Maps
Four ways to ask about emergence
Bayesian, dynamical, mechanistic, and semiotic approaches to language, artificial systems, and scientific practice.
How does a pattern arise from the processes that sustain it? I keep returning to this question across language, artificial systems, and scientific practice. The materials change, and so does the kind of explanation that seems useful.
This map holds four ways of asking together: what inference is being performed, how a process unfolds through time, what mechanism produces it, and what it means in its setting.
pattern arise?
The same orientation, across different materials
Trilingual minds
How might expectations across three languages shape the interpretation of an ambiguous word?
Neural networks
What does a predicted token distribution tell us about uncertainty, and what does it leave out?
Knowledge production
What is accepted as evidence, and how do competing claims gain or lose support?
Trilingual minds
How does competition between interpretations unfold during a multilingual response?
Neural networks
How do representations change during training, and when do new patterns become stable?
Knowledge production
How do routines and collaborations persist, change, or break down over time?
Trilingual minds
Which processes could produce the observed response, and what would distinguish their contributions?
Neural networks
Which features or circuits contribute to this output, and what changes when they are perturbed?
Knowledge production
Which instruments, procedures, and institutional arrangements enable this research practice?
Trilingual minds
How do language status, identity, and the experimental setting shape what a response means?
Neural networks
What makes a generated sequence meaningful, and whose interpretation establishes that meaning?
Knowledge production
How does a result become recognised as knowledge within a scientific community?
Select an orientation to read its question. Lines invite comparison; they do not establish a sequence or an equivalence.
How I use the map
The orientations give me different questions to put to the same material. A Bayesian account asks about inference under uncertainty. A dynamical account follows trajectories, stability, and change. A mechanistic account asks which parts and interactions produce the pattern. A semiotic account attends to signs, interpretation, and the practices in which something acquires meaning.
The connections in the diagram are prompts for comparison. They do not establish that the approaches are interchangeable, or that a successful explanation in one vocabulary automatically translates into another. That translation is part of the work.
Where it remains incomplete
An evolutionary question asks why this mechanism exists rather than another. An ecological question asks where the system’s boundaries should be drawn. Both can change the problem before any of the four orientations begins to answer it.
I keep the map provisional so these questions can alter its structure. It is a record of an orientation that can change as the work develops.