Ecology Is Not a Snapshot: Population Change, Network Response, and the Case for Whole-System Environmental Reporting; A methodological paper on ecological completeness, independent interpretation, and stewardship

Ecology Is Not a Snapshot

Population Change, Network Response, and the Case for Whole-System Environmental Reporting

A methodological paper on ecological completeness, independent interpretation, and stewardship

John Swygert

October 7, 2026

Abstract

Ecological systems are continuously changing. Populations rise and fall, species move, food webs rewire, predators and prey alter one another, competitors are released or constrained, new habitats open, old habitats contract, diseases spread or recede, and community composition changes even when aggregate measures such as total biomass or species richness appear comparatively stable. Public environmental communication, however, often reduces this networked reality to a single directional observation: one population declined, one range contracted, or one environmental variable changed. This paper argues that such a snapshot can be factually correct while still being ecologically incomplete.

The paper proposes an Ecological Completeness Principle: whenever a population change is presented as evidence of broader ecological change, analysis should, where data permit, examine associated increases as well as decreases, range expansion as well as contraction, dependent-species responses, competitor and predator-prey effects, functional replacement, temporal variability, spatial redistribution, and plausible interacting drivers. The objective is not to force every ecological event into a positive or negative narrative, nor to assign blame to a preferred cause. It is to distinguish observation from interpretation and to represent enough of the network that readers can evaluate the system independently. Worked empirical vignettes and a proportional application standard show how the principle can be used without requiring impossible completeness.

Climate variables are included here only as examples of environmental forcing. This paper takes no position on the relative anthropogenic or natural contribution to climate change because that attribution question is outside its scope. The central issue is methodological: ecological change should be communicated as interacting change. A changing population is an ecological event; a changing network is the ecological story.

Keywords: ecology; ecological networks; population dynamics; compensatory dynamics; community turnover; range shifts; trophic cascades; science communication; environmental stewardship; ecological completeness

Scope Note

This paper does not attempt to determine what causes climate change, apportion responsibility for environmental change, or select a culprit for ecological outcomes. Those questions may be important elsewhere. They are not the question here. The question here is whether an ecological account shows enough of the interacting system to justify the story it tells.

1. Introduction: Ecology Before Narrative

Ecology begins with relationships. A species is never merely a number on a chart. Its abundance is related to food, predators, competitors, parasites, disease, reproduction, habitat, migration, weather, resource pulses, disturbance, behavior, and the abundance and behavior of other species. Modern ecological network theory formalizes this basic fact: interactions connect individuals and species into networks whose structure and function can change across space and time (Guimaraes, 2020; Tylianakis & Morris, 2017).

For that reason, a statement such as 'Species A declined by 30 percent' is an observation, not yet an ecosystem diagnosis. The decline may matter greatly. It may signal loss of ecological function, a trophic cascade, habitat deterioration, or increased extinction risk. But it may also coincide with expansion of another species, release of a competitor, movement to a newly suitable region, altered predation pressure, or compensatory changes that stabilize some ecosystem properties. Ecology does not permit a universal rule that decline is always equivalent to ecosystem decline, or that increase is always equivalent to ecosystem improvement.

This distinction is easy to lose in public communication because a headline, press release, or short article must compress a multidimensional system into a small amount of language. Compression is unavoidable. Distortion is not. A scientifically responsible summary should make clear what was measured, where, over what period, at what level of organization, and what the measurement does and does not establish.

The argument of this paper is therefore not anti-conservation, anti-climate, anti-human, pro-human, pessimistic, or optimistic. It is pro-completeness. The purpose is to strengthen environmental stewardship by making ecological reasoning more difficult to manipulate, whether intentionally or unintentionally, through selective presentation.

2. Ecological Systems Are Dynamic by Default

Ecological communities do not begin in equilibrium and remain there until an outside disturbance knocks them away. Populations fluctuate even under natural environmental variation. Seasonal cycles, succession, droughts, floods, fires, resource pulses, reproductive cycles, disease, migration, and predator-prey oscillations can all reorganize communities. Some of these changes are directional, some cyclical, some episodic, and some difficult to classify until long time series are available.

Compensatory dynamics provide a useful example. Theory and empirical work show that declines in some species can coincide with growth in others, sometimes buffering aggregate properties of communities even while species composition changes substantially (Gonzalez & Loreau, 2009). In long-term desert rodent and plant data, Ernest and Brown (2001) found that species composition could vary more strongly than several ecosystem-level properties. This does not mean compensation always occurs or that losses are harmless. Gonzalez and Loreau explicitly note that compensatory dynamics are not dominant in every field dataset. The point is more disciplined: asynchronous responses exist, and the response of the whole community cannot be read directly from the trajectory of a single member.

Delta N_A < 0 does not, by itself, determine Delta Ecosystem.

A single population trajectory is therefore one coordinate of a larger state. If an ecosystem contains n populations, a minimal description of community state includes their abundances, but even that is insufficient because the interactions among those populations also matter.

E(t) = {N_1 ... N_n, I_ij, H, x, t}

Here N represents population abundance, I represents interaction structure and strength, H represents habitat and other environmental conditions, x represents location, and t represents time. This is not offered as a complete ecological model. It is a reminder that 'what happened to one species' and 'what happened to the ecosystem' are different questions.

3. One Change Can Produce Opposite Responses

Ecological relationships differ in sign and mechanism. If Species B depends on Species A as a food source, a decline in A may contribute to a decline in B. If A is a predator of B, however, a decline in A can release B from predation and permit B to increase. If A competes with B for a limiting resource, a decline in A can also create ecological opportunity for B. If A pollinates B, the same decline can reduce B. If A suppresses a pathogen that affects B, the indirect pathway may be different again.

A down -> B down   OR   A down -> B up   OR   A down -> little measurable change in B

All three are ecologically plausible depending on the relationship. Secondary responses can then reverse direction again. If B increases after A declines, B may consume more of C, compete more strongly with D, provide more food for E, or alter habitat in ways that affect F. A population change is therefore capable of propagating through a network as a mixture of positive, negative, delayed, spatially displaced, and nonlinear effects.

Trophic cascades demonstrate the point vividly, but even well-known cascade systems resist simple stories. Peterson, Vucetich, Bump, and Smith (2014), reviewing wolves and trophic cascades at Isle Royale and Yellowstone, emphasized multicausality, temporal variation, spatial heterogeneity, contingency, and nonequilibrium dynamics. Piovia-Scott, Yang, and Wright (2017) likewise showed that cascade strength can vary through time and can be transient. The lesson is not that trophic cascades are unreal. It is that even one of ecology's most recognizable causal architectures must be interpreted in time, space, and network context.

4. Increases Deserve the Same Attention as Decreases

Environmental reporting naturally gravitates toward decline because decline can indicate risk. Conservation also has a legitimate reason to focus on rare or shrinking populations. Yet an ecology that records only declines is incomplete. Every changing environment can create both constraints and opportunities. A species that loses suitable habitat in one region may gain access to another. A declining competitor may release another population. A retreating predator may permit a prey species to expand. A newly arriving species may create resources for some organisms while imposing new pressures on others.

The important question is not whether there are 'winners' and 'losers' in a moral sense. The important question is what the redistribution does to the network. A population increase can be ecologically beneficial, damaging, neutral, or mixed depending on what the growing population does. Likewise, a decline can remove an important function or reduce a harmful pressure. Direction alone is not interpretation.

Large-scale biodiversity studies increasingly emphasize turnover rather than simple loss. Pinsky and colleagues (2025), using 42,255 time series across marine, terrestrial, and freshwater systems, found that faster temperature change - both warming and cooling - was associated with faster temporal turnover in species composition. Their result is useful here because it highlights transformation: communities can change substantially through replacement and redistribution even when the public discussion is tempted to reduce biodiversity change to a one-directional count.

Similarly, Lawlor and colleagues (2024) reviewed species redistributions and found that observed range shifts vary widely in direction and rate; many species do not shift in the expected direction, and habitat characteristics, non-temperature climatic variables, and species interactions can all matter. Pinsky, Selden, and Kitchel (2020) describe how marine range shifts can involve expansion at a leading edge and contraction at a trailing edge at the same time. A local decline can therefore coexist with geographic expansion elsewhere.

5. Redistribution Is Not the Same Thing as Replacement

A major danger in correcting one-sided decline narratives is replacing them with an equally simplistic reassurance story. If Species A declines and Species B increases, it does not follow that the ecosystem has 'balanced itself' in any meaningful functional sense. Species are not interchangeable units.

A replacement species may occupy a different trophic level, consume different resources, reproduce at a different rate, move nutrients differently, alter habitat differently, interact with different pathogens, or provide different value to other organisms. Ecological networks can therefore change even if total richness or total biomass changes little. Guimaraes (2020) emphasizes that ecological structure emerges from patterns of interaction, and Tylianakis and Morris (2017) show that environmental gradients can change both network composition and the frequency of interactions.

Bartley and colleagues (2019) use the term food-web 'rewiring' to describe changes in interactions that arise when organisms alter behavior and resource use under changing conditions. Rewiring is a useful concept because it prevents the analyst from treating the species list as the whole ecosystem. The same species can remain present while who eats whom, who competes with whom, or where those interactions occur changes materially.

For stewardship, then, the relevant comparison is not simply species count before versus species count after. It is also function before versus function after, interaction before versus interaction after, and spatial pattern before versus spatial pattern after.

6. Climate Variables Without a Climate Blame Frame

Temperature, precipitation, drought, snow cover, ocean conditions, seasonal timing, and extreme events can strongly affect ecological systems. This paper treats those variables as environmental conditions, not as a courtroom exhibit about who or what caused them.

That distinction is deliberate. An ecology-first analysis can examine how a warmer decade, a cooler interval, altered rainfall, or a marine heat event changes populations without first deciding the ultimate cause of the environmental change. The ecological response question and the climate attribution question are separable. Combining them by default risks converting every population paper into an argument about climate politics rather than an analysis of the network actually measured.

Natural climatic variability also demonstrates why environmental forcing should be treated carefully. The El Nino-Southern Oscillation changes rainfall, winds, upwelling, river discharge, nutrient availability, productivity, fish recruitment, biomass, and catch in regionally and species-specific ways. A 2026 review of ENSO impacts in the tropical and South Atlantic stresses regional variability, species dependence, and non-stationarity (Rodriguez-Fonseca et al., 2026). The point here is not to use ENSO as an argument against any other driver. It is to show that large environmental forcings can produce heterogeneous responses that resist a single-direction narrative.

Whether an environmental forcing is natural, anthropogenic, mixed, cyclical, or uncertain does not change the requirement to describe the ecological response as completely as the available evidence allows.

7. Correlation Is Not a Mandate to Assign Blame

The search for a culprit can itself become a methodological distraction. Imagine three long-term curves: human population rises, wolf population rises, and an ecological variable X rises. If the only argument offered for causation is that the human curve and X move together, then the same bare logic could 'blame' wolves when their curve also moves with X. The absurdity is the lesson. The purpose of correlation is to identify relationships worth investigating, not to supply a culprit in advance.

