Does it last?
Follow narrative agency over longer periods and across more varied multilingual communities.
Featured inquiry / 2025–2026Language · Childhood · Co-creation
Before thewordsarrive.
Trans-modal narrative agency in
multilingual children’s AI co-creation.
THE ANALYTICAL FRAMEWORK
How children
take authorship.
- 01DecodingReading an image into a story
- 02NamingConnecting a word to a concept
- 03InterpretingReasoning about a character
- 04AuthoringRevising the story’s meaning
CHILD ↔ TEACHER ↔ TOOL
Four analytical levels of authorial control, observed within a mediated setting.
Three children · Two rounds of sessionsRead the evidence ↗
01 / THE QUESTIONSeeing what language alone may miss.
What becomes visible
when children can
think beyond words?
A child may have a story in mind long before they can tell it fluently in a new language.
In my online teaching practice, I asked what forms of narrative agency appear when multilingual children co-create with generative AI—and what the child, teacher, and tool each bring to that setting.
This study follows the moments when children read, question, and reshape the stories offered to them.
02 / THE STUDYClose observation. Situated knowledge.
Inside the inquiry.
Two rounds of one-to-one online sessions,
September–November 2025.
PARTICIPANTSThree children.
Ages 6–12, each living across three languages and transnational contexts.
MATERIALSMore than speech.
Session video and transcripts, digital story artifacts, and reflective teaching memos.
ANALYSISIterative, close reading.
Open coding, a refined thematic codebook, and comparison across the three cases.
THE SETTING / STORYPAINTERA shared space
A shared space
for making meaning.
As StoryPainter’s product and learning experience designer, I worked with an engineering collaborator on a drawing-to-story tool. In the study setting, it responded to a child’s drawing with an interpreted image, a short narrative, and a read-aloud.
In this inquiry, I held overlapping roles as teacher, product collaborator, and researcher. My questions and the tool’s responses were both part of the interaction.
Visit StoryPainter ↗A shared
story
SELECT A ROLE TO EXPLORE THE INTERACTIONstory
The child interprets, questions, and revises what the story offers.
03 / WHAT I OBSERVEDFrom the sessions.
Four ways to
shape a story.
Four analytical levels of authorial control.
These are not developmental stages or a sequence.
Select a lens to explore an exchange.
Context: the tool generated an image of a dark cave.
TEACHER
“What do you think is happening in this picture?”
HENG
“They are in a scary tunnel! The bats are laughing!”
READING THE EXCHANGEAn image becomes
An image becomes
a first utterance.
Heng read a detail into the image and gave it narrative meaning. The analysis interprets his personification as an act of authorship supported by visual resources.
THE TEACHER’S ROLE
An open question invited a reading of the image, leaving room for meanings beyond the generated text.
ANALYTICAL FOCUS Image → narrative claim
Context: a lantern lights a dark scene.
HENG
“Glowing means the light comes out, like this, shiny and warm?”
TEACHER
“Yes! How did you figure that out?”
HENG
“Because the lantern is yellow and the dark goes away around it.”
READING THE EXCHANGEA word finds
A word finds
an anchor.
Heng proposed a meaning for “glowing” and explained his reasoning from the image. The study describes this connection as concept-to-L2 mapping.
THE TEACHER’S ROLE
Confirmation and a follow-up question made the child’s reasoning visible.
ANALYTICAL FOCUS Visual concept → new vocabulary
Context: a character turns away from a gift.
ELENA
“She is not just sad. She is disappointed. She wanted it and it did not happen.”
TEACHER · EXCERPT
“What in the picture shows disappointed and not only sad?”
ELENA
“She looks down, away from the gift. Sad would cry, but she just turns away.”
READING THE EXCHANGEReading a feeling.
Reading a feeling.
Explaining a mind.
Elena used visual evidence to distinguish between emotions and infer a character’s unfulfilled expectation.
THE TEACHER’S ROLE
A question about the picture encouraged Elena to explain the grounds for her interpretation.
ANALYTICAL FOCUS Visual evidence → emotional inference
Context: the generated story described a “bad witch.”
ASHLEY
“The story says she is a bad witch, but I don’t want that. She is just lonely, so people are scared of her.”
TEACHER · EXCERPT
“What happens to the story if she is lonely and not bad?”
ASHLEY
“Then someone can be her friend, and the ending is happy instead of scary.”
READING THE EXCHANGETaking the
Taking the
story back.
Ashley contested the tool’s familiar character type and offered her own interpretation. Her authorship became visible in the decision to revise what the system gave her.
THE TEACHER’S ROLE
A follow-up question invited her to develop the consequences of that change for the story.
ANALYTICAL FOCUS Generated premise → child’s revision
Excerpts from the writing sample; emphasis added for this page. Participant names are pseudonyms.Read the full analysis ↗
04 / WHAT THIS SUGGESTSA proposed construct.
Trans-modal
narrative agency.
The capacity to begin, negotiate, and revise story meaning across visual and verbal modes, ahead of fluency in the target language.
A construct proposed in this teacher-inquiry study.THE INTERPRETATION
In these sessions, children personified, inferred, and revised. Images and teacher dialogue helped make those moves visible.
This agency emerged in a mediated setting: the child brought ideas, the tool offered material, and the teacher invited interpretation and revision.
The same tool could also offer generic or culturally narrow defaults. The opportunity to question those defaults is central to this account.
05 / QUESTIONS AHEADWhat this inquiry opens up.
Read the study’s boundaries
- The teacher, product co-designer, and researcher were the same person, creating both access and a stake in the outcome.
- A teacher was present throughout; the data cannot separate the tool’s contribution from the teacher’s mediation.
- Three cases over two rounds cannot establish persistence, developmental progression, or gains in second-language fluency.
- Cultural standardization was observed but was not investigated systematically.
- The analysis followed movement between images and English, rather than systematically tracking home languages.
CONTINUE READING
The full writing sample.
Before the Words Arrive: Trans-Modal Narrative Agency in Multilingual Children’s AI Co-Creation
Read the paper ↗