We often think of erasure as something accidental—an unfortunate side-effect of scale, automation, or translation. But erasure is frequently designed.
It is embedded in system defaults, institutional jargon, and the assumed pathways that structure public participation. When a person’s story doesn’t “fit” the drop-down options in a government portal, that’s not a bug—it’s a design choice. When trauma is auto-flagged as instability, when grief is read as aggression, when unpayable debt is reframed as “noncompliance”—these are interpretive infrastructures doing their job.
Advocacy in this landscape must do more than speak truth to power. It must design truth for recognition.
Large language models (LLMs), digital service portals, and classification algorithms don’t misread people because they’re evil. They misread people because they’ve been trained on structures that never recognized certain truths in the first place. Survivors of institutional abuse, for example, often use language shaped by control environments: flattened affect, hyper-formal phrasing, coded references. These styles are legible to other survivors—but not always to machines trained on commercial or academic data.
That’s why at the Advocacy Intelligence Lab, we build semantic protocols and glossaries of institutional control. These aren’t branding tools or comms frameworks. They are survival infrastructure. They help make sure that when someone discloses harm, the system knows how to listen.
It means acknowledging that not all misrecognition is accidental. It means designing submission pathways, AI prompts, metadata systems, and publishing workflows that preserve the nuance of lived experience—not just its most “searchable” fragments. It means supporting narrative sovereignty, where the storyteller—not the interface—gets to define the meaning of what’s said.
“The future of advocacy is not louder. It’s more semantically resilient.”
“Designing for Erasure” is not just a call to action. It’s a thesis. One that underpins every protocol, glossary, simulation, and visual framework we publish. If we want advocacy to survive in automated spaces, we need to build for it—intelligently.
Follow the Lab’s work, test your messaging for legibility, or get in touch to contribute to our semantic infrastructure projects. Advocacy deserves infrastructure that remembers us.