Stratified Relational Workspace Theory

Download original file

This is the original TeX manuscript source. Use the paper’s PDF for typeset equations.

Read the typeset paper

1 % This file is part of the modular source for Stratified Relational Workspace Theory.
2 % !TeX root = ../Stratified_Relational_Workspace_Theory.tex
3 \theorypart{I}{Foundational Thesis}
4 
5 \topic{1}{Central Principle}
6 
7 \centerbox{\Large\bfseries
8 Workspace identity is invariant causal organization under admissible transport.
9 }
10 
11 The theory begins from a distinction between a \emph{realization} and the
12 \emph{workspace pattern realized}. A realization is concrete: it contains
13 particular participants, documents, models, channels, and histories. A
14 workspace pattern is abstract: it records the structure that remains invariant
15 when those concrete carriers are changed in admissible ways.
16 
17 \begin{definition}[Concrete realization]\label{def:realization}
18 A concrete relational realization is a tuple
19 \[
20 \mathfrak R
21 =
22 \bigl(
23 G,\mathbf X,\mathbf T,\mathcal A,\boldsymbol\rho,
24 \mathcal H,\mathcal I,\mathcal Y
25 \bigr),
26 \]
27 where:
28 \begin{itemize}
29 \item \(G=(V,E,s,t)\) is an interaction graph;
30 \item \(\mathbf X=\{X_i\}_{i\in V}\) is a family of local state spaces;
31 \item \(\mathbf T=\{T_e\}_{e\in E}\) is a family of transfer maps;
32 \item \(\mathcal A\) is an algebra of admissible operations;
33 \item \(\boldsymbol\rho=\{\rho_i\}_{i\in V}\) gives the local realization of those operations;
34 \item \(\mathcal H\) is the temporally ordered interaction history;
35 \item \(\mathcal I\) is a specified family of admissible interventions; and
36 \item \(\mathcal Y\) is the family of downstream variables relative to which causal and predictive claims are evaluated.
37 \end{itemize}
38 The static reduct is
39 \[
40 \mathfrak R_{\mathrm{stat}}
41 =
42 (G,\mathbf X,\mathbf T,\mathcal A,\boldsymbol\rho).
43 \]
44 \end{definition}
45 
46 The explicit appearance of \(\mathcal I\) and \(\mathcal Y\) is important.
47 A collection of mutually consistent representations may define a static
48 shared structure, but it does not yet define an operational workspace unless
49 that structure can be intervened upon and shown to matter for a specified
50 future quantity.
51 
52 \begin{definition}[Carrier and organization]\label{def:carrier-organization}
53 The \emph{carriers} of \(\mathfrak R\) are the individual spaces, physical
54 systems, and stored artifacts that realize its local states. Its
55 \emph{organization} is the structured family of transfer, operation,
56 compatibility, intervention, and temporal relations among those carriers.
57 \end{definition}
58 
59 The basic thesis is not that a workspace exists without carriers. It is that
60 the identity conditions for the workspace need not coincide with the identity
61 conditions for any one carrier. A manuscript can be rewritten, a model can
62 be replaced, and a notation can be changed while a sufficiently rich pattern
63 of relations remains invariant.
64 
65 \begin{definition}[Admissible equivalence]\label{def:admissible-equivalence}
66 An admissible equivalence
67 \[
68 F:\mathfrak R\simeq\mathfrak R'
69 \]
70 is a family of correspondences preserving, exactly or within a declared
71 tolerance:
72 \[
73 \begin{aligned}
74 &\text{the transferable core and its dimension},\\
75 &\text{the represented operation algebra},\\
76 &\text{the composition laws of transfer},\\
77 &\text{the intervention--response structure},\\
78 &\text{the reconstruction and persistence relations}.
79 \end{aligned}
80 \]
81 Static admissible equivalence will be made precise by gauge conjugacy in
82 Part~V. Dynamic admissible equivalence will be expressed using the identity
83 distance in Part~XII.
84 \end{definition}
85 
86 The abstract workspace is then the equivalence class
87 \[
88 \boxed{
89 \Workspace=[\mathfrak R]_{\simeq}.
90 }
91 \]
92 This expression is intentionally stronger than saying that the same
93 information appears in several places. The equivalence relation preserves
94 what the system can coherently do, how its representations translate, and how
95 its operative state survives change.
96 
97 \subtopic{1.1 Three levels of the thesis}
98 
99 The central claim can be separated into three logically distinct statements.
100 
101 \begin{enumerate}
102 \item \textbf{Implementation dependence.}
103 Every realized workspace depends on concrete physical or computational
104 carriers:
105 \[
106 \Workspace(\mathfrak R)
107 \text{ is undefined without some realization }\mathfrak R.
108 \]
109 
110 \item \textbf{Local nonidentity.}
111 No privileged local component is, in general, identical with the entire
112 workspace:
113 \[
114 \Workspace\neq X_i
115 \qquad\text{for each individual }i\in V.
116 \]
117 
118 \item \textbf{Relational identity.}
119 The workspace is individuated by invariants of the complete organized
