificate ID for cross-referencing. Self-documenting codebook.json enables independent replication. SPSS/R/Python-ready. Output: complete research data package.

17+ filesCodebookReplicable

DFD 5.0 · GenAI Capacity Data

Input: Decision context + learner state → Process: GenAI inference (Google One AI, OpenAI, Claude, Perplexity, ElevenLabs) → Output: ai_coaching_log.json, voice_assets, feedback_loop_data

DFD 5.0 · GenAI Capacity Data

Bidirectional data exchange with frontier GenAI services. Upstream: Decision context, learner state, and domain parameters feed into Google One AI, OpenAI, Perplexity, Claude AI, and ElevenLabs APIs. Downstream: AI-generated coaching narratives, adaptive feedback text, voice assets, and research-grounded content flow back into the platform. Feedback loop: Learner interaction patterns with AI outputs (engagement, revision rates, satisfaction) generate feedback data that informs future AI response quality. Output: ai_coaching_log.json, voice_narration_assets, genai_feedback_data — the data layer that closes the human-AI learning loop.

Google One AIOpenAIPerplexityClaude AIBidirectional

Output: ai_coaching_log.json · voice_assets · genai_feedback_data

The Future · Beyond the Classroom

Hover tiles for vision details

⚡

Just-in-Time Decision Intelligence

Real-time AI guidance for any professional facing a cascading decision chain — in construction, proposal writing, or any multi-phase domain.

⚡ Just-in-Time Decision Engine

A PM facing a change order cascade, a PI facing a panel rejection — the A² cascade model becomes a live decision-support tool. Input the first domino, see every downstream consequence. Trained on thousands of student/professional decision traces across all A² platforms.

Live guidanceCascade predictionAI-powered

Construction, proposals, any interdependent decision chain

🎓

Student Success & Retention

Early warning system using engagement + decision patterns across all A² platforms to predict at-risk learners.

🎓 Predictive Student Success

Behavioral data across 10+ platforms creates a rich multi-signal early-warning dashboard. Instructors intervene before disengagement. Cross-platform patterns reveal deeper learning profiles than any single tool.

Early warningCross-platform

University-wide learning analytics

🏗️

Workforce Competency Passport

Portable, verifiable digital credential portfolio spanning multiple A² platforms and domains.

🏗️ Competency Passport

Certificates from multiple A² platforms form a multi-skill portfolio. Employers verify via certificate IDs. Professionals build credentials across construction, risk management, proposal development, and other domains.

Multi-platformPortable

Construction, academia, engineering hiring

🌍

Cross-Industry Expansion

The A² engine adapts to healthcare, transportation, energy, K-12 CTE — any domain with multi-phase cascading decisions.

🌍 Cross-Industry Platform

The 7-stage architecture is domain-agnostic. Replace "data center" with "hospital" or "transit hub". Replace "NSF proposal" with "EU Horizon grant". Cascade effects, stakeholder scoring, and formative feedback apply universally.

HealthcareEnergyK-12

Any built-environment or grant-writing discipline

🤖

AI-Trained Decision Models

Decision traces from 10+ platforms train recommendation models that learn what expert decision-makers do differently.

🤖 AI Decision Models

Thousands of decision traces become ML training data. Models learn: when experts face cascade X, they prioritize dimension Y. When strong PIs face a rejected proposal, they pivot strategy Z. This closes the loop — human data improves AI, which improves future training.

ML training dataExpert patterns

Cross-platform AI research

📊

National Benchmarking

Aggregate anonymized data across universities and workshops to establish national competency benchmarks.

📊 National Benchmarks

Multi-institution deployment across all A² platforms produces standardized competency datasets for construction management and research skill education. Compare programs, identify national skill gaps, inform ABET standards.

Multi-institutionABET

Accreditation & policy

🚀

Proposal Skill Development at Scale

Proposal development platform data reveals what separates successful proposal writers from novices — patterns that can be taught.

🚀 Proposal Intelligence

The proposal training platform traces how users build narratives, negotiate with program directors, and respond to panel feedback. At scale, this reveals the decision patterns, strategies, and revision habits that separate funded PIs from rejected ones. A new field: computational grantsmanship.

GrantsmanshipPattern mining

Faculty development, research offices

A²

Learning Studio

10+ Platforms.
Six Audiences.
One Data Architecture.

A² Learning Studio
All data stays in the learner's browser. No servers. No tracking. Full privacy.
For educational use only — not commercialized.

🚀 Try the Showcased Platforms

Explore interactive demos on Dr. Keyvanfar's Hands-On page

Kennesaw State University · Georgia Institute of Technology
Dr. Ali Keyvanfar
Dr. Arezou Shafaghat
Dr. Da Hu
Dr. M. Jonaidi
Dr. Simin Nasseri

10+

Platforms

Research Domains

17+

Export Files

200+

Decision Points / Platform

Server Calls

∞

Privacy