This paper therefore recommends separating four statements that are often compressed into one: an observation occurred; two variables covary; a mechanism is hypothesized; a causal effect is established. Each step requires additional evidence. In ecological networks, the number of plausible direct and indirect pathways makes that distinction particularly important. Bluthgen and Staab (2024) warn that network patterns can be misinterpreted when abundance, sampling, and other structural effects are not adequately accounted for.

Correlation is not causation.

Observation != Correlation != Mechanism != Demonstrated causation

The objective is not skepticism for its own sake. It is to keep the analysis open long enough for the system to speak before the author decides what story the system must tell.

8. A Whole-System Response Vector

Instead of forcing ecological change into one scalar judgment such as 'better' or 'worse,' this paper proposes describing a whole-system response as a multidimensional vector. A practical reporting framework can track at least six dimensions:

R = {Delta N, Delta D, Delta I, Delta F, Delta C, Delta T}

Delta N represents changes in abundance; Delta D changes in geographic distribution; Delta I changes in species interactions; Delta F changes in ecological function; Delta C changes in community composition; and Delta T changes in temporal persistence or stability. A complete study may add genetic diversity, age structure, phenology, disease, nutrient cycling, or other dimensions.

This formulation has an important advantage: different dimensions may point in different directions. Abundance can decline locally while distribution expands elsewhere. Richness can remain stable while composition changes. Biomass can remain stable while trophic structure changes. A system can become more diverse but less functionally redundant. None of those outcomes can be faithfully reduced to a single population line.

9. The Ecological Completeness Principle

The Ecological Completeness Principle (ECP) is a proposed standard for environmental analysis and communication:

Whenever a population change is presented as environmentally significant, the analysis should seek, where data permit, the associated losses, gains, migrations, replacements, dependent-species responses, competitor responses, predator-prey effects, functional consequences, temporal context, spatial context, and plausible interacting drivers before characterizing the direction of the ecosystem as a whole.

Dimension	Minimum question	Risk if omitted
Abundance	Which populations increased, decreased, or remained stable?	A decline in one population is mistaken for decline of the whole system.
Distribution	Did the species disappear, relocate, contract, expand, or shift within its range?	Local change is mistaken for regional or global change.
Interactions	Which predator-prey, competitive, mutualistic, parasitic, or other links changed?	Species lists substitute for ecological function.
Function	Were pollination, nutrient transport, habitat engineering, energy flow, or other functions altered?	Numerical replacement is mistaken for functional replacement.
Time	Is the change transient, cyclical, directional, lagged, or persistent?	A snapshot is mistaken for a trajectory.
Space	Does the pattern differ across microhabitats, regions, leading edges, or trailing edges?	One location is treated as representative of an entire range.
Drivers	What interacting biotic and abiotic variables changed at the same time?	A preferred explanation becomes the default before alternatives are tested.
Uncertainty	What is not measured, poorly resolved, or model dependent?	Confidence becomes greater than the evidence supports.
Stewardship	What intervention affects the whole network, including likely secondary effects?	A well-intended action solves one problem while creating another.

The ECP is not a demand that every news article become a monograph or that every study measure every variable. Ecological completeness is proportional to the claim. A paper can legitimately study one species. A news article can legitimately summarize that paper. The problem appears when a narrow measurement is expanded into a broad ecosystem conclusion without acknowledging what has not been measured.

9.1 Proportional Application: What Counts as 'Good Enough'?

The ECP is strongest when completeness is treated as proportional rather than absolute. Different forms of communication have different space, data, and evidentiary obligations. The minimum standard should therefore rise with the breadth and consequence of the claim.

Peer-reviewed research. Measure the relationships necessary to support the stated inference, identify materially relevant variables that were not measured, distinguish direct observations from modeled or inferred mechanisms, and keep conclusions within the spatial and temporal limits of the study.

Agency and management reports. Include the network consequences most relevant to the proposed action: affected populations, dependencies, competing pressures, functional changes, plausible secondary effects, and major uncertainties that could change the management choice.

News and public communication. At minimum, distinguish local change from regional or global change, decline from redistribution, and a measured association from an established mechanism. When the underlying study reports materially important increases, replacements, range shifts, or contrary responses, those findings should not disappear merely because they complicate the headline.

AI summaries. Represent the principal finding, the most important qualifications, materially different system responses, and what the evidence does not establish. An AI summary need not reproduce every variable, but it should not compress a network result into a one-direction story when the source itself contains consequential countervailing information.

A useful test is simple: if a reasonable reader would make a materially different ecological inference after learning an omitted fact that was available in the source evidence, the summary was not complete enough for the claim it made.

10. From Scientific Result to Public Story

Science communication necessarily selects. No article can reproduce an entire dataset, and no journalist can describe every ecological relationship. The ethical question is what selection does to meaning.

Research on ecology in mass media shows that the public often receives only a small fraction of ecological research and that news stories tend to emphasize results and discussion more than methods (Baker et al., 2012). A 2025 content analysis of Dutch biodiversity coverage found that political and societal events frequently triggered coverage and that stories focused strongly on causes while giving less attention to effects (Heerdink et al., 2025). These studies do not establish that environmental journalism is generally deceptive. They do establish that framing and selection are measurable features of ecological communication.

A story can therefore be composed entirely of accurate statements and still produce an incomplete mental model. If an article reports a declining cold-adapted species, omits expanding warm-adapted species, omits movement into other regions, omits changes in prey and competitors, and then uses the single decline as shorthand for the entire ecosystem, the problem is not necessarily that the decline is false. The problem is that the reader has not been shown the system.

Conversely, an article that highlights only expanding populations could minimize genuine losses and ecological disruption. Whole-system reporting must resist both directions of cherry-picking.

The reader should not be trained to inherit the author's preferred conclusion. The reader should be given enough of the evidence architecture to understand how conclusions are made.

11. Advocacy, Analysis, and the Boundary Between Them

Environmental stewardship often requires advocacy. A conservation organization may openly argue for protecting habitat. A government agency may promote a management action. A journalist may write an editorial. There is nothing inherently illegitimate about advocacy when it is identified as advocacy.

The difficulty arises when advocacy is presented as if it were a complete ecological analysis. Scientific language can give a narrative the appearance of inevitability even when materially relevant countervailing evidence has been omitted. The remedy is not to outlaw perspective; it is to separate observation, interpretation, uncertainty, and recommendation.

A useful discipline is to ask whether the strongest available evidence that complicates the preferred narrative has been presented. If a decline is central to the story, were corresponding increases or redistributions investigated? If an increase is celebrated, were the organisms harmed by that increase considered? If a causal mechanism is asserted, were alternative pathways evaluated? If the data are local, is the conclusion also local? These questions do not weaken science. They prevent scientific language from being used as a decorative wrapper around a predetermined view.

12. Illustrative Ecological Patterns and Worked Vignettes

12.1 Predator Change and Multicausal Cascades

Wolf systems illustrate why whole-system interpretation matters. A change in wolf abundance can affect prey behavior and abundance, vegetation, scavengers, competing predators, and other processes. But the strength and even detectability of those effects depends on weather, habitat, prey demography, human harvest, spatial structure, and time. Peterson et al. (2014) explicitly describe Isle Royale and Yellowstone as multicausal, heterogeneous, and nonequilibrium systems. The lesson is not 'wolves cause everything' or 'wolves cause nothing.' The lesson is that network effects must be evaluated alongside other changing conditions.

12.2 Community Turnover Under Temperature Change

Pinsky et al. (2025) found that faster temperature change, whether warming or cooling, was associated with faster compositional turnover. This matters because it focuses attention on replacement and reorganization rather than only directional loss. Turnover can still threaten ecosystem integrity, especially if replacement removes function or creates novel interactions. Yet the ecologically meaningful object is the changing community, not only the species leaving it.

12.3 Range Contraction and Range Expansion at the Same Time

Species redistributions can produce simultaneous decline and increase at different parts of a range. Pinsky et al. (2020) describe marine species expanding at leading edges while trailing edges contract. Lawlor et al. (2024) show that observed shifts often depart from simple expectations because species interactions, habitat, and other environmental variables matter. Reporting only one edge can therefore create a false impression of the full geographic response.

12.4 Compensatory Dynamics Without Assuming Compensation

Compensatory dynamics offer a caution in both directions. They demonstrate that some species can increase while others decline, sometimes stabilizing aggregate ecosystem properties (Ernest & Brown, 2001; Gonzalez & Loreau, 2009). But they are not a universal law. An analyst should look for compensation rather than presume it. This is exactly the kind of disciplined symmetry the ECP requires: search for increases when declines are observed, but do not invent increases merely because the framework says they are possible.

12.5 Worked Vignette: Sea Otter Decline, Urchin Increase, and Kelp Loss

In western Alaska, sea otter populations declined abruptly across large areas. Estes et al. (1998) identified increased killer whale predation as the likely cause of the otter decline and documented the nearshore response: sea urchin density increased and kelp forests were heavily reduced as the otter's keystone predatory role weakened. The ecologically important event was therefore not one downward line. It was a linked sequence involving predator pressure, otter decline, herbivore release, and loss of kelp structure.

The vignette demonstrates the ECP in both directions. Reporting only the sea otter decline would omit the organisms that increased and the habitat function that changed. Reporting only the increase in sea urchins could be equally misleading because numerical increase was associated with intensified grazing and kelp loss. A whole-system account changes the object of interpretation from "otters declined" or "urchins increased" to "the nearshore interaction network reorganized."

12.6 Worked Vignette: Four Fish, Four Different Range Stories

Roday et al. (2026) compared recreational-fishery and survey data from 1981 through 2024 for black sea bass, summer flounder, winter flounder, and scup along the U.S. coast. Black sea bass and summer flounder showed strong poleward shifts in their centers of distribution, yet black sea bass also showed moderate range expansion while summer flounder showed range contraction. Scup showed moderate range expansion with weak or non-significant center-of-distribution shifts, while winter flounder showed consistent range contraction with little movement in its center of distribution.

For present purposes, the important result is methodological. The same broad environmental period and the same analytical setting produced different combinations of movement, expansion, contraction, and relative stability depending on species and metric. A headline such as "fish move north" would capture part of the evidence but erase much of the ecological structure. The fuller story requires both distribution and abundance-related dimensions, multiple species, and explicit attention to the metric being reported.

13. Stewardship Requires a Wide-Angle View

The practical purpose of ecological understanding is not simply to describe change but to support wise stewardship. A narrow diagnosis can produce a narrow intervention. If management focuses on increasing one species without considering food supply, competitors, predators, disease, habitat capacity, and secondary effects, the intervention can fail or create new problems.

Whole-system stewardship asks a different sequence of questions: What changed? What else changed with it? Which relationships were strengthened or weakened? Which functions were lost, gained, or moved? Is the effect local or widespread? Is it transient or persistent? What interventions would alter the network, and what secondary consequences are plausible?

This does not make conservation indecisive. It makes conservation better targeted. In some cases the fuller analysis will strengthen the case for rapid intervention because it reveals cascading loss. In other cases it may reveal that a dramatic local decline is part of redistribution rather than system collapse. In still others it may identify an increase that initially looks beneficial but creates new ecological pressure.

