AI and Contemplative Practice: An Inquiry into Professional Learning Infrastructure for Early Childhood Educators
It's 10pm. A teacher is thinking about tomorrow's classroom. She's noticed a pattern - her bilingual students watch quietly during whole-group times but engage actively during small-group work. She wants to understand what's happening so she can support them better. She's meeting with her director tomorrow and wants to bring something more than questions.
She picks up her phone. She remembers something from a recent professional development session - they talked about using AI to help think through classroom observations. She starts to type their names, then stops. The session guidance was explicit: don't enter identifying information into AI platforms - no names, birth dates, addresses, or any information that could identify specific children or families.
Scenario One: The Gate
Without the guidance about how to structure her thinking, she types what first comes to mind: "How do I get bilingual children to participate more in whole group?"
The AI responds with a list: "Here are some strategies to increase participation:
Use visual supports to reduce language demands
Incorporate home languages into instruction
Create smaller group activities within whole group time
Build on students' cultural backgrounds
Provide more wait time after asking questions
Use gestures and body language alongside verbal instruction"
She reads the list. She knows these things. Most of them are already part of her practice.
She puts down her phone.
Scenario Two: The Processing Space
She sits with that for a minute. Then she finds her notes from the professional development session:
"Share observations without identifying children. Ask AI to help you think through what you're seeing before jumping to solutions. Use multiple prompts - this isn't one-and-done. When you're ready for strategies, ask AI to point you toward evidence-based resources: peer-reviewed research, DEC Recommended Practices, position statements from the field. Don't just accept AI's suggestions - ask where they come from."
She picks up her phone again.
First prompt: She types: "I'm observing two bilingual students in my Pre-K classroom. During whole group time (18 children on the carpet, 12 minutes before transition), they watch quietly but rarely raise hands or respond when I ask questions to the group. During small group time (4 children at a table, materials-based activity, no time limit), these same two children talk with each other in their home language, point to what they're doing, and try out English words with peers. What might explain this difference in participation?"
The AI responds: "I notice you're seeing these children engage very differently depending on the context - quiet during whole group but actively communicating during small group. That's a really important observation. Let me think through some of the structural differences between these two settings that might help explain what you're seeing.
In whole group, you have 18 children, a 12-minute time constraint, and questions directed to the whole group. In small group, you have 4 children, no time pressure, and materials they can reference while talking. Some things that might be worth considering: How much time do children have to process before they need to respond? Can they use anything besides verbal English to communicate their thinking - like gestures, materials, or their home language? Are they responding directly to you as the teacher, or are they able to talk with peers?
What do you notice when you think about those differences?"
Second prompt: She writes: "In whole group, I ask a question and usually one child raises their hand right away. I call on that child, they answer, and I move to the next question because I need to cover content in 12 minutes. Children respond directly to me. In small group, I ask a question about what they're building, and children respond to each other while I listen. They point to their structures, they use their home language with each other, then try English. There's no hand-raising - they just talk when they have something to say. I'm seeing that whole group requires immediate English verbal response with no supports, while small group allows processing time, peer talk, gesture, and home language."
The AI responds: "You've really clarified the structural differences here. In whole group, the setup requires immediate verbal response in English, directed to you, with limited processing time because you're moving through content. In small group, children have time to think, can use multiple forms of communication (gestures, materials, home language), and can talk with each other rather than performing for the teacher.
Before we jump to strategies, I'm curious what you notice about what these two children can do when the structure supports them. When they have processing time, materials to reference, peer interaction, and the option to use their home language - what are they showing you about their capabilities? And what happens when the structure requires immediate English-only verbal responses?"
Third prompt: She types: "They CAN participate when they have time to process, when they can use materials or gestures to support their meaning, when they can talk with peers in their home language first, and when there's no pressure to respond immediately in English. They seem to shut down when whole group requires quick English responses without those supports. I think I've been interpreting their quiet watching as 'not participating' when really they're participating differently - they're processing, they're listening. The problem isn't them. The problem is that my whole group structure doesn't support the kind of processing and language use they need."
The AI responds: "This is a significant reframing - you've moved from 'how do I get them to participate' to 'how is my whole group structure creating barriers to their participation.' That's exactly the kind of shift that can lead to meaningful changes in practice.
