Seeing Past the First Reading: What AI Can Build in EI/ECSE/ECE Preparation
At George Mason University (2012), in my first semester as a faculty member, I took off points on a journal assignment for punctuation and incomplete sentences. I was preparing students to work in systems where they would be expected to write IFSPs and IEPs, and to communicate professionally with families in writing. What I missed was that journal writing was processing space, the place where students connect their own ideas and experiences with course content before they can show what they know in more forward-facing work.
I came to the faculty role as the daughter of professional immigrants from a country where teachers are respected, where a title is not a formality but a recognition of what the work requires. I asked to be called Dr. Gupta. I was a young Brown woman in a field that had not expected me to hold the title I held, and I knew it every time I walked into a room and had to establish, again, that I belonged there. I let that fear get the best of me.
A student disagreed with her journal grade. She went to her advisor, then to the chair. I met with the chair and it became clear the student had already spoken to her. She asked me, with a smile, to consider restoring the points, in a way that made clear it wasn't a consideration. She asked me to meet with the student and then follow up with her afterward. I restored the points and met with the student. And I kept thinking about what I had missed. I would honor the processing space going forward. What I would hold them to was professional competence in assignments where it showed.
So I built one. I divided the class into five groups and gave each one a de-identified parent newsletter dense with errors, incomplete sentences, thoughts that started and stopped. I asked each group to evaluate the teacher who had written it. They were not kind. Then I showed them the originals, error free, as written. They read them differently. And then I told them both versions were mine, and that I had introduced every error deliberately into newsletters I had written.
The room went quiet. The student who had disagreed with me turned bright red.
I asked them what they thought now, about what a grammatical error signals, about who decides what professional competence looks like, and about what a family loses when the person communicating with them has been read too quickly.
In the follow-up meeting with the chair, I told her how it had gone. She had looked me up and down every time I walked into her office. Sometimes her eyes dropped to my belly. I was never sure what she was looking for. I had learned to read her face the way you learn to read any room where the terms of your belonging are not yet settled. And then I kept talking. I told her about the assignment I had developed for the class. Her eyes widened. Her eyebrows rose. There was a stillness before she sat back in her chair. I don't recall her ever asking what I had been trying to do. What I remember is the surprise on her face.
In the department work that followed, she asked me to write my contributions in complete sentences. This included feedback on CAEP assessments and revisions to certification materials. Write in complete sentences, she said, to the person she had asked to forgive a student for incomplete ones.
What the students needed before any of that could matter was time: to read, to discuss, to form their own conclusions about the person who wrote those newsletters. What I was building, without yet having language for it, was whatKemmis and colleagues (2014) call a practice architecture, the conditions that determine whether that space exists at all. And what I kept returning to, across different institutions and different courses, was the same question underneath it: what does it take to design an experience where a practitioner feels the gap between what she first perceived and what was actually there?
That question was still with me when I submitted a proposal as an assistant professor at Hunter College CUNY (2019) for graduate students in early childhood special education to visit an exhibition at the Cooper Hewitt Smithsonian Design Museum called Face Values: Exploring Artificial Intelligence. The exhibition invited visitors to examine their own faces through cameras and software, to sit with what emotional recognition technology read when it looked at a face, and what it missed. Not as a lesson about AI, but as a direct experience of a gap. What the technology perceived, and what was actually there. I wanted candidates to sit with that, and to let it open something about what they themselves bring into rooms where families are depending on them to see clearly.
The proposal was not funded.
Both experiences pointed toward the same question: what does it take to build the conditions for practitioners to see past their initial perceptions in professional preparation?In an earlier post in this series, I talked about searching for peer-reviewed literature on the use of AI in EI/ECSE. I found nothing. This is an opportunity, an opening, for the field to build AI into preparation across EI/ECSE/ECE contexts thoughtfully, responsibly, and responsively in ways that serve the children, families, and practitioners at the center of this work.
What both experiences had in common was the gap between what was visible on first reading and what was actually there. The newsletter assignment made students feel that gap before they could articulate it. The Cooper Hewitt proposal was designed to do the same thing through a different entry point. Both wereUDL in practice: multiple means of engagement, processing space built in before performance, different modes of knowing required before the insight could arrive. Practitioners in EI/ECSE/ECE carry that same gap into every family conversation, every hallway exchange with a specialist, every moment when a child's behavior requires reading past the first thing they see.
The GMU experience was itself an absence of practice architecture. Kemmis and colleagues (2014) describe practice architectures as the conditions of language, time, and power that make particular practices possible. In that department, those conditions were absent. Language was monitored, time for reflection was not built in, and the expectations shaping my role were never made explicit. And yet I was expected to create those conditions for my students. That contradiction is worth sitting with, because it is not uncommon in professional preparation across EI/ECSE/ECE.
NYC's Office of Technology and Innovation offers a useful guidepost for thinking about what AI means in that context: it is a machine-based system that produces predictions, recommendations, or decisions in response to human-defined objectives (NYC OTI, 2024). The practitioner defines the objective. What she brings to the interaction, how she frames her questions, what she examines, whether she traces responses back toDEC Recommended Practices,NAEYC position statements, or peer-reviewed research specific to the field, shapes what the tool can do. UDL makes that possible by designing the learning experience around multiple means of engagement, representation, and action and expression (CAST, 2018), so that the practitioner's capacity to ask better questions can develop.
Knowing how you relate is not the same as being able to see clearly in the moment, and for those of us whose brains are wired differently, the gap between the two can be particularly wide. That is what the newsletter assignment and the Cooper Hewitt proposal were both reaching toward. Both experiences asked practitioners to stay with a gap long enough to see past their first reading. AI, designed well and rooted in evidence-based practice, can extend the conditions for that kind of seeing beyond what any institution can provide in formal time and space. Preparation programs that integrate it responsibly, ethically, and with a responsive orientation can build practitioners who see past the first read.
References
CAST. (2018). Universal Design for Learning Guidelines version 2.2.http://udlguidelines.cast.org
Division for Early Childhood. (2014). DEC recommended practices in early intervention/early childhood special education.http://www.dec-sped.org/recommendedpractices
Gupta, S. S. (2026, April 1). AI and contemplative practice: An inquiry into professional learning infrastructure for early childhood educators. Ecological Learning Partners.https://ecologicallearningpartners.com/blog/ai-and-contemplative-practice-an-inquiry-into-professional-learning-infrastructure-for-early-childhood-educators
Gupta, S. S. (2026, March 29). Building processing space: Three assignments that apply UDL to early childhood teacher preparation. Ecological Learning Partners.https://ecologicallearningpartners.com/blog/building-processing-space-three-assignments-that-apply-udl-to-teacher-preparation
Gupta, S. S. (2026, February 19). When supporting families and supporting teachers aren't separate problems: 2-Care. Ecological Learning Partners.https://ecologicallearningpartners.com/blog/when-supporting-families-and-supporting-teachers-arent-separate-problems-2care
Gupta, S. S. (2026, February 3). Essential infrastructure: Designing practice architectures beyond verbal fluency. Ecological Learning Partners.https://ecologicallearningpartners.com/blog/essential-infrastructure-designing-practice-architectures-beyond-verbal-fluency
Kemmis, S., Wilkinson, J., Edwards-Groves, C., Hardy, I., Grootenboer, P., & Bristol, L. (2014). Changing practices, changing education. Springer.https://doi.org/10.1007/978-981-4560-47-4
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
Sarika S. Gupta, Ph.D., is a learning ecologist in early childhood and educational systems.