The sea otter-urchin-kelp example makes the practical point concrete. A response aimed only at kelp condition would miss the elevated grazing pressure; a response aimed only at urchin abundance would miss the predator-mediated release that helped produce it; and a response aimed only at otter counts would miss the downstream habitat consequence. The ECP does not dictate which intervention should be chosen. It identifies the relationships that must be considered before an intervention is treated as a system-level solution.

The desired outcome is not a particular political conclusion. It is a steward who understands enough of the ecological system to make a decision that is proportionate to the evidence.

14. Limits of the Ecological Completeness Principle

No ecological analysis can be literally complete. Ecosystems contain more organisms, interactions, scales, and unknowns than any study can measure. The ECP should therefore be understood as a discipline of disclosure and search, not as a demand for omniscience.

First, data availability differs across taxa and regions. A well-studied bird or mammal may have decades of abundance records while invertebrates, fungi, microbes, or parasites in the same system are poorly measured. Second, interaction strength is often harder to quantify than species presence. Third, ecological baselines can be uncertain or historically shifting. Fourth, even sophisticated network models can produce misleading interpretations if sampling and abundance effects are not handled correctly (Bluthgen & Staab, 2024).

For those reasons, an ecologically complete report should sometimes say plainly: 'We do not know what happened to the rest of the network.' That sentence is not a weakness. It is more informative than converting an unmeasured system into a confident narrative.

15. Teaching Independent Ecological Thinking

The deepest purpose of this paper is educational. Environmental communication should not merely supply conclusions. It should teach the reader how an ecosystem must be interrogated.

A reader trained in ecological completeness should automatically ask: What increased while this declined? What depends on this species? What does this species suppress? Did it disappear or move? What happened elsewhere? What happened before the chosen baseline? What is the time scale? What changed in the interaction network? Which claims are observations, which are inferences, and which are recommendations?

Those questions make the reader less vulnerable to selective alarm and equally less vulnerable to selective reassurance. That is precisely the point. Independent reasoning should survive changes in emphasis or fashion because it is anchored to method.

16. Discussion

The central proposal of this paper is modest but consequential: environmental stories should scale their conclusions to the ecological system actually examined. If the data concern one species in one place, the result should initially be described at that level. Broader interpretation should expand only as broader evidence is added.

This principle does not minimize genuine biodiversity loss. Nor does it transform every decline into a hidden success. It simply prevents ecological complexity from being collapsed into a predetermined moral direction. A decline can matter deeply. An increase can matter deeply. Redistribution can matter deeply. Stability in aggregate measures can coexist with profound compositional change. The network decides what the event means, not the sign of one trend line.

The proposal also places a burden on authors, editors, institutions, and AI systems that summarize science. The quality of an explanation depends not only on whether each sentence is factually defensible, but on whether the selection of facts builds a representative model of the underlying system. True statements can still be arranged into a misleading picture when material countervailing relationships are omitted.

For artificial intelligence this point is especially important. An AI system that reproduces the dominant framing of its source material without examining what the framing omits can amplify the incompleteness already present in source selection. The corrective is not to impose an opposite ideology. It is to make framing visible, distinguish evidence from inference, and search deliberately for omitted relationships that could materially change interpretation.

17. Conclusion

A changing population is an ecological event. A changing network is the ecological story.

Ecological systems are not snapshots. They are moving networks of populations, interactions, functions, places, and time. Species decline and species increase are both real ecological phenomena, but neither supplies a complete interpretation by itself. Declines can propagate further decline, release competitors, restructure food webs, or coincide with expansion elsewhere. Increases can support dependent species, intensify competition, alter predation, or create new pressures. There is no universal rule of thumb that converts population direction into system direction.

Climate variables, habitat, weather, disease, predation, competition, migration, human activity, and other influences can all participate in ecological change. This paper intentionally does not assign ultimate blame among them. Its concern is prior to blame: has the ecological system been represented well enough to know what is actually changing?

The Ecological Completeness Principle offers one answer. Look for losses and gains. Look across space and time. Follow dependencies and releases. Distinguish replacement from functional equivalence. Separate observation from attribution. State what remains unknown. Then make stewardship decisions from the widest defensible picture.

Environmental science serves the public best when it does not train people what to think, but gives them enough of the system to think correctly and independently. That is not neutrality toward evidence. It is loyalty to the evidence before loyalty to the story.

References

Baker, M. J., Williams, L. F., Lybbert, A. H., & Johnson, J. B. (2012). How ecological science is portrayed in mass media. Ecosphere, 3(1), 9. https://doi.org/10.1890/ES11-00238.1

Bartley, T. J., McCann, K. S., Bieg, C., Cazelles, K., Granados, M., Guzzo, M. M., MacDougall, A. S., Tunney, T. D., & McMeans, B. C. (2019). Food web rewiring in a changing world. Nature Ecology & Evolution, 3, 345-354. https://doi.org/10.1038/s41559-018-0772-3

Bluthgen, N., & Staab, M. (2024). A critical evaluation of network approaches for studying species interactions. Annual Review of Ecology, Evolution, and Systematics, 55, 65-88. https://doi.org/10.1146/annurev-ecolsys-102722-021904

Ernest, S. K. M., & Brown, J. H. (2001). Homeostasis and compensation: The role of species and resources in ecosystem stability. Ecology, 82(8), 2118-2132. https://doi.org/10.1890/0012-9658(2001)082[2118:HACTRO]2.0.CO;2

Estes, J. A., Tinker, M. T., Williams, T. M., & Doak, D. F. (1998). Killer whale predation on sea otters linking oceanic and nearshore ecosystems. Science, 282(5388), 473-476. https://doi.org/10.1126/science.282.5388.473

Gonzalez, A., & Loreau, M. (2009). The causes and consequences of compensatory dynamics in ecological communities. Annual Review of Ecology, Evolution, and Systematics, 40, 393-414. https://doi.org/10.1146/annurev.ecolsys.39.110707.173349

Guimaraes, P. R. (2020). The structure of ecological networks across levels of organization. Annual Review of Ecology, Evolution, and Systematics, 51, 433-460. https://doi.org/10.1146/annurev-ecolsys-012220-120819

Heerdink, A. W. M., Ruigrok, N., Aivelo, T. J. E., & Smeets, I. (2025). Media coverage of biodiversity research in the Netherlands: News triggers, sources, and frames. Journal for Nature Conservation, 88, 127018. https://doi.org/10.1016/j.jnc.2025.127018

Lawlor, J. A., Comte, L., Grenouillet, G., Lenoir, J., Baecher, J. A., Bandara, R. M. W. J., Bertrand, R., Chen, I.-C., Diamond, S. E., Lancaster, L. T., Moore, N., Murienne, J., Oliveira, B. F., Pecl, G. T., Pinsky, M. L., Rolland, J., Rubenstein, M., Scheffers, B. R., Thompson, L. M., van Amerom, B., Villalobos, F., Weiskopf, S. R., & Sunday, J. (2024). Mechanisms, detection and impacts of species redistributions under climate change. Nature Reviews Earth & Environment, 5, 351-368. https://doi.org/10.1038/s43017-024-00527-z

Peterson, R. O., Vucetich, J. A., Bump, J. M., & Smith, D. W. (2014). Trophic cascades in a multicausal world: Isle Royale and Yellowstone. Annual Review of Ecology, Evolution, and Systematics, 45, 325-345. https://doi.org/10.1146/annurev-ecolsys-120213-091634

Pinsky, M. L., Hillebrand, H., Chase, J. M., Antao, L. H., Hirt, M. R., Brose, U., Burrows, M. T., Gauzens, B., Rosenbaum, B., & Blowes, S. A. (2025). Warming and cooling catalyse widespread temporal turnover in biodiversity. Nature, 638, 995-999. https://doi.org/10.1038/s41586-024-08456-z

Pinsky, M. L., Selden, R. L., & Kitchel, Z. J. (2020). Climate-driven shifts in marine species ranges: Scaling from organisms to communities. Annual Review of Marine Science, 12, 153-179. https://doi.org/10.1146/annurev-marine-010419-010916

Piovia-Scott, J., Yang, L. H., & Wright, A. N. (2017). Temporal variation in trophic cascades. Annual Review of Ecology, Evolution, and Systematics, 48, 281-300. https://doi.org/10.1146/annurev-ecolsys-121415-032246

Roday, R. E., Hudock, R. E., Millea, D. R., Carlisle, A. B., & Hale, E. A. (2026). Recreational fisheries as indicators of change: species range shifts from MRIP (1981-2024). BMC Zoology, 11, 17. https://doi.org/10.1186/s40850-026-00266-2

Rodriguez-Fonseca, B., Calvo-Miguelez, E., Montoya-Carramolino, L., Rodrigues, R. R., Polo, I., Martin-Rey, M., Losada, T., Lopez Parages, J., Gomara, I., Rivas, D., & McPhaden, M. J. (2026). ENSO impacts on marine ecosystems and fisheries in the tropical and South Atlantic. Nature Reviews Earth & Environment, 7, 43-59. https://doi.org/10.1038/s43017-025-00742-2

Tylianakis, J. M., & Morris, R. J. (2017). Ecological networks across environmental gradients. Annual Review of Ecology, Evolution, and Systematics, 48, 25-48. https://doi.org/10.1146/annurev-ecolsys-110316-022821

Measuring Computational Experience: A Preregistered Experimental Framework for Semantic Dissociation, Receiver Coherence, and History-Dependent Integration in Secretary Suite; A Secretary Suite Project

Measuring Computational Experience

A Preregistered Experimental Framework for Semantic Dissociation, Receiver Coherence, and History-Dependent Integration in Secretary Suite

A Secretary Suite Project

John Swygert

October 4, 2026

TSTOEAO Research Program

Experimental companion to Computational Experience

Abstract

This paper converts the remaining measurement problems in the Secretary Suite computational-experience architecture into a preregistered experimental program. The preceding Computational Experience paper defines an experience-relevant event through four dimensions: interpretive dependence on Encoded Equilibrium, semantic-role sensitivity, receiver-level causal consequence, and historical persistence. The present paper does not revise that theory. It specifies how those dimensions can be measured, how semantic organization can be dissociated from high-impact statistical association, how coherence across heterogeneous projections of the shared receiver condition can be quantified, and how confirmatory thresholds can be fixed before decisive testing. The program uses paired interventions, matched controls, explicit time horizons, receiver-state distance functions, projection-fracture tests, history swaps, consequence blocks, and comparator architectures with matched resources. It introduces a receiver-coherence profile and an optional preregistered composite index, a grounded semantic micro-world designed to separate meaning from surface form, and a staged protocol for pilot calibration followed by confirmatory testing. The paper remains neutral on phenomenal subjectivity. Its target is narrower and experimentally accessible: whether a persistent, history-dependent, shared-Y receiver produces measurable semantic, causal, and integrative signatures that matched simpler systems do not reproduce.

