Reclaiming Agency
Educators as Architects of AI-Enhanced Learning Architectures
Associate Professor in Learning Sciences and Technologies
University of Pennsylvania Graduate School of Education
Language Educator Symposium 2025
Spring 2023
My first AI policy
Here is Alex
I wasn’t answering a simple question.
“Can I use X for Y?”
These are questions on the architecture for learning.
Revealing the Architecture
Shearing Layers of Change
Each layer changes at different pace, and acts as a platform for the layer above it.
Stuff (days-months) Furniture, books
Space Plan (3-30 yrs) Interior layout
Services (7-15 yrs) Wiring, plumbing
Skin (20 yrs) Exterior surfaces
Structure (30-300 yrs) Load-bearing
Site (eternal) Geographical setting
Learning Architectures
Layers in Learning Architectures
Alex: “Can I use ChatGPT to outline my paper?”
Buildings
| Stuff |
| Space Plan |
| Services |
| Skin |
| Structure |
| Site |
Classrooms
| Practices (days): Activities, interactions |
| Plans (wks-mons): Lessons, assignments |
| Protocols (mons-yrs): Norms, routines |
| Pedagogies (yrs): Approaches, relations |
| Principles (decades): Foundations, lens |
| Philosophy (centuries): Epistemologies |
Principles of Thinking Like an Architect
1. Recognize the Layers
What layer am I actually working at?
Locate which layer a problem or decision primarily affects. A student question might seem like a practice issue, but may be more relevant to the pedagogy layer.
Look deeper—surface questions often have slow-layer reasons. A fast-layer question may stem from ambiguity in the slow layers below.
Principles of Thinking Like an Architect
2. Work Across the Layers and with Tensions
What other layers should I also consider?
- Fast learns, slow remembers. Fast gets attention, slow has power. Practices experiment and adapt; principles and philosophy provide stability and wisdom
- Tensions arise between layers during sudden change. New technologies can create ripple effect across layers.
Human Experience in the Architecture
“Architecture is a mediation between the world and our minds. Good architecture tells us something about the world… and finally, it tells us who we are.”
— Juhani Pallasmaa
Dwelling Precedes Building (Martin Heidegger)
be → dwell (wohnen) → build
- Dwelling = being genuinely “at home”; not just occupying space
- Building should serve dwelling, not determine it
Space is a social product (Henri Lefebvre)
A triad that’s always in tension:
- Conceived space: Expert-designed plans
- Perceived space: How people actually use it
- Lived space: Emotional, symbolic experience
The political problem: Conceived space dominates—expert abstractions override lived reality
“The task of architecture is to defend the autonomy of human experience, not to condition, dominate or dictate.”
— Juhani Pallasmaa
Principles of Thinking Like an Architect
3. Begin with Dwelling, Not Building
How do learners actually dwell in the space?
- Ask: “How do we dwell?”, not “What can we build?”
- Start with lived experience, not technological capability.
- Let perceived and lived space inform design: Resist expert abstractions dominating learner reality and experience
Principles of Thinking Like an Architect
4. Mediate, Don’t Dominate
Does this defend autonomy or impose control?
- Center human autonomy: They navigate; algorithms support
- Connect learners to different worlds, without prescribing the journey or shying away from struggle
- Resist conceived-space domination: Don’t let institutional metrics dictate real learning
A Tale of Two AI Apps
Language Tutor: “Complete exercises. Here is your badge.”
- Conceived space dominates: expert metrics override lived experience
- Fragments language from culture and identity
- Corrects every “error”—little space for struggle
- Placeless, generic—no belonging or attachment
Learners occupy, never truly dwell. App dictates and conditions behavior.
Language Pathfinder: “What worlds do you want to explore?”
- Lived experience informs design: goals, culture, identity matter
- Integrates language with authentic communication contexts
- Honors struggle and situated perspectives as essential to learning
- Enables attachment, memory, belonging—a place to return to
Learners dwell, feeling “at home.” App mediates and promote autonomy.
Reclaiming Agency as Educators
The best AI tools for learning have not been created yet
e.g., cultural bias (Tao et al., 2024), gender bias (Du et al., 2025), pedagogical bias (Chen et al., 2025)
Action 1. Think Like Architects, Not Users
| Principles | Users ask | Architects ask |
|---|---|---|
| Layers | Does this tool work? | What layer am I working at—Practices or Principles? |
| Tensions | Can I use this for grading? | How does this change affect slow layers? |
| Dwelling | Will it help students? | What is authentic learning and does this tool serve it? |
| Autonomy | Should I allow this tool to be used? | How does this tool mediate relations? Does it dominate or defend student autonomy? |
Moving from thinking
and critiquing
to building
Action 2. Be an Educator–Builder
The promise: AI enables educators to build, with taste
- Express professional craft in new mediums: Create tools that embody your pedagogical expertise—what you know about learning that tool engineers don’t
- Build with architectural intention: Respect layers (fast/slow), dwelling (authenticity), and autonomy (mediate, don’t dominate)
- Build with purpose, aesthetics, and care: Enable learning experiences that are not just functional, but meaningful, beautiful, worthy of attachment
Example: ChatLab
An app I built for my learning theories class
Principles in Action:
- Learning is multifaceted, requiring dialogue and reflection [Dwelling]
- ChatLab facilitate connections with theorists [Mediate]
- Students set the topic, ‘invite’ voices, and critically analyze and reflect [Autonomy]
Results: New practices and protocols, but same pedagogy. Surprising connections by students that deepen understanding.
Action 3. Maintain and refresh professional identity
For teacher educators, department chairs, professional leaders
- Position educators as architects: Support educators building learning architecture that embody their professional expertise
- Innovate practices, preserve principles: Revamp fast layers (practices, tools) without mindlessly bulldozing slow layers (pedagogical values)
- Build collective knowledge: Create repositories of exemplars, frameworks, workflows, infrastructures
- Change with intention: Adapt responsibly—be aware but critical of fashion, protect foundations, evolve with purpose
Example: CraftPad
An AI-embedded canvas for teachers to design knowledge-building lessons
Principles in Action:
- Teachers craft designs based on pedagogical principles and their goals [Autonomy]
- Teacher design is iterative [Dwelling]
- Teachers and AI interact in many ways on the canvas that foreground their design [Mediate]
- Built-in AI coach for continual reflection [Dwelling]
Result: Teachers designing with AI iteratively, not consuming AI-generated content. Professional agency is nurtured.
Reclaim Agency: Educators as Architects
Four Principles
- Recognize the Layers
- Work Across Layers and With Tensions
- Begin with Dwelling, Not Building
- Mediate, Don’t Dominate
Three Actions
- Think like architects—not users
- Be educator-builders—build with taste
- Maintain and refresh professional identity—adapt responsibly
Let’s reclaim the agency to shape the learning architecture for generations to come
Bodong Chen
cbd@upenn.edu • bchen.net
penn-wonderlab.github.io
Penn • 12/2025