An interactive exhibition about memory and reconstruction. Three rooms, one memory — the factual record, the emotional register, the versions that conflict. The visitor chooses which one is theirs.
Gather
Before any making, the partnership was defined. Six levers shaped how AI would operate in this project: not as a generator, but as a configured collaborator.
A creative research partner and critical design collaborator. Rather than providing answers, it helped question assumptions about memory, identify tensions in my thinking, and develop concepts that could be translated into an exhibition experience.
To investigate how memories change over time and why those changes matter. AI helped uncover deeper questions about emotional truth, false memory, and personal identity — eventually leading to the design of an interactive exhibition.
Exploratory rather than definitive. Instead of presenting memory as a solved scientific problem, it was encouraged to acknowledge uncertainty, challenge assumptions, and keep multiple interpretations open.
When ideas became too broad, redirect toward specific people, moments, and experiences. Rather than proposing solutions immediately, ask questions that helped clarify the emotional tension behind the project.
Not allowed to finalize concepts too early or generate finished solutions before the underlying problem was understood. Could suggest directions, but needed to prioritize questioning, critique, and reflection over execution.
Flora, a 28-year-old art history graduate student, sits with a photograph of four people she used to know. The memory attached to the image shifts every time she returns to it. She cannot determine which version is closest to what actually happened — and she is beginning to wonder whether that question is the right one.
Your prompt
"What do you understand about the emotional core of this project?"
Claude
"You're assuming that distortion is a problem to solve — that false memory is the villain. But what if the project is actually about something else: the right to a memory that's emotionally true, even if factually off? That reframe changes what the exhibition does."
What it changed
It shifted the entire frame. I stopped trying to correct memory and started exploring reconstruction as meaning-making. Distortion was no longer a flaw to design around — it was the subject.
What was the question underneath your question? What made this worth a full quarter?
Flora feels uncertainty about whether the memories that shape her identity are reliable. She wants to trust her recollections, yet senses that they have been altered by emotion, distance, and time.
Flora repeatedly revisits photographs, messages, and keepsakes. She pauses over details, compares different recollections, and questions why the same memory seems to change each time she remembers it.
I am interested in memory reconstruction because memories shape our identities, yet they are constantly changing. I want to explore why something that may not be factually accurate can still feel emotionally true and meaningful.
Explore
Three sources shaped the direction. For each, AI helped surface what the source held beyond its surface argument.
TED Talk
How Reliable is Your Memory? — Elizabeth Loftus
The leading researcher on false memory demonstrates how memories are constructed, not recorded. Every time we remember something, we alter it slightly — and the alteration feels like the truth.
Podcast
Did That Really Happen? — Hidden Brain
Explores the malleability of memory through personal accounts and neuroscience, revealing how our emotional state at the moment of retrieval shapes what we recover — and what we lose.
Video Essay
Staring Contest — Tomcbumpz
A visual essay on obsessive memory loops — how we return to the same images hoping to extract certainty from something that has already changed in the looking.
Using Claude to synthesize these sources revealed a pattern that none of them named directly: memory is not about accuracy but about meaning. Loftus shows how memory fails factually; Hidden Brain shows how it succeeds emotionally; Staring Contest shows what it looks like to be caught in the loop. Together they point toward an exhibition that doesn't correct memory — it makes the reconstruction visible.
What it revealed
The three sources approach memory from different angles — scientific, emotional, obsessive — but they all land in the same place: reconstruction is inevitable, and it is also where meaning lives. The project needed to inhabit that space, not resolve it.
Narrow
This tension creates an opportunity to help people reflect on memory as an active process rather than a fixed record. Rather than treating distortion as a flaw, the design can frame reconstruction as a fundamental part of how identity is formed and reformed over time — making that process visible, inhabitable, and meaningful.
How might we design an experience that makes the process of memory reconstruction visible — not as a failure of accuracy but as a form of meaning-making?
How might we help someone sit with the uncertainty of an altered memory rather than trying to resolve it?
If every memory is already a reconstruction, is there a design space for intentional re-remembering — and what would it mean to author your own past?
Your prompt, from when you were still framing this as a false memory project
"My project might be about false memory, but I'm not sure that's the right frame anymore."
Claude
"You're not about false memory — you're about the right to a memory that's emotionally true, even if factually off. False memory implies error. What you're exploring implies something closer to authorship: who gets to shape the story of what happened?"
What it opened
I stopped designing for correction and started designing for the experience of reconstruction itself. The exhibition doesn't ask visitors to remember accurately — it asks them to choose which version they want to carry.
Flora sits alone with an old photograph of four friends. She can't remember which version of that night is true anymore — her memory, someone else's story, or the image in front of her. The question isn't which one is accurate. The question is which one is hers.
Midterm reflection
After presenting, feedback centered on two things: defining what the visitor actually does inside the exhibition, and determining whether reconstruction happens to them or through them. That distinction became the hinge of the final design.
