Zwicky
Last updated: Wednesday, 13 November 2024
An experiment in computer-supported morphological analysis, exploring how computational approaches might make systematic analysis feel more like structured play. Named for Bulgarian-born Swiss astronomer Fritz Zwicky, Zwicky seeks to bridge the gap between conversation and rigorous problem decomposition.
While traditional morphological analysis tools tend to emphasise rigid structure and fixed frameworks, this experiment explores how computational approaches might support more fluid, conversational engagement with complex problem spaces. How might we make morphological analysis feel more like structured play? The aim is to create tools that can flex between intuitive dialogue and rigorous decomposition, starting with simple command-line implementations and growing organically from there.
I’m interested in using open source LLMs to help extract dimensions from unstructured text, designing “dialogic” interfaces for exploring problem spaces, and experimenting with ways to make multi-dimensional navigation more intuitive. This could include generating “diegetic artifacts” (imagined objects or documents) from particular morphological combinations, or other experimental approaches to representation and interaction. Intent is to start small, with command-line tools and documentation-driven development.
Core focus
- Tools for morphological analysis that feel like structured play
- Supporting problem exploration over solution-hunting, encouraging users to engage with complex issues rather than rushing to a “best” solution
- Enabling gradual mastery through sustained use
- Experimenting with natural language approaches to dimension extraction1
Technical explorations2
- Developing computationally-efficient ways to represent multi-dimensional problem spaces and solution landscapes
- Local-first architecture, extensible framework design, emphasis on data formats and protocols
- Exploring more nuanced approaches to cross-consistency assessment, moving beyond binary feasibility to gradient-based or multi-criteria evaluations
- Experimenting with multimodal outputs, diagrams, and visualisations
Future possibilities
- Handling temporality and change within morphological frameworks
- Generating interactive scenarios from morphological combinations
- Supporting collaborative, multi-user exploration of problem spaces
-
Designing layered, adaptive documentation, capable of supporting users with varying levels of expertise
- [?] What’s the smallest useful implementation?
- [?] How might different interfaces (CLI, GUI, API) support different kinds of thinking? Which are best placed to support both free-form brainstorming and structured analysis within the same tool?
- [?] What existing tools could we build upon?
- [?] How can we manage the “combinatorial explosion” inherent in complex morphological analyses?
- [?] What are the implications of focusing on problem exploration rather than solution generation? How might this affect the tool’s utility in different contexts?
-
Exploring various approaches to dimension extraction, from simple text processing to more sophisticated dialogue-based interactions. The goal is to find natural breakpoints between free exploration and structured analysis. ⤴︎
-
Current technical direction: Exploring simple text processing tools and data formats that could support both structured analysis and playful exploration. Emphasis on extensibility over complexity. ⤴︎
Tags: activity
Backlinks