I'm Kohen a User

Why a Designer Speaks About AI — and Proposes an Open-Source Framework

Category:
Personal Reflections
Published:

This image reflects the tone and underlying structure of the article.

From design practice to human–AI interfaces, and why I’m building “Mapping the Prompt (MTP).”

Preface

I’m Kohen (廣円, /ˈkoʊ.ən/), a designer by training and an independent researcher working on human–AI interfaces. My work explores the structural, cognitive, and cultural dimensions of AI systems—especially large language models (LLMs)—and how people actually engage with them in daily practice.

This essay explains why a designer talks about AI and why I’m proposing an open-source framework called Mapping the Prompt (MTP). It is a companion to my /about page and a concise introduction to what I do, why I do it, and how to collaborate.

The Context: Design in the Age of Generative AI

Generative AI now drafts visuals, UI scaffolds, and exploratory variations in seconds. Services like “instant first draft” tools are compressing the early phases of design work that once required hours of human exploration.

As the boundary between design and engineering blurs, roles must be re-defined. The question is no longer “Can AI make something?”—it’s “How should humans and AI think together?”

Designers are not merely arranging pixels. We are designing interfaces of engagement: the places where human intention meets model behavior. In that sense, AI has become a new environment for design—one that exposes assumptions about language, culture, cognition, and power.

A Different Angle: Applying Language and Culture to AI Design

A common research approach is to study LLMs as linguistic objects: test syntactic competence, probe universals, quantify biases. That work is essential—and I read it closely.

My angle is applied and interface-first:

  • Bring linguistic and cultural knowledge into the design of AI interactions.
  • Treat prompts as UI elements, not just text strings.
  • Make ambiguity visible and adjustable, rather than pretending it isn’t there.

Two recurring pain points motivate this:

  1. Language-internal ambiguity
    For Japanese, genitive chaining (e.g., “X の Y の Z…”) and honorifics can create layered, under-specified relations. A model may choose a plausible—but unintended—parse, shifting meaning or politeness in ways humans feel immediately.
  2. Culture and implicitness
    In high-context settings, much is left unsaid and supplied by shared background. Models trained on mixed-context corpora can oscillate in voice or stance (e.g., “expert assertiveness → sudden apology”), changing both content and impression.

Rather than fixing this post hoc with longer prompts, I aim to structure intention up front.

Introducing Mapping the Prompt (MTP)

MTP is an open-source framework for structuring ambiguous intentions in human–AI interaction.
It is not a new NLP algorithm. It’s a shared coordinate layer that sits next to the chat box and lets people specify, see, and adjust what they mean—beyond raw text.

Core ideas:

  • Shared coordinates for intention
    Users and models share a lightweight grid of intention nodes. Instead of writing a paragraph to “sound more exploratory but maintain factual caution,” one can adjust a small set of interpretable controls mapped to those tendencies.
  • Twenty nodes, two “faces”
    MTP currently uses a 20-node structure (two complements, “A-side” and “B-side”) to capture tensions like opening/closing, stillness/motion, stance/softening. These are design metaphors, not hard linguistic primitives—used to surface and tune intent.
  • Model-agnostic and UI-first
    Because it’s a UI layer, MTP remains model-agnostic. It can wrap different providers while respecting their safety constraints, offering a consistent way to talk about intention.
  • From text to touch
    Think of MTP as a prompt mixer you can see and manipulate. You still write text; the coordinates make your implicit levers explicit—and adjustable during the conversation.

GitHub: imkohenauser/mtp ↗

Why Open Source?

Open source is not a badge—it is the only practical way to:

  • Invite scrutiny of the coordinate design and UI affordances.
  • Let diverse communities (languages, professions, accessibility needs) adapt the framework.
  • Bridge research and practice: practitioners can instrument and iterate; researchers can formalize and test.

If intention is culturally situated, the framework that represents it must be collectively examined.

What I Study (and Build)

These themes anchor my writing, prototypes, and collaborations:

  • Language as cognitive interface
    How do grammar, register, politeness, and metaphor shape what a model “hears” and “returns”?
  • Prompt design as UI design
    Moving from “clever incantations” to structured controls, so intentions become inspectable and revisable.
  • Cross-cultural gaps
    Designing for Japanese–English (and beyond) requires understanding what is left implicit in each culture and how models interpolate.
  • Structural thinking for everyday AI use
    Small, composable structures (like MTP nodes) can help everyday users steer without writing treatises.
  • Practice meets evidence
    Observations from real interactions feed back into testable questions—toward future academic publication.

A more systematic overview of categories and audience is in /about.

Methods and Writing Style

I use a layered narrative:

  1. Observed friction (a misparse, a tone shift, a UI confusion)
  2. Analysis (linguistic, cognitive, design)
  3. Cultural framing (why this feels off to humans)
  4. Actionable structure (UI affordances, MTP nodes, prompt patterns)

The goal is not to mystify AI, but to make its handles visible.

From Blog Posts to Papers

Publishing code and essays is valuable, but scholarship is how ideas become portable knowledge. My plan:

  • Formalize MTP as a framework others can instrument.
  • Design small studies on Japanese prompt ambiguity and human perception of model stance/voice.
  • Cross-disciplinary articulation across HCI, design research, and cognitive/linguistic perspectives.

I do not claim the title “AI researcher” as an institutional role.
I am a designer and independent researcher, aspiring HCI / AI researcher, working toward publishable, reproducible contributions.

Current Projects

  • MTP (open source): core specification, UI patterns, and example integrations.
  • Japanese prompt challenges: characterization of genitive chaining, honorifics, and ellipsis in everyday tasks.
  • Design patterns for human–AI tone control: making stance, caution, and evidentiality controllable without verbose prompts.

Code and drafts live here:

Who This Is For

  • Developers and researchers interested in structured prompting, evaluation, and UI affordances.
  • Designers and writers shaping human–AI workflows beyond “prompt hacks.”
  • Educators teaching critical, cross-cultural AI literacy.
  • Teams building products where tone, stance, and cultural fit matter.

If you see your work in these lines, we likely have something to build together.

Collaboration

I welcome conversations and co-authorship across design, HCI, linguistics, and cognition.
If your team explores themes like metaphor in interaction, structural meaning, or cross-cultural alignment, let’s talk.

TL;DR

  • I’m a designer and independent researcher working on human–AI interfaces.
  • I build MTP, an open-source coordinate layer for intention that treats prompts as UI, not just text.
  • My focus: language/culture gaps, structured prompting, and everyday steering.
  • I’m working toward publishable HCI/AI research and invite collaborators.

Post Navigation

Search This Site