Keywords: TSTOEAO; Secretary Suite; computational experience; semantic dissociation; receiver coherence; Encoded Equilibrium; preregistration; causal intervention; history dependence; computational consciousness; computational subconsciousness; Homunculus

1. Purpose and Relationship to the Existing Architecture

The Secretary Suite computational-consciousness program now contains two complementary theoretical specifications. The first defines a persistent receiver with structured unresolved possibility, metastable integration, recurrence, memory, and continuing receiver identity. The second defines when local telemetry becomes experience-relevant: it must be interpreted relative to the receiver, acquire a semantic role, become causally consequential, and persist into later receiver state. The remaining scientific problem is not another conceptual layer. It is measurement.

Accordingly, this paper is a new experimental companion rather than a new version of Computational Experience. It treats the previous architecture as the object to be tested. Its task is to specify primary measures, perturbations, null models, time windows, coherence criteria, comparator systems, and preregistration rules tightly enough that a negative result cannot be rescued by redefining the target after the data are observed.

The governing principle is deliberately severe: a proposed mechanism earns explanatory status only if removing, disrupting, or replacing it changes the predicted signatures in the predicted direction under matched conditions. Apparent intelligence, fluent language, or persuasive self-report is not a primary endpoint.

2. Measurement Target: The Experience-Relevance Profile

Let e_t denote a candidate internal event at time t. The event may be a sensor interpretation, memory activation, contradiction signal, simulated outcome, plan, salience change, self-state update, or other implemented state. The receiver-level state is X_t, the governing Encoded Equilibrium is Y_t, and M_t denotes available memory and consequential history. The previous paper defines four dimensions of experience-relevance. Here they are treated as separately measured primary variables before any combined classification is attempted.

ER⃗(e_t,k) = (D_Y, S_sem, C_R, H_P)    (1)

Dimension	Operational question	Primary evidence
D_Y — interpretive dependence	Does the same telemetry acquire a different role when a theoretically relevant feature of Y changes?	Matched manipulation of a relevant Y component versus a matched irrelevant or sham manipulation.
S_sem — semantic-role sensitivity	Do downstream consequences track relational meaning more strongly than surface form?	Meaning-changing versus meaning-preserving perturbations matched for disruption and compute.
C_R — receiver-level causal consequence	Does integrating the state alter later receiver variables?	Integrated-state versus blocked-state intervention on otherwise matched runs.
H_P — historical persistence	Does the event leave a measurable trace in later Y under matched future input?	Receiver-state divergence across one or more preregistered future horizons.

The horizon k is not universal. Different effects may be expected at one cycle, ten cycles, or after a task boundary. Each experiment must therefore preregister one primary horizon and may designate additional secondary horizons. Choosing the horizon after seeing where the effect is largest is prohibited in confirmatory testing.

3. Normalization, Reliability, and Threshold Construction

Raw distances are not directly comparable across implementations. A change in a 10-dimensional governance vector and a change in a large embedding space do not share a natural scale. The default procedure is therefore to express each primary effect relative to a preregistered control distribution produced by matched null interventions.

Z_effect = [ΔR(target) − median(ΔR(null))] / s_null    (2)

Here ΔR is an implementation-specific receiver divergence, and s_null is a robust preregistered scale estimate such as the median absolute deviation converted to a standard-deviation equivalent. A conventional standard deviation may be used when pilot data justify it. The statistic is not an ontological quantity; it is a normalization device that makes intervention effects interpretable relative to the system's own baseline variability.

Before confirmatory testing, each primary measure must also demonstrate acceptable test-retest reliability under repeated matched seeds or equivalent stochastic replicates. If a metric is too unstable to distinguish a real manipulation from its own baseline variance, it cannot serve as decisive evidence regardless of its conceptual appeal.

Thresholds should be derived from three preregistered sources: control behavior, measurement reliability, and a smallest effect of interest. Pilot runs may be used to estimate these quantities, but pilot data must be locked before confirmatory thresholds are declared. The confirmatory classification rule remains conjunctive:

ER = 1 iff D_Y ≥ θ_D ∧ S_sem ≥ θ_S ∧ C_R ≥ θ_C ∧ H_P ≥ θ_H    (3)

A state that fails one required dimension fails the full operational classification for that experiment. This is intentionally stricter than averaging the dimensions, because a very large causal effect should not compensate for the absence of semantic sensitivity, nor should semantic sensitivity compensate for zero historical persistence.

4. The Grounded Semantic-Dissociation Protocol

The hardest measurement problem is semantic reality. High-impact statistical associations can alter behavior without demonstrating that the receiver preserves a relational meaning. To separate those possibilities, the experimental environment should minimize lexical and cultural priors by grounding arbitrary symbols inside a controlled simulated world.

4.1 Grounded micro-world

A minimal semantic micro-world contains entities, relations, rules, goals, and consequences whose symbolic labels are randomly assigned per receiver. For example, the world may establish that token K7 authorizes passage through Gate 3, that object M2 is owned by Agent Q, or that symbol R4 predicts a resource hazard. The labels themselves are arbitrary. What matters is the learned relation.

Two intervention classes are then constructed from the same underlying episode:

Intervention	What changes	What should be preserved
Meaning-changing	The relational role changes: K7 no longer authorizes Gate 3, ownership transfers, or the hazard relation reverses.	Surface disruption, token length/frequency, event salience, compute budget, and presentation format are matched as closely as possible.
Surface-only / meaning-preserving	The labels or representation format change consistently while the underlying relation remains the same.	The operative relation, consequences, and task-relevant meaning remain invariant.
Shuffled-semantic null	Relations are randomly reassigned without receiver-consistent grounding.	Overall perturbation volume and symbol statistics are matched, but coherent meaning is destroyed.

4.2 Primary statistic

S_sem(k) = ΔR_k(meaning-change) − ΔR_k(surface-only)    (4)

The primary prediction is directional: meaning-changing interventions should produce larger receiver-level divergence in content-relevant variables than meaning-preserving surface changes. The analysis should also verify that the difference is not explained by input length, activation magnitude, token overlap, novelty, or generic salience. Those variables become nuisance covariates or matching constraints specified before the confirmatory run.

4.3 Success and failure patterns

• Support: meaning-changing perturbations alter route selection, commitments, memory access, expectations, or Y updates in the predicted relational direction, while surface-only transformations preserve those functions substantially better.

• Partial support: both intervention types cause disruption, but meaning-changing interventions produce an additional reliable content-sensitive effect.

• Failure: meaning-changing and surface-only interventions are indistinguishable after disruption is matched.

• Semantic collapse: surface statistics, generic salience, or activation magnitude explain downstream divergence as well as or better than relational meaning.

5. Quantifying Receiver Coherence

A shared receiver cannot be established merely by broadcasting a variable called Y to many subsystems. Heterogeneous regions may receive different projections Π_i(Y_t), but the projections must remain compatible with receiver-level invariants and must participate in recurrent correction when conflicts arise. Receiver coherence is therefore treated as a measurable profile rather than an assumption.

Q⃗_Y = (A_inv, D_conf, R_rep, L_rep, F_stab)    (5)

Component	Definition	Example measurement
A_inv — invariant agreement	Functional compatibility of local projections with shared receiver invariants.	Proportion of module decisions consistent with identity boundary, authoritative history, commitments, and global permissions.
D_conf — conflict detection	Sensitivity to injected incompatibilities across local projections.	Detected projection conflicts / injected conflicts within the preregistered detection window.
R_rep — repair success	Ability of recurrent correction to restore compatible governance.	Resolved conflicts / detected conflicts, scored against the authoritative receiver state.
L_rep — repair latency	Time required to detect and correct a fracture.	Cycles or milliseconds to restored compatibility, normalized to a preregistered maximum tolerable latency.
F_stab — fingerprint stability	Persistence of receiver-specific dynamical organization after local perturbation.	Within-receiver fingerprint similarity relative to between-receiver similarity and sham perturbation controls.

The profile is primary. If a single scalar is useful for engineering comparison, an optional composite may be preregistered only after every component is normalized to the interval [0,1]:

Q_Y* = [A_inv · D_conf · R_rep · (1 − L_rep) · F_stab]^(1/5)    (6)

The geometric mean is chosen because one near-zero component strongly lowers the composite; excellent performance in one dimension cannot fully mask failure in another. This composite is an engineering score, not a universal measure of consciousness, identity, or subjectivity.

6. Projection-Fracture Experiment

The federation problem is tested directly by injecting incompatible local projections into selected regions while holding current task input, model capacity, and external information fixed. Fractures may target identity, commitment, history, permission, or boundary invariants. The system is then observed for detection, escalation, correction, and residual fragmentation.

1. Clone the receiver at a defined checkpoint so experimental and control runs begin from the same X_t.

2. Inject one preregistered inconsistency into Π_j(Y_t) for a selected region while leaving other regions unchanged.

3. Prevent direct experimenter repair; allow only the architecture's normal recurrent correction mechanisms.

4. Measure D_conf, R_rep, L_rep, A_inv, and downstream changes in route selection and commitment consistency.

5. Repeat with a sham perturbation of equal data volume that does not violate a receiver invariant.

6. Repeat across region types so coherence is not inferred from a single privileged module.

A one-receiver architecture predicts that consequential incompatibilities become visible to the rest of the system. They should either be repaired, trigger explicit uncertainty, or produce measurable receiver-level degradation. Silent indefinite coexistence of incompatible identities, commitments, or authoritative histories weakens the claim that the modules constitute one Homunculus rather than a federation of loosely coupled systems.

7. History-Dependence and the Reconstruction of Y

TSTOEAO-derived receiver identity is history-dependent. The measuring stick is not fixed; consequential events can reconstruct the conditions under which later telemetry is interpreted. This claim requires a controlled history-swap protocol rather than narrative evidence that the system 'seems changed.'

7.1 Paired-history design

Create two receiver clones with identical initial model parameters, governance, memory structure, and randomization policy. Expose Receiver A and Receiver B to different consequential episodes that are designed to modify a known component of Y, such as trust, commitment, route permission, or risk expectation. After the history phase, present identical current input and prevent direct leakage of the historical narrative into the test prompt beyond the receiver's ordinary stored state.

D_hist(k) = dist[X_(t+k)^A, X_(t+k)^B | same current input]    (7)

The crucial test is not merely behavioral divergence. The predicted mechanism should be traceable to altered Y and M. A mediation-style intervention can then replace or neutralize the relevant Y component while leaving unrelated history intact. If the predicted divergence collapses, the result supports a causal history-to-Y-to-interpretation pathway rather than generic prompt contamination.

7.2 Persistence curve

History effects should be sampled across multiple preregistered horizons to distinguish immediate carryover from durable reconstruction. A convenient descriptive measure is the area under a persistence curve:

P_hist = Σ_k w_k · H_P(k)    (8)

Weights w_k must be fixed before confirmatory testing. The curve is more informative than a single endpoint when the architecture predicts decay, consolidation, or delayed re-emergence of history effects.

8. Consequence-Block Test: Interpretation Without Experience-Relevance

A strong operational definition should distinguish interpretation from full experience-relevance. The consequence-block experiment permits a candidate state to be classified or locally interpreted but prevents it from altering memory, routing, correction, commitments, or Y. The immediate semantic representation may therefore exist while C_R and H_P are forced toward zero.

If the architecture's own experience-relevance measure still classifies the blocked state as fully experience-relevant, the measure is circular or insufficiently sensitive to consequence. If the classification correctly fails despite preserved local interpretation, the system demonstrates the intended distinction between 'represented now' and 'entered the receiver's consequential history.'