Now that you're seeing this as a structural issue rather than a deficit in the children, you can start thinking about what modifications might reduce those barriers. What are you thinking about in terms of wait time - could you build in processing time after questions? What about language supports - are there ways to incorporate visual or material references during whole group so children don't have to rely only on verbal English? And what about response options - could children turn-and-talk with a peer before responding to the whole group?"
Fourth prompt: She writes: "Based on what I'm seeing, I need research on: extending wait time in whole group for multilingual children, allowing home language use during whole group processing, and providing visual or material supports so children don't have to rely only on verbal English responses. Can you point me toward evidence-based approaches for these modifications?"
The AI responds: "Yes, there's solid research on all three of these areas. Let me point you to some key resources:
For wait time with multilingual learners, Mary Budd Rowe's classic 1986 research in the Journal of Teacher Education documented that extending wait time from 1 second to 3-5 seconds significantly increased participation from multilingual learners. Mariana Souto-Manning's 2013 book Multicultural Teaching in the Early Childhood Classroom examines how wait time functions differently across cultural communication patterns.
For home language use, the NAEYC position statement Advancing Equity in Early Childhood Education (2019) specifically addresses supporting home language alongside English learning. Linda Espinosa's 2015 book Getting It Right for Young Children from Diverse Backgrounds provides research on why maintaining home language supports rather than hinders English acquisition.
For the structural modifications you're considering, the Division for Early Childhood's Recommended Practices (2014) include specific guidance on creating responsive interactions for diverse learners. Practice E4 addresses creating opportunities for children to use home languages alongside English. The practices document is available at http://www.dec-sped.org/recommendedpractices
These resources should give you some research grounding as you think about modifying your whole group structure. Would it help to think through what a specific modification might look like in practice?"
She had organized her thinking. Tomorrow when she talks with her director, she's bringing observations about what supports their participation, what creates barriers, and what specific structural changes she wants to try.
Four interactions, drawing on the four-pass Listening Guide assignment from teacher preparation: writing what you're thinking, separating observations from interpretations, noticing patterns, synthesizing for action.
Table showing four-pass contemplative mapping structure with examples from teacher scenario: write what you're thinking, separate observations from interpretations, notice patterns, synthesize for action.
What AI Is (And What It Isn't)
According to NYC's Office of Technology and Innovation, artificial intelligence is "a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions."
The critical pieces:
Human-defined objectives - A person asks the question
Machine- and human-based inputs - Trained on existing text (with all the biases that contains)
Automated analysis - Finds patterns from training data
Options for information or action - Gives possibilities to consider
What's missing: It doesn't observe actual classrooms. It doesn't know specific children or families. It doesn't understand particular community contexts. It doesn't replace relationship or the conversation and planning that happens with colleagues and families.
I've been thinking about AI in educational practice for a while now. This week, in my data science seminar at CUNY, we heard from Jiahao Chen, the head of AI and machine learning for NYC's Office of Technology and Innovation. Chen has been working at the center of AI ethics, AI regulation, and municipal use of data. His background spans computational chemistry, machine learning research at MIT, developing responsible AI practices at Capital One and JP Morgan Chase for regulatory compliance, and co-founding Responsible AI to work on risk management for employment decision systems.
Listening to Chen describe how NYC is thinking about AI governance made me consider how I'd be using AI in a university faculty setting. Many teachers don't have the luxury of stepping back to examine new technologies before they're mandated to use them.
When I refer to AI in this blog, I mean Claude specifically, not ChatGPT. Claude is the tool I used to structure my thinking for this piece, and it's the tool I'm proposing teachers consider. The question isn't which AI platform to mandate but whether we can build professional learning infrastructure that supports teachers to use tools like Claude contemplatively rather than as efficiency shortcuts.
Anthropic, the organization that builds Claude, has developed resources for educators—but they're focused on higher education. Claude for Education provides university-wide access with features like Learning Mode, which guides students' reasoning rather than providing answers. The AI Fluency for Educators course teaches faculty how to integrate AI into their practice. The Teach For All partnership brings Claude to 100,000+ teachers across 63 countries, though the focus is K-12 and above. What's missing: implementation guidance for early childhood educators, early interventionists, or preschool special education teachers. The resources that exist don't extend to the educators working with our youngest children.
