PAIDEIA (Program for AI-Driven Education, Integration, and Advancement) is GRC's framework for practical AI fluency. It combines a common foundation in context engineering, agent systems, evaluation, and local inference with applied modules for research, coding, writing, and scientific computing.
PAIDEIA has been used within GRC and with interdisciplinary collaborators. Its curriculum and assessment methods continue to develop.
Purpose
AI fluency is more than learning a particular interface. Researchers need to decide where AI belongs in a workflow, supply the right context, connect models to tools, evaluate outputs, and preserve scholarly judgment.
PAIDEIA teaches those decisions through work on real research problems. The framework emphasizes reproducibility, provenance, privacy, and careful validation alongside technical fluency. Its aim is to help scientists redesign workflows without weakening disciplinary standards.
Learner Pathways
People enter with different experience and different responsibilities. PAIDEIA organizes instruction around three pathways that share the same technical foundation.
Newcomer
Build accurate mental models of what current systems can and cannot do. Guided exercises develop the judgment needed to select appropriate tools and evaluate their outputs.
Practitioner
Connect isolated uses of AI into a coherent workflow. The curriculum introduces context engineering, tool integration, agent orchestration, and systematic evaluation.
Advanced Practitioner
Design multi-model systems, operate local inference infrastructure, and manage privacy, cost, reliability, and performance across complex research workflows.
Curriculum Architecture
PAIDEIA has two layers. Every learner studies a Universal AI Fluency Foundation covering transferable technical and evaluative skills. Domain Modules then apply those skills to the methods, data, constraints, and standards of a particular field.
The shared foundation gives collaborators a common vocabulary. Domain modules preserve the practices that make each discipline rigorous.
Applied AI Fluency
Discipline-specific data, workflows, safety constraints, evaluation methods, and agent configurations.
Core Competencies
Eight modules covering the concepts and practices needed to use AI responsibly in technical work.
Universal AI Fluency Foundation
The foundation develops a common technical vocabulary and a repeatable approach to workflow design.
AI Landscape and Mental Models
How foundation models behave, where their capabilities end, and how to reason about uncertainty, context limits, cost, and failure modes.
Context Engineering
How to write, select, compress, and isolate context for complex work. Learners diagnose context loss, contamination, conflict, and overflow.
Agent Architectures and Orchestration
How agents plan, act, retain state, recover from errors, and coordinate with people or other agents. The module also asks when a simpler workflow is preferable.
Tool Integration
How agents connect to databases, APIs, file systems, schedulers, and scientific instruments. The module covers the Model Context Protocol, permission boundaries, and integration design.
Local Inference and AI Sovereignty
How to select, deploy, and optimize local models when privacy, controlled data, cost, or institutional independence precludes a cloud-only workflow.
Evaluation, Trust, and Reproducibility
How to validate outputs against domain knowledge, trace provenance, calibrate confidence, and design reproducible evaluations for non-deterministic systems.
Human-AI Collaboration
How to divide work between people and models, review prompts and outputs, establish validation gates, and retain human responsibility for consequential decisions.
Managing Heterogeneous Agent Fleets
How to assign tasks across models with different capabilities, privacy properties, costs, and latency. Learners study routing, monitoring, fallback, and resource allocation.
Applied Modules
The current curriculum applies the foundation to four areas of scientific and technical work.
AI for Science
Scientific computing, simulation, experimental design, and data analysis in AI and high-performance computing environments.
Topics include scientific data formats, multimodal data, instruments and schedulers, literature-grounded hypothesis generation, and physical validation of generated results.
AI for Coding
Software development with AI assistance across prototyping, implementation, testing, debugging, refactoring, review, and deployment.
The module emphasizes multi-file context, test-based validation, code review, and choosing among chatbot, copilot, and agent workflows.
AI for Writing
Academic writing, technical documentation, reports, and professional communication with explicit safeguards for scholarly integrity.
Topics include iterative drafting, citation verification, source management, technical style, and review across different audiences.
AI for Research
Literature review, question refinement, hypothesis generation, experimental design, analysis, and knowledge synthesis.
The module covers multi-agent review, knowledge graphs, provenance, reproducible analysis, and evaluation of generated claims.
PAIDEIA is also being extended toward administration, law, policy, and economics. The law curriculum is being developed with the Chicago-Kent College of Law. Economics work is being developed with the University of Piraeus through visiting scholar Harry Agyropoylos.
Learning Formats
PAIDEIA is a framework rather than a single course. Its curriculum can be taught through formats matched to a learner's experience, schedule, and research problem.
Intensive Bootcamps
Hands-on instruction, team projects, and a capstone problem for a research group or cohort.
Agentic Hackathons
Team challenges focused on agent architecture, context design, and the validity of results under practical resource constraints.
Prompt Review Sessions
Structured peer review of prompts, context architectures, agent configurations, and validation plans.
Local Inference Workshops
Deployment and operation of local models for private, controlled, or air-gapped environments.
Semester Courses
The complete foundation and an applied module, developed through projects connected to a student's research.
Research Integration Sprints
Focused redesign of an existing research workflow, with an operational pipeline as the outcome.
Practice and Assessment
PAIDEIA treats AI systems as research instruments that require calibration and review. Learners work on their own code, data, writing, and analysis rather than synthetic exercises.
Design the Workflow. Instruction begins with the research process and its constraints, then selects the appropriate models and tools.
Engineer the Context. Learners make source selection, compression, isolation, and provenance explicit parts of system design.
Validate the Result. Assessment focuses on the quality, traceability, and reproducibility of work rather than tool usage alone.
Preserve a Private Path. Local inference remains part of the curriculum for work that cannot use external model providers.
Retain Human Responsibility. AI can extend a researcher's reach, but disciplinary judgment and consequential decisions remain with people.
Teach for Transfer. Shared methods allow learners to adapt as models, interfaces, and domain requirements change.