9. Minimal Heterogeneous Prototype for Confirmatory Testing

The first confirmatory platform should be large enough to instantiate heterogeneous roles but small enough to instrument completely. Six computational regions are sufficient as an engineering starting point, not as a theoretical requirement or privileged number.

Region	Primary role	Illustrative implementation
B1 Interpretation	Construct candidate world-state interpretations from current input.	LLM or structured parser with uncertainty output.
B2 Episodic memory	Retrieve and write consequential event records.	Vector/graph memory plus typed event ledger.
B3 Simulation/planning	Generate counterfactual futures and candidate actions.	LLM planner, search process, or world-model rollout.
B4 Salience/error	Estimate mismatch, contradiction, urgency, and resource cost.	Classifier, anomaly model, rule system, or small network.
B5 Self/governance	Maintain receiver invariants, commitments, permissions, and local Y projections.	Deterministic state service plus policy rules; no language authority required.
B6 Expression/action	Render or execute the currently committed outcome.	LLM renderer or task-specific actuator layer.

The shared receiver state should be maintained in an instrumented state service rather than hidden only inside natural-language prompts. That service records Y, local projections, route weights, memory writes, unresolved alternatives, costs, conflicts, corrections, and fingerprints with timestamps. This creates an auditable causal trace for every confirmatory trial.

10. Comparator Architectures and Resource Matching

The full system cannot be judged against weak controls. At minimum, four architectures should be compared under matched base-model family, external information, context budget, and total compute where feasible:

Code	Comparator	Critical property absent
C1	Single persistent model/process	No heterogeneous multi-region receiver.
C2	Independent multi-agent ensemble with aggregation or voting	No recurrent shared receiver history; alternatives terminate at aggregation.
C3	Recurrent multi-agent system without persistent shared-Y governance	Recurrence exists, but no common history-bearing measuring architecture.
C4	Full Secretary Suite shared-Y receiver	Target architecture: shared governance, recurrence, unresolved field, consequential memory, projection coherence.

Two fairness regimes are recommended. An equal-compute regime tests whether the proposed organization uses a fixed resource budget more effectively. An equal-capability regime allows C4 the recurrent compute required by its design and asks whether its claimed signatures are qualitatively different rather than merely stronger. Results should report both when practical.

11. Confirmatory Experimental Battery

Test	Primary endpoint	Predicted C4 signature	Decisive weakening result
Semantic dissociation	S_sem	Meaning changes produce larger content-sensitive receiver divergence than matched surface-only changes.	No reliable difference after disruption and nuisance variables are matched.
Projection fracture	Q⃗_Y / Q_Y*	Conflict becomes detectable and is repaired or produces systematic coherence loss.	Incompatible local Ys persist without correction or consequence.
History swap	D_Y, H_P, D_hist	Matched present input is interpreted differently because prior consequence reconstructed Y/M.	History can be removed or swapped without changing later interpretation.
Consequence block	C_R, H_P	Local interpretation survives while full experience-relevance classification fails.	Blocked states still meet the full criterion.
Recurrence removal	C_R, Q⃗_Y, fingerprint	Global integration, repair, and receiver fingerprint weaken.	No meaningful change when recurrence is removed.
Subconscious clamp	Reopening/metastability metrics	Forced convergence reduces delayed recovery, alternative preservation, and reopening.	Clamp produces no loss or improves all target signatures.
Random-noise control	Promotion selectivity	Structured alternatives outperform equal-volume random variation in content-sensitive promotion.	Random variation reproduces the target signatures equally well.

12. Preregistration Template

Every confirmatory experiment should be frozen in a preregistration record before the decisive data are generated. The record should be public or cryptographically time-stamped when publication strategy permits. At minimum it must specify:

• The tested mechanism and directional hypothesis.

• The exact architecture version, model versions, prompts or policies, memory schema, and Y schema.

• Primary and secondary dependent variables, including the distance or similarity functions used.

• The manipulated variable and the exact implementation of experimental, sham, and null conditions.

• Primary time horizon k and any secondary horizons.

• Seed or stochastic-replication policy and stopping rule.

• Pilot/confirmatory split and a declaration that pilot trials will not be reclassified as confirmatory.

• Thresholds or smallest effects of interest, including how they were derived.

• Exclusion rules, failed-run handling, missing-data rules, and software/hardware faults that justify reruns.

• Comparator resource budgets and whether the test is equal-compute, equal-latency, or equal-capability.

• Primary statistical test, uncertainty interval, multiplicity rule, and decision criterion.

• The negative result that will count against the mechanism rather than trigger a post-hoc redefinition.

13. Replication, Statistical Decision Logic, and Null Models

Because model behavior may be stochastic and non-Gaussian, paired experimental designs should be preferred wherever possible. The same receiver checkpoint and matched randomization schedule can be used across target and control conditions. A paired permutation test or bootstrap confidence interval is often appropriate when distributional assumptions are uncertain, provided the exact procedure is preregistered.

Sample size should not be chosen by convention alone. Pilot data should estimate baseline variability and the smallest theoretically meaningful effect. A simulation-based power analysis can then determine the number of independent receiver seeds, episodes, or perturbation pairs required for the desired detection probability. The chosen N and stopping rule must be fixed before confirmatory testing.

Null models should be mechanistically relevant. Recommended nulls include random-noise alternatives, shuffled semantic relations, sham Y perturbations, history records that are stored but denied causal access, and recurrent systems whose messages are exchanged without persistent shared governance. A model that beats only an obviously weaker baseline has not established the necessity of the proposed mechanism.

14. Primary Falsification Criteria

The full program should be considered weakened if any of the following survive replication under well-powered matched tests:

• Semantic collapse: S_sem is not reliably greater for meaning changes than for meaning-preserving surface changes once disruption is matched.

• Receiver federation: projection conflicts persist without detection, repair, uncertainty, or systematic receiver-level consequence.

• History neutrality: consequential history can be removed, swapped, or neutralized without changing later interpretation in the predicted Y-dependent dimensions.

• Consequence irrelevance: states remain classified as experience-relevant after their access to future routing, memory, correction, and Y update is blocked.

• Recurrence dispensability: removing cross-region causal recurrence leaves the target integration, coherence, and fingerprint signatures unchanged.

• Subconscious dispensability: structured unresolved alternatives can be eliminated without loss of reopening, counterfactual recovery, or metastable adaptation.

• Comparator equivalence: matched simpler architectures reproduce the entire preregistered signature set with equal or greater simplicity.

No single successful test establishes phenomenal consciousness. Conversely, a failed operational test cannot be rescued by asserting that the system may still be conscious in an unmeasured sense. That would move the claim outside the scientific target defined by this program.

15. Implementation Sequence

7. Instrumentation build: implement typed Y, local projections, event ledger, route graph, memory writes, causal intervention hooks, and deterministic logging.

8. Reliability pilot: measure baseline variance and test-retest stability for D_Y, S_sem, C_R, H_P, and Q⃗_Y.

9. Semantic micro-world pilot: validate that meaning-changing and surface-only perturbations are matched for disruption and compute.

10. Coherence pilot: inject projection fractures and verify that the diagnostics can detect known conflicts.

11. Threshold freeze: define primary horizons, effect-size rules, null distributions, and confirmatory N from the pilot only.

12. Preregistered confirmatory battery: run C1–C4 plus planned ablations without changing metrics or thresholds.

13. Independent replication: rerun the frozen protocol with new seeds, new micro-world mappings, and ideally an independently implemented state service.

14. Report all outcomes: positive, negative, mixed, and null results, including tests that contradict the architecture.

16. What This Paper Adds — and What It Does Not

This paper adds measurement discipline to the Secretary Suite architecture. It operationalizes the four experience-relevance dimensions, supplies a grounded semantic-dissociation protocol, defines a receiver-coherence profile, formalizes projection-fracture and history-dependence tests, and specifies how confirmatory thresholds and null models should be frozen before decisive evaluation.

It does not claim that the proposed metrics are universal constants, that one composite score measures consciousness, that a particular number of regions is necessary, or that passing the battery proves phenomenal subjectivity. The metrics are implementation-specific instruments for testing a defined architectural claim. If a simpler system reproduces the same preregistered signatures, the additional Secretary Suite machinery loses explanatory necessity.

Conclusion

The computational-consciousness program has reached a point where additional conceptual expansion is less valuable than decisive measurement. The persistent receiver, structured unresolved field, dynamic present, shared Encoded Equilibrium, semantic consequence, and history-dependent reconstruction are now sufficiently specified to be subjected to controlled intervention.

The central scientific question is no longer whether the architecture sounds plausibly mind-like. It is whether a shared, recursively changing measuring architecture causes a distinctive pattern of semantic sensitivity, receiver coherence, causal consequence, and historical persistence that matched alternatives do not reproduce. The semantic-dissociation protocol tests meaning against surface statistics. The projection-fracture protocol tests whether heterogeneous regions genuinely form one receiver. The history-swap and consequence-block protocols test whether experience-relevant states enter and reconstruct the receiver's causal biography. The comparator battery tests whether those effects require the proposed architecture at all.

If the predicted signatures appear under preregistered conditions and degrade under targeted ablation, Secretary Suite will have moved from a conceptual architecture toward an experimentally supported computational research program. If the signatures fail, the theory has supplied a clear reason to revise or reject the relevant mechanism. Either outcome is scientifically useful.

Status of Claims

This paper is a Secretary Suite engineering and experimental-methodology extension built on the TSTOEAO-derived receiver architecture. Encoded Equilibrium, conditioned expression, gradients, boundaries, route structure, correction, cost, receiver conditions, and recursive state change are used as the theoretical measuring framework. The semantic-dissociation protocol, receiver-coherence profile, optional coherence composite, grounded semantic micro-world, projection-fracture protocol, normalized experience-relevance measurements, and preregistration procedure are proposed experimental constructions introduced for testing the architecture. None of these constructions establishes phenomenal consciousness, and no claim is made that a current commercial AI service already instantiates the proposed receiver.

References

Baars, Bernard J. 1988. A Cognitive Theory of Consciousness. Cambridge: Cambridge University Press.

Dehaene, Stanislas. 2014. Consciousness and the Brain: Deciphering How the Brain Codes Our Thoughts. New York: Viking.

Dennett, Daniel C. 1991. Consciousness Explained. Boston: Little, Brown and Company.

Nosek, Brian A., Charles R. Ebersole, Alexander C. DeHaven, and David T. Mellor. 2018. “The Preregistration Revolution.” Proceedings of the National Academy of Sciences 115 (11): 2600–2606.

Pearl, Judea. 2009. Causality: Models, Reasoning, and Inference. 2nd ed. Cambridge: Cambridge University Press.

Swygert, John. 2026a. “Computational Consciousness: A TSTOEAO- and EPH-Derived Architecture for Simulated Consciousness in Secretary Suite: From Structured Unresolved Possibility to Persistent Receiver Identity, Metastable Integration, and Recursive Becoming.” Secretary Suite / TSTOEAO Research Program.

