A constrained learning environment for planning-before-execution.
An instructional design artifact demonstrating how constrained environments can teach planning, execution, and revision through literal system behavior.
https://glatzp.github.io/Training-In-Dungeon/
Training in Dungeon is an interactive instructional environment that teaches learners to define intent, plan completely before acting, observe literal execution, and revise their thinking after failure.
Unlike many educational tools, it deliberately removes hints, previews, and mid-run corrections so that planning, and not reaction, becomes the focus.
Training in Dungeon
A constrained learning environment for planning, execution, and judgment
- Learning Experience Design
- Instructional Design
- Human-AI Interaction
- Constraint-Based Learning
- Educational Technology
- JavaScript, HTML, CSS
Many discussions about AI, coding, and digital tools focus on what systems can produce. Far fewer address the upstream skills required to use those systems responsibly.
This project explores a simple but underdeveloped question:
What happens when learners must commit to a plan before acting and take responsibility for the outcome?
Training in Dungeon removes hints, previews, and corrective scaffolds in order to make planning, failure, and revision visible. The goal is not engagement or efficiency, but judgment.
Training in Dungeon is a browser-based instructional design experiment centered on deliberate planning and literal execution.
Learners must write a complete set of instructions before anything moves. Once execution begins, the system follows those instructions exactly as written. It does not infer intent, offer corrections, or intervene mid-run.
When an outcome does not match the goal, the mismatch is attributable to the plan itself rather than timing, dexterity, or hidden rules.
This project is intentionally not a coding tutorial, a Scratch alternative, a gamified learning experience, or an AI tool.
The command system exists only to make thinking executable and observable. It is a constraint, not the learning objective.
The design targets a recurring pattern: learners act before fully specifying intent, anticipating consequences, or defining success.
Across its scenarios, the environment surfaces six common breakdowns in planning and system use: unclear intent, imprecise instructions, insufficient anticipation, unexamined errors, shallow revision, and weak evaluation of results relative to goals.
These breakdowns appear in programming, writing, engineering, and AI-assisted work, but are rarely addressed explicitly at an early stage.
Several common educational features are deliberately excluded.
There are no previews, hints, adaptive corrections, scoring systems, or time pressure. These omissions are intentional. They create space for learners to pause, anticipate outcomes, observe failure clearly, and revise their thinking without external judgment.
Failure is treated as information rather than as a penalty.
This environment works best for learners who are able to slow down, reflect on outcomes, and revise their thinking over multiple attempts.
It has been designed with upper elementary and middle school learners in mind, but the underlying mechanism is transferable to older students and adult contexts where planning and judgment matter.
The dungeon is not a narrative goal or a game mechanic. It is a constraint.
A limited grid, minimal commands, and visible consequences reduce distraction and keep attention on the quality of planning rather than interface mastery or creativity.
Success is not defined by speed, efficiency, or completion.
Success is defined as increased awareness of how planning, assumptions, and revision shape outcomes under literal system behavior.
This repository represents an expanded proof of concept demonstrating multiple planning failure modes and structured variation.
It is shared as a thinking artifact rather than as a finished curriculum or product.
This project intentionally sits upstream of tools, platforms, and AI systems. Its purpose is to strengthen the planning and judgment habits those systems quietly depend on.
This project is licensed under the MIT License. See the LICENSE file for details.