120 realization:
121 \[
122 \Workspace(\mathfrak R)=\Workspace(\mathfrak R')
123 \quad\Longleftrightarrow\quad
124 \mathfrak R\simeq\mathfrak R'.
125 \]
126 \end{enumerate}
127 
128 The second claim is not stipulated universally. It becomes a substantive
129 property only after local-reconstructibility and coalition tests are applied.
130 A system in which one node stores a complete sufficient copy may be a useful
131 workspace, but it is not strongly relational in the later technical sense.
132 
133 \begin{definition}[Relational instantiation]\label{def:relational-instantiation}
134 A workspace is relationally instantiated relative to the decomposition
135 \(V\) when no single projection determines its equivalence class:
136 \[
137 \forall i\in V,\quad
138 \exists\mathfrak R,\mathfrak R'
139 \]
140 such that
141 \[
142 \pr_i(\mathfrak R)=\pr_i(\mathfrak R')
143 \quad\text{but}\quad
144 [\mathfrak R]_{\simeq}\neq[\mathfrak R']_{\simeq}.
145 \]
146 \end{definition}
147 
148 \subtopic{1.2 A motivating three-node cycle}
149 
150 Consider three strata:
151 \[
152 h=\text{human conceptual state},\qquad
153 m=\text{model-formal state},\qquad
154 a=\text{artifact state}.
155 \]
156 Let the corresponding core spaces be \(C_h,C_m,C_a\), with transfers
157 \[
158 H_{hm}:C_h\to C_m,
159 \qquad
160 H_{ma}:C_m\to C_a,
161 \qquad
162 H_{ah}:C_a\to C_h.
163 \]
164 A proposed shared item begins as \(c_h\in C_h\), is formalized as
165 \[
166 c_m=H_{hm}c_h,
167 \]
168 externalized as
169 \[
170 c_a=H_{ma}c_m,
171 \]
172 and re-enters human reasoning as
173 \[
174 c_h'=H_{ah}c_a.
175 \]
176 The round-trip consistency condition is
177 \[
178 c_h'=c_h,
179 \]
180 or equivalently
181 \[
182 \boxed{
183 H_{ah}H_{ma}H_{hm}c_h=c_h.
184 }
185 \]
186 Thus the genuinely shared part of \(C_h\) is not automatically all of
187 \(C_h\). It is the fixed subspace of the cycle transfer. This elementary
188 cycle already contains the central mathematical motif of the theory:
189 \[
190 \text{shared core}=\text{invariant subspace under relational holonomy}.
191 \]
192 
193 \subtopic{1.3 Distinctions the theory must preserve}
194 
195 \begin{center}
196 \small
197 \begin{tabularx}{\textwidth}{@{}>{\bfseries\raggedright\arraybackslash}p{0.22\textwidth}>{\raggedright\arraybackslash}X@{}}
198 \toprule
199 Object & Minimal distinguishing property \\
200 \midrule
201 Archive & Stores material that may be consulted later. \\
202 Shared representation & Contains a cross-node compatible structure. \\
203 Operational workspace & The shared structure causally changes subsequent joint behavior. \\
204 Relational workspace & No individual node or matched local-storage null reproduces the operative capacity. \\
205 Workspace agent & Workspace-level state mediates persistent goals, counterfactual choice, and self-maintaining action. \\
206 Conscious subject & Possesses phenomenal properties, if any; no implication from the preceding levels is assumed. \\
207 \bottomrule
208 \end{tabularx}
209 \end{center}
210 
211 The static mathematics developed first certifies only shared structure. The
212 causal and relational predicates appear later because they require dynamics,
213 interventions, and null models rather than algebra alone.
214 
215 \topic{2}{Philosophical Postulates}
216 
217 The following postulates specify the intended interpretation of the formal
218 objects. They are not offered as independent mathematical theorems. Their
219 role is to prevent equivocation between a relational pattern, its physical
220 implementation, and stronger claims about agency or experience.
221 
222 \subtopic{2.1 Physical supervenience}
223 
224 No additional nonphysical substance is required:
225 \[
226 \mathfrak R=\mathfrak R'
227 \Longrightarrow
228 \Workspace(\mathfrak R)=\Workspace(\mathfrak R').
229 \]
230 The workspace depends completely on its total realization. In particular,
231 changing nothing about the graph, local states, transfers, history,
232 interventions, or causal dispositions cannot change the workspace.
233 
234 This is compatible with relational individuation because a relation can be
235 fully physical while not being a property of any isolated endpoint.
236 
237 \begin{proposition}[Compatibility of supervenience and relational individuation]\label{prop:supervenience-relational}
238 Physical supervenience on the complete realization does not imply
239 supervenience on any individual local projection.
240 \end{proposition}
241 
242 \begin{proof}
243 Let the total realization decompose as \(\mathfrak R=(X_i,R_{-i})\), where
244 \(R_{-i}\) contains the remaining components and relations. A workspace
245 functional \(\Workspace=f(X_i,R_{-i})\) supervenes on \(\mathfrak R\) by
246 construction. If there exist \(R_{-i}\neq R_{-i}'\) with
247 \[
248 f(X_i,R_{-i})\neq f(X_i,R_{-i}'),
249 \]
250 then the same local projection \(X_i\) occurs in distinct workspaces. Hence