What Remains is an interactive exhibition that makes the process of memory reconstruction visible. Visitors move through rooms that correspond to the same memory experienced differently — factual, emotional, contested — and are asked to choose which version they want to keep. The exhibition doesn't correct memory. It holds the space where reconstruction happens.
Implement
The first making move, before the loop tightened. Generate to understand, not to finish.
What the making showed
Generating early spatial images of the rooms showed me that the visual register needed to feel inhabited — not designed. The first pass looked too controlled, too gallery-clean. The breakthrough was letting the images feel uncertain: materials slightly out of focus, light diffused, edges soft.
Convergence
The ideation kept returning to one structural question: what does it feel like to move from one register of memory into another? That threshold — the moment between rooms — became the most important design decision in the project.
Generating early images of the three rooms helped test whether the visual language held across registers — factual, emotional, contested. The images gave a baseline for what each room needed to feel like before any spatial decisions were made.
Before
Flora before What Remains: the photograph, the uncertainty, the memory she cannot verify.
After
Flora after What Remains: the sealed card, the chosen version, the question she decided to live with.
Three versions. Each one sharpened the visual argument.
What I Learned
The three hero candidates had different emotional registers. The frosted glass image was the one that held the right tension: present but unreachable, something preserved that cannot quite be recovered. That became the visual anchor for all promotional materials.
Three rooms. Same memory. Factual record → Emotional register → Contested versions. The visitor chooses which one they carry. The architecture is the mechanism.
The second image (inputs and outputs) shows a relationship between sensory tech and focus that the others do not. However what is missing is the emotional and psychological relevance of the ring. Here it is simply systematic. It's missing why a ring is preferable to other tech and what the goal of a quiet UX design actually is.
The visual world every image must belong to. Not a style reference: a set of rules that every generation is measured against.
Strongest image from moodboard
Strongest image from moodboard
What Remains is an exhibition about memory, not an archive of it. It does not ask the visitor to remember correctly. It asks them to choose — which version they carry, which feels true even when it cannot be verified. Every image must belong to this world: aged paper, diffused light, objects slightly out of focus. Nothing is definitive. Nothing declares itself. The language is intimate and unresolved — it holds space for contradiction without explaining it away. Never show certainty. Never show resolution. Everything in the frame should feel like something held rather than something known.
This or That Exercise:
12 images across four registers: room atmosphere, material detail, visitor presence, and the conflict moment. What held: the frosted and diffused images maintained the right ambiguity. What broke: any image with clean edges or directional light felt too resolved — it gave answers where the seed needed to hold questions.
What changed
The seed needed more permission to be ambiguous. Every time I pushed toward clarity in the prompt, the output resolved something the exhibition needed to leave open. The revision was: remove all adjectives that describe a specific emotion. Let the material carry it.
Storytelling
Four beats. Who Flora is, the memory she carries, the moment she enters What Remains, and what she carries out.
Flora, 28. She studies photographs of events she wasn't at. She is comfortable with the idea that history is constructed. She is less comfortable with the idea that her own memory is too.
She holds a photograph of four friends from a night she remembers differently than the image suggests. She cannot tell which version is closest to what actually happened — and she is beginning to wonder if that question is the right one.
Three rooms. The factual record. The emotional register. The contested versions. Flora moves slowly. In Room 3 she encounters a version that matches neither the photograph nor her own account. She holds her answer card for a long time.
She seals the card. She has written: mine, even if it isn't right. She carries that out of the exhibition. The uncertainty is hers now — chosen rather than suffered.
I want the viewer to leave the exhibition feeling uncertain in a good way — carrying a version of someone else's memory and unsure whether it matters that it's not theirs.
What the video taught me
The video works best when it does not explain what the exhibition is. It should make the viewer want to stand inside it — not understand it from the outside.
Evaluate
The clearest discovery was about what AI cannot hold. It kept resolving the uncertainty that the exhibition needed to preserve. Every time I asked it to help me describe the experience, it would explain it — and explanation was exactly what the work was resisting. I had to learn to use it for structural pressure, not for voice.
AI is useful for stress-testing ideas. It is not useful for holding ambiguity. That part stays with the designer.
Week 1: Before
I thought memory was a subject. I didn't yet understand that it was a methodology. I came in wanting to make something about false memory — about correction, accuracy, the gap between what happened and what we recall. I had sources. I had a persona. I didn't have a position.
Week 10: Now
The project is not about false memory. It is about the right to a version of a memory that feels true — even without proof. That distinction took the entire quarter to find. AI helped me pressure-test it. The position was always mine.
Class Gallery
A breadcrumb trail across the process. Not polished. Real. Every image here is a decision point.
Initial brainstorming: lots of ideas swimming around.
Defining the visitor and the kind of memory they carry
SWOT analysis, narrowing down the scope.
First idea for cover images — still when set on making an app.
12 ideas for how the "app" could show up for people.
