Patrick Glatz | Learning & Decision Systems

Featured Exhibit: Conceptual Progression

A sequential navigation of core design concepts. Visitors are encouraged to explore this progression in order.

Concept 1: Instructional Pressure

Regulation First

The Invisible Variable

In high-stakes learning environments, the exact level of "instructional pressure"—the balance of demand and support that drives student engagement—is completely invisible to the designer until a student silently disengages or fails.

The Visual Intervention

The Regulation First environment surfaces this pressure using a structured diagnostic matrix mapping student response rates, prompting density, and error clusters in real-time.

The Decision Outcome

By rendering instructional pressure visible, educators can immediately adjust demand levels to keep learners in the optimal zone of active engagement, preventing silent frustration and behavioral drift before they manifest.

View Paper (PDF) →

Primary Artifact: Zone of Engagement

System Interfaces

Concept 2: Planning

Training in Dungeon

The Invisible Variable

In AI navigation and execution tasks, the model's planning phase—the sequence of decisions it intends to make—is completely invisible, hidden behind a black box that generates immediate actions.

The Visual Intervention

The Dungeon environment forces the system to commit to its entire instructional route (using commands like Forward, Left, Right) in writing before the first physical movement occurs.

The Decision Outcome

Surfacing the plan prior to execution isolates logical planning errors from execution mistakes. If navigation fails, the user can verify whether the plan itself was flawed or if the model diverged during run-time, allowing precise code adjustments.

View Paper (PDF) →

Interactive Application Sandbox

dungeon-game.local

System Workflow

Environment Preview

Concept 3: Adoption

AI Enablement Framework

The Invisible Variable

The trust, relevance, and workflow fit that determine whether people actually adopt AI after the training ends.

The Visual Intervention

The framework makes adoption visible through five stages: Pain, Possibility, Responsibility, Structure, and Practice.

The Decision Outcome

Designers and administrators can identify adoption friction, build trust, and support responsible AI use before a rollout becomes another unused technology initiative.

Download Case Study (PDF) →

Primary Artifact: Enablement Framework

Prompt Structure Analysis

Case Study Documentation Practical AI Adoption.pdf

Download Project Materials

Explore the workshop materials developed to support responsible AI adoption in educational settings. These resources include participant materials, facilitator notes, and presentation assets.

AI Enabled Educators – Participant Workbook PDF

Reflection activities, prompt frameworks, guided practice, and implementation planning for workshop participants.

View PDF Download PDF

AI Enabled Educators – Facilitator Guide PDF

Slide-by-slide facilitator notes, discussion prompts, and implementation guidance for workshop delivery.

View PDF Download PDF

AI Enabled Educators – Presentation Deck PDF

The workshop slide deck used during instruction and demonstration activities.

View PDF Download PDF

Concept 4: Delegation

AI Authorship Alignment

The Invisible Variable

In AI-assisted workflows, the gradual erosion and transfer of cognitive ownership (authorship) from the human to the AI system is invisible, masked by high-quality, polished outputs.

The Visual Intervention

The AI Authorship Evaluator uses a structured rubric auditing agency, cognitive engagement, and evidentiary control.

The Decision Outcome

Exposing this ownership shift makes it impossible for users to unconsciously outsource reasoning. It lets designers set explicit boundaries on where human judgment must dominate, maintaining intellectual and evidentiary control.

View Paper (PDF) →

Interactive Evaluator Sandbox

authorship-evaluator.local

Authorship Evaluation Process

Authorship Fit Assessment

Bridge Project: Model Behavior

Operational Integrity System

The Invisible Variable

In advanced AI pipelines, task drift—the model silently shifting from the requested problem to an easier or adjacent interpretation—is invisible because the generated text remains coherent.

The Visual Intervention

OIS exposes this drift by implementing a pre-generation control matrix (Prompt Auditor) and a post-generation validation check against an independent database.

The Decision Outcome

Surfacing execution boundaries allows users to reject and recompile drift errors programmatically, providing a verifiable audit trail for safety-critical systems.

Explore Governance Repository →

Governance Landing Page

Operational Integrity System Infographic Diagram

OIS Framework Principles & Interface

Research and Writing

Supporting investigations focusing on structural visibility across different decision domains.

This section traces a single observation across four pieces of work: AI systems can satisfy every visible instruction while silently solving a different problem than the one they were actually given. "You Just Got GPTed" names the pattern. Advisory Drift in Adaptive Systems and System Behavior Under Input Compression investigate it empirically, in decision-making and constraint-translation contexts. The Governed Nonfiction Writing Companion responds to it directly, designing a system that refuses to fill that silence on a student's behalf.

You Just Got GPTed

You Didn't Use AI Wrong. You Just Got GPTed.

Open Article PDF →

Advisory Drift

In advisory contexts, users are influenced not only by what a system recommends, but by how that recommendation changes. This study examines how recommendation shifts occur and explores the risks introduced when those revisions happen without clear justification.

Open Case Study PDF →

Behavior Under Compression

This study examines how AI-generated accommodations change under realistic classroom input conditions. Rather than testing correctness, the focus is on consistency, reliability, and constraint adherence when information becomes limited or incomplete.

Open Research PDF →

Governed Writing Companion

This study explores how authorship and judgment change within highly reliable AI-assisted writing workflows. The focus is not on output quality, but on preserving visibility into the decisions that shape the final work.

Open Research PDF →

About

Observational Methodology & System Control

Patrick Glatz is an educator and systems designer with more than two decades of experience, beginning in behavioral analysis and vocational program management before moving into K-12 instructional design.

In July 2025, a cycling accident left him with a herniated disc and unable to walk for several months. Rather than wait to find out whether he'd return to physically demanding special education work, he used the time to build deep technical fluency in AI, a deliberate choice rather than a casual interest.

His work explores how people learn, adopt new technologies, and make decisions within complex systems. Recent projects examine AI enablement, instructional design, behavioral systems, and human-AI collaboration.

Across projects, a common principle remains constant:

"Making invisible processes visible so people can make better decisions."

AI Governance & Operational Integrity

For formal frameworks, technical specifications, and the Operational Integrity System (OIS) codebase, visit the sibling site.

Go to Governance.PatGlatz.com →