Research on AI in early childhood and special education teacher preparation is emerging but limited. I scanned the Journal of Early Intervention (JEI), Topics in Early Childhood Special Education (TECSE), Early Childhood Research Quarterly, (ECRQ) Early Childhood Education Journal (ECEJ), and the Journal of Early Childhood Teacher Education (JECTE) and found no peer-reviewed research on AI in teacher preparation or professional learning specific to early intervention or early childhood special education. A recent study documents one early childhood teacher educator's experience integrating AI tools into teacher preparation coursework (Jambunathan, 2025), though not specifically for early intervention or early childhood special education. Research on early childhood teachers' perceptions of AI shows that while teachers recognize AI's potential usefulness, they emphasize the need for adequate training and express concerns about maintaining human interaction in teaching (Lee & Yoon, 2025). The Journal of Special Education Preparation (JOSEP) published a special issue on AI integration in special education teacher preparation (Vol. 4 No. 2, 2024), though it does not specifically address early intervention or early childhood special education. A separate JOSEP special issue focused on early childhood special education teacher preparation (Vol. 4 No. 1, 2024) but does not address AI at all. The research gap at the intersection of AI, early intervention, and early childhood special education is real.
This gap is also an opening. Contemplative mapping — systematic observation combined with contemplative inquiry — makes visible the relational dynamics and institutional patterns shaping practitioners' work. Applied in teacher preparation, it's structured around multiple passes: writing what you're thinking, separating observations from interpretations, noticing patterns that weren't visible initially, synthesizing for action. The four-pass structure in this blog adapts that methodology for AI-mediated professional learning. The question is whether approaches built for face-to-face reflective practice translate to AI-mediated professional learning for teachers already in classrooms.
What's also missing from existing research: any examination of AI as a contemplative tool for teacher professional learning. No studies on structuring AI use around multiple passes, pausing, or examining assumptions. The research focuses on AI as content (teaching about AI) or AI as an efficiency tool (e.g., lesson planning, grading).
This blog proposes approaching AI as a contemplative tool rather than an efficiency tool - grounded in what contemplative mapping has built in teacher preparation but extending into territory where research doesn't yet exist.
When Teachers Need Language They Don't Yet Have
In a study I conducted in partnership with NYC's Division of Early Childhood Education examining how teachers navigate decisions about special education referrals and inclusion (Gupta et al., 2024), teachers struggled with how to describe what they were seeing. One teacher tried to describe differences in abilities and development by saying "I don't know how else to describe it but for lack of a better word, she's slow... I don't mean to sound - I just don't know how to word it. She's a little bit behind but I don't know if it's just because she's not ready or if she actually needs help."
Another teacher spoke about her desire to help children and families: "That's where the emotions come into play and frustration, because definitely I really do want to help you. Not that I can't, but at the moment no one's allowing me to do that."
Teachers had the commitment. Systems failed to provide shared language to describe developmental differences without defaulting to distancing phrases or deficit framing. Systems failed to provide the conditions - time, collaborative structures, shared protocols - that would support developing that language together.
Administrators described the bind: "Teachers need time to plan together, to coordinate. But I can't give them that time without coverage, and I don't have coverage. So they're planning in the hallway, they're texting each other at night. It's not sustainable."
What Contemplative AI Use Could Look Like
Imagine one of these teachers at 10pm, preparing for tomorrow's team meeting about a child. She opens her phone.
First interaction: She types what she's been thinking: "I have a child who seems slow. She doesn't talk much. She doesn't play with other kids. I don't know if she needs help or if she's just not ready."
She reads what she wrote. She sees the word "slow" and hears the other teacher's voice from the study saying "I don't know how else to describe it but for lack of a better word..." She realizes she's doing the same thing.
She revises her question: "I'm trying to describe a child's development to my team tomorrow. She's developing differently from other children in language and social engagement. I don't want to use deficit language. What am I actually observing? How can I describe this in ways that open up conversation about supports rather than labeling her?"
Second interaction: The AI responds by helping her separate observations from interpretations. What specifically does she observe about language? About social engagement? What contexts is she observing in? What has she tried?
She starts writing. The child uses 2-3 word phrases while peers use 4-5 words. She watches other children during play but doesn't join in. During small group, she participates more. During whole group transitions, she shuts down.
The teacher pauses. She sees a pattern she hadn't named before - the child engages more in smaller, quieter contexts.
Third interaction: She asks: "I'm noticing this child participates more in smaller, quieter contexts. What might this tell me about what she needs? What environmental modifications might I try before assuming she needs evaluation?"
The AI prompts questions about sensory environment, language demands in different contexts, peer grouping, wait time. The teacher realizes she hasn't tried reducing group size, providing more processing time, or checking whether language demands match the child's current capacity.