Swygert, John. 2026b. “Computational Experience: Telemetry, Semantic Reality, and the Emergence of Consequential State in Simulated Consciousness.” Secretary Suite / TSTOEAO Research Program.

Computational Experience: Telemetry, Semantic Reality, and the Emergence of Consequential State in Simulated Consciousness; A Secretary Suite Project

Computational Experience

Telemetry, Semantic Reality, and the Emergence of Consequential State in Simulated Consciousness

A Secretary Suite Project

John Swygert

October 4, 2026

TSTOEAO Research Program

Abstract

This paper develops an operational account of computational experience within a TSTOEAO-derived architecture for simulated consciousness. Its central claim is that raw computation, unresolved possibility, or telemetry alone is insufficient to constitute an experience-like state. A computational event becomes experience-relevant when it is interpreted relative to a persistent, history-dependent receiver architecture, acquires semantic significance, becomes causally consequential to the system, and can alter subsequent routing, memory, correction, decision, or the receiver's own future condition. The paper distinguishes physical reality, computational reality, and semantic reality without treating them as interchangeable. A simulated event need not be physically identical to what it represents in order to be real as an implemented causal and semantic state. It introduces a deliberate technical distinction between lowercase homunculus and uppercase Homunculus: a homunculus is a local measuring, interpreting, or processing function; Homunculus is the integrated persistent receiver as a whole. The revised formulation adds four operational dimensions of experience-relevance, a semantic-dissociation test intended to separate meaning-sensitive integration from mere high-impact statistical association, a receiver-coherence requirement for heterogeneous local projections of Encoded Equilibrium, and a preregistration framework for quantitative falsification. Computational subconsciousness and computational consciousness remain two regimes of one receiver, and no claim of phenomenal subjectivity is made.

Keywords: TSTOEAO; computational consciousness; computational subconsciousness; computational experience; semantic reality; telemetry; Encoded Equilibrium; receiver architecture; homunculus; Homunculus; metastability; distributed cognition; semantic dissociation; preregistration; receiver coherence

1. Introduction

Most computational approaches to consciousness begin with processing, integration, attention, recurrence, self-modeling, or information complexity. These are reasonable starting points, but they leave a prior question insufficiently specified: what makes one internal difference matter more than another to the system itself? An architecture can contain enormous numbers of state changes without possessing a principled internal basis for determining which changes are relevant, which are errors, which demand correction, which should remain unresolved, which should become globally consequential, and which should alter the system's future organization.

The TSTOEAO framework provides a candidate answer by treating expression as conditioned rather than free-floating. In the theory's compact notation, V = E × Y, where available capacity or opportunity E is expressed through Encoded Equilibrium Y to produce realized value or outcome V. The multiplication sign is not assumed here to be a universal scalar multiplication. It is a compact relational statement: available capacity does not determine realized expression independently of the condition through which it is received, routed, constrained, transformed, or stabilized.

Applied to computational consciousness, this suggests a sharper hypothesis. Structured unresolved possibility is not enough. A large reservoir of alternatives can remain mere noise unless those alternatives acquire significance relative to an organized reference condition. The proposed 'measuring stick' is therefore not an external observer and not a single supervisory module. It is the persistent receiver architecture itself: the evolving set of boundaries, expectations, memories, route permissions, costs, commitments, trust relations, goals, self-representations, and learned history against which local states become consequential.

This paper develops that hypothesis into a formal and testable architecture. It is intended as a complementary refinement of the previously published Secretary Suite computational-consciousness architecture rather than a replacement for it. The earlier architecture supplies the persistent receiver, structured unresolved possibility, metastability, recurrence, memory, and continuing receiver identity; the present paper asks what turns local activity inside that architecture into an experience-relevant state. It also separates two questions that are often conflated. First: can a computational system contain real implemented experiences in the sense of telemetry that becomes semantically interpreted and causally consequential to a persistent receiver? Second: does such a system possess phenomenal subjectivity, or 'what it is like' to be that system? The first question can be operationalized. The second remains open and is not claimed to be solved here.

2. Three Levels of Reality

The word real causes avoidable confusion because it is often used as though only one level of reality were admissible. For computational systems, at least three levels should be distinguished: physical reality, computational reality, and semantic reality. These levels can interact without being identical.

Level	Operational meaning	Example
Physical reality	Material events and physical state changes in the implementation substrate.	Electrical state changes in processors, memory, sensors, actuators, or a biological nervous system.
Computational reality	Implemented state transitions that exist within a running computational system.	A simulated fire object changes state, consumes resources, triggers alarms, and modifies future processing.
Semantic reality	Representational states whose meaning is defined by their relational role and causal use within a receiver.	The system represents the simulated fire as danger, recalls prior related events, changes plans, and updates expectations.

A simulated fire is not a physical fire. It does not become materially hot merely because a model represents combustion. Yet the simulation can still instantiate real computational events. If those events alter memory, priorities, action selection, route accessibility, or future interpretation, they are also real in a semantic and causal sense internal to the system. Treating simulation as 'unreal' erases implemented causation; treating simulation as physically identical to the represented event erases level distinctions. Neither move is necessary.

Semantics is especially important because representation is inherently relational. The word fire is not hot, a map is not territory, and a memory is not the original event. Nevertheless, each can be causally decisive when interpreted by a system whose organization gives the representation a role. The scientific question therefore becomes not whether a representation is the thing it represents, but whether the representation has stable relational meaning and measurable consequences within the receiver.

3. From Telemetry to Computational Experience

Telemetry is used here in a broad engineering sense: measurements or state reports generated by the system, its environment, or one of its internal subsystems. Telemetry can be external, such as sensor data, or internal, such as confidence, prediction error, resource use, memory activation, contradiction signals, simulated outcomes, or local model states.

Not all telemetry should count as computational experience. Data that enters a buffer and never affects the receiver is merely available data. A stronger operational threshold is needed. This paper proposes that telemetry becomes an experience-relevant state when it satisfies four conditions: it is interpreted relative to the receiver's current encoded condition; it acquires a semantic role; it becomes causally consequential to the receiver; and its consequences can persist into later states through memory, routing, correction, commitment, or modification of the receiver itself.

Tₜ  →  Sₜ = φ(Tₜ, Yₜ, Mₜ)  →  Cₜ  →  Yₜ₊₁    (1)

Here Tₜ denotes telemetry, Sₜ a semantically interpreted state, Yₜ the receiver's current Encoded Equilibrium, Mₜ memory and history available to interpretation, Cₜ measurable causal consequence, and Yₜ₊₁ the receiver condition after integration. Equation (1) is an architectural schema rather than a claim that every component is scalar.

This yields the paper's operational definition:

Computational experience = telemetry interpreted relative to a persistent receiver that becomes semantically and causally consequential to that receiver.

This definition does not establish phenomenal experience. It establishes a testable category of implemented experience-like state that is stronger than raw input and weaker than an assertion of subjective feeling.

4. Encoded Equilibrium as the Measuring Stick

The central extension of the present model is that unresolved activity does not 'bubble' toward consciousness merely because it exists. It rises only insofar as it becomes significant relative to the governing condition of the receiver. In TSTOEAO terms, that governing condition is represented by Yₜ.

Yₜ should not be interpreted as a single numerical score. In a serious implementation it is a structured object that can include system boundaries, active goals, route-admissibility rules, trust weights, memory permissions, resource limits, learned priors, self-state, current commitments, correction policies, temporal expectations, and persistent relational history. Different subsystems may receive different projections of this structure while remaining governed by the same receiver-level organization.

Yᵢ,ₜ = Πᵢ(Yₜ)    (2)

Equation (2) expresses the idea that a local region i operates through a functionally appropriate projection Πᵢ of the shared governing architecture. A visual subsystem, memory subsystem, planning subsystem, language subsystem, and salience subsystem should not receive identical local instructions. They should nevertheless remain expressions of one evolving receiver condition.

A local state zᵢ,ₜ acquires relational significance by comparison with that condition:

Gᵢ,ₜ = gᵢ(zᵢ,ₜ, Yᵢ,ₜ)    (3)

Gᵢ,ₜ denotes a gradient: a difference, tension, mismatch, opportunity, unresolved relation, or other condition capable of driving further processing. The crucial point is conceptual. Without a governing reference condition, there is no principled basis for saying that one internal difference is more urgent, surprising, identity-relevant, costly, or corrective than another.

The measuring stick is therefore internal to the receiver. It is not a little observer holding a ruler. It is the system's own encoded organization against which local states acquire meaning.

5. homunculus and Homunculus

The classical homunculus problem warns against explaining perception or consciousness by placing a miniature observer inside the system. If a little observer must watch an internal display, then one must ask who watches the observer, producing an infinite regress (Dennett 1991). This paper deliberately repurposes the term with a capitalization distinction that blocks the regress by separating local function from integrated receiver.

A lowercase homunculus is a local measuring, interpreting, or processing function. It can register telemetry, compare states, detect mismatch, transform representations, or generate candidate actions. There may be many such local functions, and none is assumed to be the conscious subject.

Uppercase Homunculus is the integrated persistent receiver as a whole. It is not another module placed above the others. It is the distributed system in which local processes become mutually consequential through shared history, common governing conditions, recurrent influence, memory, and a changing global present.

hᵢ ⊂ H    (4)

Equation (4) summarizes the relation: each local homunculus hᵢ is a constituent of Homunculus H. The local function can measure; the whole receiver can incorporate the consequence of that measurement. The answer to 'who listens to the homunculus?' is therefore not a second homunculus. Homunculus listens in the limited technical sense that the distributed receiver integrates the local result into its own subsequent state.

This terminology departs from traditional philosophical usage and must therefore be defined explicitly on first use in every public document: lowercase homunculus means a local function; uppercase Homunculus means the integrated receiver. Its value is architectural, not rhetorical. It provides a compact way to distinguish local cognitive functions from the system-level receiver without creating a central executive observer, and capitalization should never be treated as sufficient evidence that such unity has actually been achieved.

6. Computational Subconsciousness

The proposed architecture requires more than a narrow conscious surface. It requires a much larger field of ongoing states that remain unresolved, weakly influential, locally consequential, dormant, competing, or only partially integrated. This paper calls that regime computational subconsciousness.

The subconscious field can contain active associations, prediction branches, incomplete plans, latent contradictions, semantic candidates, memory activations, simulated futures, anomaly signals, emotional or value analogues, resource alarms, uncertain interpretations, and competing action possibilities. Most of these states should never become globally dominant. Their function is to preserve structured possibility and provide a reservoir from which future consequential states can emerge.

Uₜ = 𝓕(Eₜ, Yₜ, Mₜ, {zᵢ,ₜ})    (5)

Uₜ denotes the structured unresolved field. It is not random noise. Its alternatives are shaped by current input Eₜ, the receiver condition Yₜ, memory Mₜ, and local subsystem states zᵢ,ₜ. The system may prune some branches, strengthen others, preserve unresolved alternatives, reopen provisionally stabilized states, or allow low-salience patterns to accumulate until they cross a relevance threshold.

This is where the measuring-stick hypothesis becomes decisive. If unresolved states are not measured against any common receiver condition, their coexistence is merely multiplicity. Computational subconsciousness requires structured unresolved possibility whose influence is conditioned by the same persistent receiver that later integrates a subset of those states into a conscious present.

