You and I
A 2D serious game designed for in-home co-play between autistic children and their caregivers, introducing pronoun scaffolding for gestalt language processors.
- When
- Winter 2025 – June 2026
- Where
- WWU KIND Lab
- Role
- Lead researcher
- Built with
- Unity, C#, WebGL
Gestalt language processing and pronoun reversal
Most people acquire language analytically, learning to build longer sentences word by word. There is another way. Gestalt language processors begin with longer sentences and learn how to break them down and mix and match. Those longer sentences are memorized chunks, known as gestalt phrases, and they lead to scripting back phrases the child has heard verbatim. Instead of "I want a snack," a child might repeat "Do you want a snack?" because that is the script they heard.
Scripting like this often leads to pronoun reversal, where "you" is learned in place of "I" since that is how the child is referred to. Pronouns are hard on their own, and using them correctly takes more than grammar: it takes theory of mind, social cognition, and memory. When pronouns are misaligned it can lead to social isolation and reduced joint attention, and joint attention is itself foundational to learning pronouns.
Where that left us
Three separate systematic reviews from 2024 showed that existing language tools for autistic children target vocabulary and social skills, not gestalts or perspective-taking. There is work that introduced pronouns using a mirrored avatar to help introduce the "self" perspective, but they never tested it with children.
On joint media engagement, children get the most out of media when a caregiver engages alongside them, but that research focuses on turn-taking and social skills rather than language. On design guidelines, the literature gives useful principles: multimodal redundancy, familiar voices, simple age-appropriate language, and tapping rather than dragging as the default mechanic. But existing guidelines tie language to a child's age and ability. None of them speak to perspective-taking, or to the way gestalt language processors actually develop language.
The gap: no existing tools teach pronouns as deictic, perspective-shifting language for gestalt learners. That is what You and I does.
Introducing identity and language
The game is built for dual participation, so parent and child play together and joint media engagement stays at the center. Identity comes next. Self-insertion is fundamental to perspective-taking, so the child builds a customizable avatar and can even put their own face on it. To keep play error-free there is no gamified testing and no wrong answers, just modeling and practice. We attempted to scaffold language in a gestalt-style way by introducing pronoun concepts in isolated chunks on a scene-by-scene basis: you and I in the produce aisle, he and she in baking, then the full mix later on.
To introduce differing perspectives, the game centers on the main player and adds their two friends, Alice and Bob. The friends model pronouns between characters: "you" toward the player, "I" when a character means themself, "he" or "she" between the two. We surface the deictic shift by modeling that the same person is "you," "I," or "she/he" depending on who is speaking.
It is designed to be co-played, so the parent reads the dialogue aloud while the child navigates, which lets the parent scaffold perspective shifts as needed. When the child picks a wrong item, the feedback redirects them back into the story rather than marking them wrong.
The participatory evaluation
It was important to build trust with the families before the studies, so we asked if we could bring anything that would help their child feel more comfortable. Calculators, bubbles, stickers, you name it. That, paired with the pre-gameplay caregiver interviews, gave us the chance to build trust with the children and get to know them on a personal level before engaging in research. During one session I watched TV with a child while they decided whether they wanted to play the game we made. Another researcher played on the trampoline and explored the yard.
This ended up being the first phase of the framework we followed, the foundation phase from Co-Design Beyond Words, a framework for getting design insight from non-verbal and minimally speaking children. After the pre-gameplay interviews we trained the parents and showed them how to introduce You and I to their child. That led to the second phase, the interaction phase, captured multimodally with video cameras so we could observe micro-interactions between the parent and their child, turn taking, and moments where the parent scaffolded to help with perspective confusion. After gameplay we ran a child-facing survey with laminated visual aids and concluded with a post-interview with the parent.
We concluded the framework with the reflection phase. All the audio from the pre and post interviews was transcribed verbatim by the lead researcher. Video recordings were transcribed for verbal content by another researcher and re-coded for emotional engagement cues. The lead researcher applied three rounds of inductive coding. Themes were proposed to the research group on sticky notes and we built an affinity diagram, which we reviewed together.
Findings
We conducted in-home participatory design sessions with four families: three with autistic gestalt learners and one with Down syndrome. Data was collected through pre- and post-gameplay caregiver interviews, a post-gameplay child survey, and video recordings of gameplay sessions.
Our findings reveal that self-insertion, specifically mapping the player's own face onto their avatar, was central to resolving perspective confusion, with avatar customization alone falling short for some learners. Some gestalt learners benefit from atomic, phrase-level repetition targeting individual pronouns and verbs, rather than the scene-by-scene scaffolding the game proposed. Finally, we surface a tension between accessibility and learning: the design choices that reduced cognitive load for some learners enabled others to bypass pronoun content entirely.
Design recommendations
First, perspective-taking needs a visual anchor. Children were confused about who their character was, and the selfie feature resolved it. And do not assume children grasp the teaching convention itself, like what a speech bubble does.
Second, match language to the child's natural language acquisition level, and use pronoun slotting so it is predictable where pronouns appear, keeping the structure stable while the material varies.
Third, existing text-to-speech can affect how children pronounce words, so families prefer a familiar-sounding voice. But they cannot sustain recording those voices, so researchers should ship the game with pre-recorded voices to minimize that workload.
Fourth, the same accessibility scaffolds, tapping instead of dragging and interactive hints, can let some children skip the learning. Customization by ability is essential: toggle motor accommodations and difficulty per learner, not by age.
Method takeaways
Caregiver support groups were what actually yielded participants, and families screen research for its stance on neurodiversity. We recommend researchers state that up front, leading with a social model stance. Testing in real homes surfaced insights a lab would have hidden.
Feedback works best embedded in play. Only one of four children finished the post-session survey. We should have emphasized more of the Co-Design Beyond Words interaction components, to draw out microinteractions that would derive more meaning than a standardized questionnaire.
Gestalt language learners require a distinct design to support their learning. The manuscript is in preparation.