Fourth interaction: She writes: "Help me organize what I want to say to my team. I've been observing patterns in this child's engagement. I want to describe what I'm seeing and suggest we try some environmental modifications first. I need language that centers the child's strengths and what she can do in different contexts."
She drafts what she'll say. She revises it. She had organized her thinking to bring to tomorrow's three-minute hallway conversation with her co-teacher before the team meeting.
he didn't ask one question - she made multiple passes. She paused to see what became visible. She used AI to organize thinking so the brief time she has with actual humans can focus on collaborative problem-solving rather than starting from scratch.
The processing happened at 10pm when she had time. The collaboration will happen tomorrow when she has three minutes.
Questions for Using AI (Not a Framework to Implement)
Contemplative AI use requires questioning at three stages:
Before: What is the teacher trying to understand about how participation works in different contexts? Does she need processing space to organize thinking, or does she need a conversation with people who know this child and family? Can she ask this question without entering children's names, family names, or details that could identify specific individuals?
During: What assumptions shape how she phrases questions? Whose cultural knowledge is missing from AI's response? What would change if she asked this differently - trying multiple approaches rather than accepting the first response?
After: What did this help her see about her practice, and what questions does she need to bring to colleagues and families now? Where does she need to slow down rather than implement strategies?
What the Redesigned System Could Look Like
Protected planning time: Teachers have protected planning time during work hours. During that protected time, they use AI to organize observations before meeting with co-teachers. The AI interaction prepares them for collaboration; it doesn't replace it.
Reflective supervision: Teachers participate in reflective supervision where supervisors ask about their thinking processes and outcomes. In supervision, teachers share how they used AI to examine their own assumptions or map infrastructure gaps. The supervisor helps them see patterns across multiple AI interactions over time.
Learning communities: Learning communities meet regularly with facilitated protocols for examining practice. Teachers bring AI-organized thinking to these communities as data - "here's how my questioning changed across four interactions" - and the community examines what that reveals about shared assumptions or system-level patterns.
Organizational structures: Organizational structures value contemplative practice through time allocation, professional development focus, and accountability measures that assess quality of thinking rather than speed of implementation.
AI within infrastructure: AI supports processing that fuels human collaboration. It doesn't replace the infrastructure - it functions within infrastructure that actually exists.
Closing: What We Should Be Building
When I searched the major early childhood and special education journals, I found no research on AI in teacher professional learning for early intervention or early childhood special education. Anthropic's educator resources serve higher education and K-12. The educators working with our youngest children don't have implementation guidance.
These questions connect to larger conversations about professional learning infrastructure, about who gets left out when new tools emerge, about whether we're removing gates or building new ones.
References
Division for Early Childhood. (2014). DEC recommended practices in early intervention/early childhood special education. http://www.dec-sped.org/recommendedpractices
Espinosa, L. M. (2015). Getting it right for young children from diverse backgrounds: Applying research to improve practice (2nd ed.). Pearson.
Gupta, S. S., Cheatham, G. A., Strassfeld, N., Zhu, X., Medellin, C., & Nagasawa, M. (2024). Examining the ecology of preschool inclusion in New York City: A mixed-methods study underway. Contemporary Issues in Early Childhood. https://doi.org/10.1177/14639491241229229
Jambunathan, S. (2025). Integrating artificial intelligence into early childhood teacher education. Contemporary Issues in Early Childhood. https://doi.org/10.1177/14639491251340141
Lee, J. O., & Yoon, J. (2025). Exploring perceptions of early childhood teachers on the use of artificial intelligence in early childhood education. Journal of Early Childhood Teacher Education. https://doi.org/10.1080/10901027.2025.2600033
National Association for the Education of Young Children. (2019). Advancing equity in early childhood education. https://www.naeyc.org/resources/position-statements/equity
NYC Office of Technology and Innovation. (2024). Artificial intelligence: Principles and definitions. https://www.nyc.gov/assets/oti/downloads/pdf/artificial-intelligence-principles-definitions.pdf
Rowe, M. B. (1986). Wait time: Slowing down may be a way of speeding up! Journal of Teacher Education, 37(1), 43–50. https://doi.org/10.1177/002248718603700110
Souto-Manning, M. (2013). Multicultural teaching in the early childhood classroom: Approaches, strategies and tools, preschool-2nd grade. Teachers College Press.
Sarika S. Gupta, Ph.D., is the founder of Ecological Learning Partners LLC.