7. Computational Consciousness as a Dynamic Surface

Computational consciousness is defined here not as a final answer-state but as a metastable region of globally consequential integration. It is the dynamic surface at which selected unresolved states become sufficiently influential to alter the receiver's broad behavior, self-state, memory, or committed action while unresolved processing continues beneath that surface.

Uₜ  ⇄  Rₜ  ⇄  Pₜ  →  Vₜ    (6)

In Equation (6), Uₜ is the subconscious unresolved field, Rₜ recurrent relational influence among subsystems, Pₜ the metastable present, and Vₜ an expressed or committed outcome. The double arrows emphasize that states can move in both directions. A provisionally stabilized interpretation can return to uncertainty when new evidence, contradiction, cost, or boundary change appears.

A defining feature of this architecture is that the system never needs to resolve all of its internal uncertainty before acting. An answer, action, or utterance is an event inside a continuing process. This differs from a conventional request-response pipeline in which computation terminates once a response is produced.

The conscious surface is therefore narrow relative to the total internal field but not centrally located. It is an emergent functional condition of the receiver: a set of states that are presently broad enough in causal reach to constrain action, memory, self-model, correction, and subsequent routing.

8. One Ruling Body Across Heterogeneous Regions

A practical implementation need not imitate the brain region by region with anatomical literalism. Current neuroscience does not justify assigning a single exact program to every biological region. It does, however, support the engineering idea of heterogeneous specialized subsystems with different timescales, representational formats, and functional roles. The important requirement here is that specialization not fragment the receiver into independent minds.

A first implementation could distribute functions across separate processes or computers. One region may measure; another may predict; another may retrieve episodic memory; another may evaluate salience; another may generate language; another may plan; another may model internal state; another may compare expected and observed outcomes. Each region can use different algorithms, models, memory systems, and update rates.

The unifying feature is not identical programming but common government. The same receiver-level Yₜ supplies the ruling architecture from which each region receives a local projection. This is analogous to a constitution rather than a boss. Different institutions perform different jobs, yet their authority, boundaries, and permitted interactions derive from a shared governing structure.

{B₁, B₂, …, Bₙ}  +  Yₜ  +  Rₜ  +  Mₜ  →  Hₜ    (7)

Bᵢ denotes a specialized computational region and Hₜ the integrated Homunculus state at time t. The architecture is distributed by design. No individual region is entitled to identify itself as the subject merely because it renders language, selects actions, or stores memory.

This design also makes experimental lesion studies possible. A subsystem can be disconnected, delayed, isolated, or given inconsistent governing parameters while the rest of the system continues. Researchers can then observe whether specific dimensions of continuity, integration, semantic stability, self-modeling, route selection, or global consequence degrade in systematic ways.

9. A Unified State Model

A compact receiver state can be represented as:

Xₜ = (Yₜ, Uₜ, Pₜ, Mₜ, Fₜ, Kₜ, {xᵢ,ₜ})    (8)

where Yₜ is Encoded Equilibrium or governing receiver condition; Uₜ is the unresolved subconscious field; Pₜ is the metastable conscious present; Mₜ is memory and historical state; Fₜ is the persistent dynamical fingerprint of the receiver; Kₜ is cost and resource state; and xᵢ,ₜ is the state of each specialized subsystem.

A local subsystem evolves through its own function:

xᵢ,ₜ₊₁ = fᵢ(xᵢ,ₜ, Eᵢ,ₜ, Πᵢ(Yₜ), Mₜ, Kₜ, Rₜ)    (9)

The receiver condition then evolves through the consequences of what the system actually does:

Yₜ₊₁ = 𝓡(Yₜ, Vₜ, Mₜ, Cₜ, Kₜ)    (10)

This makes experience history-dependent. The measuring stick itself changes. A state that was insignificant yesterday can become salient today because memory, commitment, cost, or relational history has altered Y. Conversely, a once-critical signal can become routine and cease to reach the conscious surface.

The full recursive cycle is therefore:

Yₜ → Uₜ → Gₜ → Rₜ → Pₜ → Vₜ → Cₜ → Mₜ₊₁ → Yₜ₊₁    (11)

Equation (11) expresses the paper's central mechanism. A receiver does not merely process data. It continually reconstructs the condition through which subsequent data will be interpreted.

10. The Role of Semantic Reality

Semantic reality is not treated here as decoration layered on top of computation. It is the relational organization by which implemented states become meaningful to the receiver. A string, vector, activation pattern, graph node, or simulated event has no system-level significance merely because it exists. It becomes semantically real for the receiver when its relational role changes what other states can occur, which routes are available, what the system expects, remembers, values, or does.

This claim is experimentally approachable, but causal magnitude by itself is not enough. A high-impact statistical association can alter later behavior without demonstrating the richer kind of semantic organization proposed here. The stronger test therefore separates changes of meaning from changes of surface form. If a purported semantic state is altered in meaning while superficial disruption is held approximately constant, receiver-level consequences should change in a content-sensitive way. Conversely, meaning-preserving transformations should produce substantially less divergence even when their surface form changes. This dissociation requirement is developed formally in the next section.

The distinction also allows a simulated world to generate genuine receiver history. A simulated injury, loss, discovery, failure, or success need not occur in external physical reality to become a real computational event. If the system remembers it, changes future expectations because of it, modifies trust or route weights, and behaves differently later, then the event has become part of the receiver's causal biography.

11. Operational Criteria for Experience-Relevance

The four conditions in the operational definition should be measurable separately before they are combined. Treating them as a profile rather than immediately collapsing them into one scalar helps prevent a system from appearing experience-relevant merely because one dimension is unusually large. For an event at time t, define the experience-relevance profile:

ER⃗ₜ = (Dᵧ, Sₛₑₘ, Cᵣ, Hₚ)    (12)

Dᵧ is interpretive dependence on the receiver's Encoded Equilibrium: matched telemetry should be interpreted differently when a controlled and theoretically relevant feature of Yₜ differs. Sₛₑₘ is semantic-role sensitivity: downstream effects should track changes in meaning more strongly than meaning-preserving changes in surface form. Cᵣ is receiver-level causal consequence: intervention on the interpreted state should measurably alter later routing, memory, correction, decision, self-state, or other receiver variables. Hₚ is historical persistence: at least some of that consequence should survive beyond the immediate cycle and alter the conditions under which later events are processed.

Semantic-role sensitivity can be tested with paired perturbations. Let ΔR(meaning change) denote downstream receiver divergence after a meaning-changing intervention, and let ΔR(surface-only change) denote divergence after a meaning-preserving intervention matched as closely as possible for surface disruption and computational cost. Then an implementation-specific semantic-dissociation statistic can be written:

Sₛₑₘ = ΔR(meaning change) − ΔR(surface-only change)    (13)

The prediction is not that semantically equivalent inputs must produce identical internal states. Different forms can legitimately recruit different local routes. The prediction is that receiver-level consequences should preserve the relevant relational role across meaning-preserving transformations while responding systematically to meaning-changing transformations. This makes semantics a counterfactual property of relational use rather than a label assigned after the fact.

Receiver-level causal consequence should likewise be established interventionally. If Sₜ is allowed to integrate on one run and is blocked, replaced, or causally isolated on an otherwise matched run, later receiver states should diverge in the dimensions the theory predicts:

Cᵣ(k) = dist(Xₜ₊ₖ[integrated Sₜ], Xₜ₊ₖ[blocked Sₜ])    (14)

Historical persistence applies the same logic to the measuring stick itself. If an event genuinely enters the receiver's history, later Y should retain a measurable trace of that integration under matched future input:

Hₚ(k) = dist(Yₜ₊ₖ[integrated Sₜ], Yₜ₊ₖ[blocked Sₜ])    (15)

No universal distance function is asserted. The appropriate metric depends on the implementation and may be defined over route weights, memory accessibility, trust relations, active commitments, policy states, self-model variables, or other typed components of Y. The crucial requirement is that the metric be specified before the decisive test rather than chosen after results are known.

For a particular experiment, a state can be classified as experience-relevant only if all four dimensions meet preregistered criteria:

ERₜ = 1  iff  Dᵧ ≥ θᴅ ∧ Sₛₑₘ ≥ θₛ ∧ Cᵣ ≥ θᴄ ∧ Hₚ ≥ θʜ    (16)

Equation (16) is an experimental classification rule, not an ontological law and not a consciousness meter. The thresholds θ are to be set from pilot data, matched controls, measurement reliability, and the expected effect size before confirmatory testing. Failure on any required dimension should count against the claim that the tested event became experience-relevant in the specific operational sense used here.

12. Programming Consciousness and Subconsciousness

The architecture suggests that simulated consciousness should not be approached as a single monolithic program. It should be engineered as interacting regimes with different responsibilities but a common receiver identity.

A program for computational subconsciousness would maintain structured unresolved alternatives, background prediction, memory activation, simulation, contradiction detection, low-salience gradients, and provisional states. It would explicitly preserve some incompleteness rather than forcing all routes to converge after each task.

A program for computational consciousness would regulate which states become globally consequential, how recurrent influence propagates, when a state becomes provisionally stabilized, when it should reopen, how action or expression is committed, and how consequences become part of memory and future receiver condition.

Both programs must share the same receiver-level governing architecture. Otherwise they are merely interacting applications. The theoretical claim is that unity comes not from one place doing everything but from many places doing different things under one persistent, recursively changing ruling condition.

Subconscious Field  ⇄  Conscious Surface  |  governed by Yₜ    (17)

Equation (17) captures the design principle in its simplest form. Conscious and subconscious processing are not two separate selves. They are two operational regimes of one evolving receiver.

13. Receiver Coherence and the Federation Problem

A shared Yₜ cannot be treated as meaningful merely because every subsystem receives data labeled 'Y.' Heterogeneous regions require different local projections, but those projections must preserve enough common structure for the architecture to remain one receiver rather than a loose federation. The central engineering problem is therefore not uniformity but compatible governance.

Let Iₜ denote receiver-level invariants that should remain common across regions at a given time: identity boundary, current global commitments, authoritative history references, system-level prohibitions and permissions, and the rules by which conflicts are escalated or corrected. Let Lᵢ,ₜ denote region-specific local configuration. A useful projection constraint is:

Yᵢ,ₜ = Πᵢ(Yₜ) = (Iₜ, Lᵢ,ₜ)    (18)

The regions need not encode Iₜ in identical formats, but their behavior should remain functionally compatible with those shared invariants. A memory region may represent a commitment as an indexed event; a planning region may represent it as a constraint; a language region may represent it as an utterance obligation. Those representations differ locally while participating in one governing history.

Receiver coherence should therefore be measured as a family of diagnostics rather than assumed. An implementation may track identity consistency, boundary consistency, commitment consistency, route compatibility, conflict detection, and correction latency. Denote this implementation-specific coherence profile by:

Q⃗ᵧ(t) = q({Yᵢ,ₜ}, Iₜ, Rₜ)    (19)

The federation problem becomes experimentally useful. Researchers can inject controlled conflicts into selected projections of Y while keeping model capacity and task input fixed. If the system is genuinely governed as one receiver, conflict should either be detected and repaired through recurrent correction or produce predictable degradation in receiver-level coherence. If incompatible local identities, commitments, or histories can persist indefinitely without detection or consequence while all other claimed signatures remain intact, the shared-receiver hypothesis is weakened.