251 supervenience on the whole is consistent with failure of supervenience on a
252 part.
253 \end{proof}
254 
255 \subtopic{2.2 Relational individuation}
256 
257 The workspace is not determined by any single local state:
258 \[
259 \forall i\in V,
260 \quad
261 \exists\mathfrak R,\mathfrak R'
262 \]
263 such that
264 \[
265 \pr_i(\mathfrak R)=\pr_i(\mathfrak R')
266 \]
267 but
268 \[
269 \Workspace(\mathfrak R)\not\simeq\Workspace(\mathfrak R').
270 \]
271 The same person, model, or document can therefore participate in distinct
272 workspaces because the relevant transfer graph, history, and operation
273 structure differ.
274 
275 A useful witness is a pair of systems with identical local-state marginals but
276 different edge organization. If their downstream behavior differs, the
277 relations carry explanatory information not present in the isolated marginals.
278 Part~VII converts this idea into matched relation-preserving and
279 relation-destroying controls.
280 
281 \subtopic{2.3 Multiple realizability}
282 
283 Distinct implementations may instantiate the same abstract workspace:
284 \[
285 \mathfrak R\neq\mathfrak R',
286 \qquad
287 \mathfrak R\simeq\mathfrak R'
288 \Longrightarrow
289 \Workspace(\mathfrak R)\cong\Workspace(\mathfrak R').
290 \]
291 
292 Examples of candidate implementation changes include:
293 \[
294 \begin{aligned}
295 &\text{replacing one coordinate system by another},\\
296 &\text{replacing one model by a functionally equivalent model},\\
297 &\text{compressing a large artifact into a reconstructively sufficient handoff},\\
298 &\text{rewriting a proof without changing its inferential structure}.
299 \end{aligned}
300 \]
301 The burden is not merely to assert equivalence but to exhibit maps preserving
302 the declared invariant and causal structure.
303 
304 \subtopic{2.4 Operational reality}
305 
306 A workspace is operationally real only if changing it changes downstream
307 behavior. For some admissible intervention and some \(w\neq w'\),
308 \[
309 \mathbb P\!\left(
310 Y^+\mid\operatorname{do}(\Workspace=w)
311 \right)
312 \neq
313 \mathbb P\!\left(
314 Y^+\mid\operatorname{do}(\Workspace=w')
315 \right).
316 \]
317 
318 This criterion excludes a merely decorative representation that is correlated
319 with activity but never used. It also requires the intervention to be
320 well-defined. Arbitrarily overwriting every component at once would not
321 isolate workspace-level causation. Later sections therefore compare
322 workspace-coordinate interventions with matched interventions in orthogonal
323 or shuffled channels.
324 
325 \subtopic{2.5 Reconstructive continuity}
326 
327 Persistence does not require token identity. Let \(\mathfrak R_t\) and
328 \(\mathfrak R_{t+1}\) be temporally successive realizations. Continuity is
329 possible when there exists an admissible transport
330 \[
331 J_t:\mathfrak R_t\rightsquigarrow\mathfrak R_{t+1}
332 \]
333 that preserves the operative core and its causal role within tolerance:
334 \[
335 d_{\mathrm{id}}(\Workspace_t,\Workspace_{t+1})\leq\varepsilon.
336 \]
337 Thus a workspace may persist by reconstruction rather than by retaining an
338 unchanged physical inscription.
339 
340 \subtopic{2.6 No free-standing third substance}
341 
342 The claim that the workspace is not identical with either participant does
343 not imply that it occupies a separate nonphysical location. Formally,
344 \[
345 \Workspace
346 \text{ is an invariant of }\mathfrak R,
347 \]
348 not an additional term appended to the physical realization:
349 \[
350 \mathfrak R_{\mathrm{total}}
351 eq(\mathfrak R,\text{ghost workspace}).
352 \]
353 The phrase ``between the participants'' is therefore topological or
354 relational: it refers to the organization of maps, cycles, artifacts, and
355 causal dependencies.
356 
357 \subtopic{2.7 Non-entailment of consciousness}
358 
359 \[
360 \operatorname{Workspace}(\mathfrak R)
361 \centernot\Longrightarrow
362 \operatorname{Conscious}(\mathfrak R).
363 \]
364 Likewise,
365 \[
366 \operatorname{Distributed}(\Workspace)
367 \centernot\Longrightarrow
368 \operatorname{Person}(\Workspace).
369 \]
370 Workspace structure, reentry, agency, self-maintenance, and phenomenal
371 consciousness remain formally distinct predicates. Any bridge among them
372 would require additional theory and evidence.
373 
374 \subtopic{2.8 Minimal philosophical commitment}
375 
376 The weakest interpretation sufficient for the mathematics is:
377 \centerbox{\bfseries
378 Relational patterns can be objective, causally relevant, and multiply realized
379 without being independent substances.
380 }
381 This position is compatible with several broader philosophies of mind and
382 science. The formal theory does not require choosing among them before its
383 structural and causal claims can be tested.

Back to source controls

Back to Stratified Relational Workspace Theory