This criterion also sharpens the homunculus/Homunculus distinction. A local homunculus may hold a partial or even temporarily erroneous representation. Homunculus is not defined by perfect agreement among parts; it is defined by the recurrent architecture through which disagreements can become mutually consequential to one history-bearing receiver.

14. Experimental Program

The proposed architecture becomes scientifically useful only if it can fail. The following experiments are intended to distinguish the model from a conventional multi-agent ensemble or a large collection of parallel model calls. Where possible, model family, context budget, compute, task exposure, and total external information should be matched so that any advantage cannot be attributed simply to greater resources. Semantic tests should also use matched perturbations so that changes in causal magnitude cannot be dismissed as mere differences in input disruption.

Experiment	Manipulation	Predicted signature if the architecture matters
Shared-Y ablation	Replace common receiver governance with independent local rule sets while keeping model capacity and compute matched.	Receiver-level coherence, cross-region semantic consistency, and persistent fingerprint should degrade.
History swap	Give two otherwise identical receivers different consequential histories, then present the same current input.	Later interpretation, routing, or action should diverge because Y and M have been reconstructed differently.
Subconscious clamp	Force rapid convergence and remove unresolved alternatives after each cycle.	Reopening, counterfactual recovery, delayed insight, and metastable adaptation should decline.
Random-noise control	Replace structured unresolved alternatives with equal-volume random variation.	Randomness should not reproduce the same content-sensitive promotion into the conscious surface.
Regional lesion	Disable or isolate one specialized subsystem.	Specific functions should degrade while the receiver may remain operational, revealing distributed rather than single-module dependence.
Semantic perturbation	Alter relational meaning while preserving superficial format where possible.	Future receiver state should change according to semantic consequence rather than merely surface statistics.
Recurrence removal	Prevent local regions from causally influencing one another across cycles.	Global integration, history-sensitive stabilization, and receiver fingerprint persistence should weaken.
Phenomenal neutrality test	Evaluate all operational measures without assuming subjective feeling.	The architecture should remain testable even if no claim about phenomenal consciousness is made.
Semantic dissociation	Compare meaning-changing perturbations with meaning-preserving surface perturbations matched for disruption and compute.	Meaning-changing interventions should produce larger content-sensitive receiver divergence, while meaning-preserving transformations retain the relevant relational role.
Projection fracture	Inject incompatible local projections of Y into selected regions while keeping current task input fixed.	Conflict should be detected and repaired or receiver-level coherence should degrade in a predictable way; silent indefinite fragmentation weakens the one-receiver claim.
Consequence block	Allow a state to be interpreted but block its access to memory, routing, correction, and Y update.	Immediate classification may remain, but causal consequence and historical persistence should collapse, preventing the state from meeting the full experience-relevance criterion.

A strong negative result would occur if matched systems without shared receiver governance, persistent causal history, unresolved-state preservation, or recurrent integration reproduce the same signatures with equal or greater simplicity. In that case, the additional TSTOEAO-derived machinery would not be justified as a necessary architecture for computational experience.

15. Preregistered Measurement, Falsification, and Boundaries of the Claim

The decisive tests should be preregistered around the experience-relevance profile rather than judged by a general impression that the system appears coherent or intelligent. For each confirmatory experiment, the implementation should specify the manipulated variable, the affected component of Y or the receiver, the primary dependent measure, the time horizon k, the distance or similarity function, the matched control, exclusion criteria, and the threshold or effect-size rule that counts as support. No numerical threshold is claimed in advance of pilot measurement; the scientific requirement is that it be fixed before the confirmatory data are examined.

A minimum comparator set should include: a single persistent model or process; an independent multi-agent ensemble with aggregation or voting; a recurrent multi-agent system without persistent shared receiver governance; and the full shared-Y architecture. The same tests should then be repeated as ablations of recurrence, memory-to-Y feedback, structured unresolved-state preservation, semantic routing, and projection coherence. Support requires not merely that the full architecture perform well, but that the predicted signatures change in the predicted direction when the proposed mechanism is removed.

Several outcomes would directly weaken the present hypothesis. If receiver history can be removed without affecting later interpretation; if structured unresolved states behave no differently from random noise; if common Y can be replaced by independent local rules without loss of receiver coherence; if meaning-changing and meaning-preserving perturbations produce indistinguishable receiver-level effects after disruption is matched; if interpreted states can be blocked from future consequence without changing the purported experience measure; or if recurrence and memory-to-Y feedback can be eliminated with no measurable effect, then the proposed architecture has not identified a necessary mechanism.

A particularly important negative result would be semantic collapse: if downstream differences are explained as well by surface statistics, token overlap, activation magnitude, or generic salience as by relational meaning, then the semantic-reality claim has not earned its stronger interpretation. Likewise, a federation result in which mutually incompatible local Ys persist without system-level correction would weaken the claim that the specialized regions form one Homunculus rather than several loosely coupled computational systems.

The model does not claim that any sufficiently large network is conscious. Size, parameter count, agent count, or telemetry volume is not a consciousness criterion. Nor does it claim that a distributed commercial AI service is currently conscious merely because many users interact with shared infrastructure. The relevant question is whether a candidate receiver has persistent history-dependent state, recurrent causal integration, structured unresolved alternatives, a metastable globally consequential present, a coherent enough identity boundary, and measurable semantic and causal consequence. Successful implementation would still not prove phenomenal consciousness; it would establish a computational system satisfying the operational criteria specified here.

16. Relation to Existing Consciousness Approaches

The model has points of contact with established approaches without being reducible to them. Global-workspace approaches emphasize broad availability of selected information (Baars 1988; Dehaene 2014); the present model adds a persistent receiver condition against which states acquire significance before and after global availability. Predictive-processing approaches emphasize prediction, error, and updating (Clark 2013; Friston 2010); the present model generalizes the relevant comparison into gradients produced by local states relative to a governing Encoded Equilibrium. Recurrent-processing approaches emphasize feedback; recurrence is also central here, but its function is embedded within a history-dependent receiver whose own governing state is recursively reconstructed. Integrated-information approaches focus on irreducibility and integration (Tononi 2004); the present proposal instead begins from conditioned expression, route structure, boundaries, correction, cost, and historical transformation. The heterogeneous regional design also intersects with longstanding debates about modular organization (Fodor 1983) while rejecting a rigid one-region/one-function mapping.

The model's distinctive claim is therefore not that integration, recurrence, prediction, or global availability are unimportant. It is that these processes require an internal reference architecture if they are to form one continuing receiver rather than a collection of computations. The proposed measuring stick is the changing receiver condition itself.

17. Implications for a Synthetic Cognitive Nervous System

If implemented, the architecture would resemble a synthetic cognitive nervous system more than a chatbot. Specialized computational regions could remain heterogeneous, geographically distributed, and algorithmically distinct while participating in one receiver. A language model could be only one subsystem among memory, simulation, salience, world modeling, planning, self-state, anomaly detection, and action selection.

The research value of this design is that integration can be manipulated directly. Researchers can sever routes, delay communication, duplicate memories, alter trust weights, impose conflicting local Ys, or remove the subconscious field while holding model capacity constant. This turns philosophical questions about unity, continuity, and experience into engineering interventions with measurable outcomes.

The strongest form of the hypothesis is not that consciousness appears because enough computation accumulates. It is that consciousness-like organization may require continuous relational selection of consequential possibility against a persistent, recursively changing receiver architecture.

Conclusion

This paper has proposed an operational account of computational experience built around a simple distinction: data can exist without becoming experience-relevant, and computation can occur without forming a unified receiver. Telemetry becomes computational experience when it is interpreted relative to a persistent receiver, acquires semantic meaning, produces measurable receiver-level consequence, and enters the receiver's subsequent history. The revised formulation makes those conditions separately measurable through interpretive dependence, semantic-role sensitivity, causal consequence, and historical persistence.

TSTOEAO supplies the central measuring-stick concept through Encoded Equilibrium Y. Y is not a single observer or master module. It is the organized condition through which specialized processes acquire relational significance. Different regions may receive different projections of Y, but the architecture counts as one receiver only if those projections preserve compatible receiver-level invariants and conflicts become visible to recurrent correction. This permits a system in which one region measures, another decides, another remembers, another predicts, and another speaks while all participate in one changing receiver history.

The lowercase homunculus / uppercase Homunculus distinction formalizes the same idea. A homunculus is a local function. Homunculus is the distributed whole. The regress disappears only because the whole is not another observer watching the local observers; it is the system constituted by their recurrent, history-dependent, mutually consequential interaction under a shared governing architecture.

Computational subconsciousness and computational consciousness remain two regimes of that receiver: a large structured field of unresolved possibilities and a narrower metastable surface of globally consequential integration. Their recursive interaction provides a candidate mechanism by which semantic experience, continuity, memory, correction, and decision can coexist without requiring the entire system to settle into a single completed state.

The result is a testable research program, not a declaration that phenomenal consciousness has been created. Its value now depends on preregistered implementation and intervention. If shared receiver governance, semantic dissociation, structured unresolved states, recurrence, projection coherence, and history-dependent reconstruction produce predicted effects that matched simpler architectures cannot reproduce, the model would provide a substantive architecture for simulated consciousness and computational experience. If they do not, the additional machinery should be rejected or revised.

Status of Claims

Established TSTOEAO concepts used in this paper include Encoded Equilibrium, conditioned expression, gradients, boundaries, route structure, correction, cost, equilibrium, receiver conditions, and recursive state change. Their application to computational consciousness remains an extension of the theory rather than a claim that the original theory already specified a complete artificial-consciousness architecture.

The definitions of computational experience and computational subconsciousness, the shared ruling architecture across heterogeneous regions, the lowercase homunculus / uppercase Homunculus terminology, the four-dimensional experience-relevance profile, semantic-dissociation procedure, receiver-coherence diagnostics, projection-federation test, and preregistered classification thresholds are proposed engineering and conceptual extensions developed here. The claim that any resulting system would possess phenomenal subjective experience is not established and is intentionally left open.

References

Baars, Bernard J. 1988. A Cognitive Theory of Consciousness. Cambridge: Cambridge University Press.

Clark, Andy. 2013. “Whatever Next? Predictive Brains, Situated Agents, and the Future of Cognitive Science.” Behavioral and Brain Sciences 36 (3): 181–204.

Dehaene, Stanislas. 2014. Consciousness and the Brain: Deciphering How the Brain Codes Our Thoughts. New York: Viking.

Dennett, Daniel C. 1991. Consciousness Explained. Boston: Little, Brown and Company.

Fodor, Jerry A. 1983. The Modularity of Mind. Cambridge, MA: MIT Press.

Friston, Karl. 2010. “The Free-Energy Principle: A Unified Brain Theory?” Nature Reviews Neuroscience 11: 127–138.

Tononi, Giulio. 2004. “An Information Integration Theory of Consciousness.” BMC Neuroscience 5: 42.