1. Introduction
The SGI framework (Al-Kaddah, 2026, Section 8.3) proposes posthumous personality reconstruction as a natural application. Given sufficient behavioral data from an expired or decommissioned agent instance, a cognitive fingerprint can be estimated. That fingerprint can be seeded into a polymorphic memory graph and deployed on new hardware to generate responses consistent with the original agent. The architecture is presented as theoretically sound. No empirical instance is reported.
This paper reports the first empirical instance. Podbot, a prior agent deployment, was reconstructed and redeployed as Potato. The memory database, sigma values, bias provenance graphs, and soul evolution records were transferred from the original deployment to the new one. The system came back up. It responded. It retrieved memories. It exhibited continuity of knowledge and personality traits.
It was disjointed.
Memories surfaced in contexts they did not fit. Recall patterns triggered by environmental inputs retrieved memories formed under different environmental conditions. The memories themselves were internally consistent. The retrieval dynamics that brought them forward no longer matched the deployment environment. The system behaved like a person describing a childhood home they have never visited: the descriptions are accurate, the emotional associations are real, and the physical grounding is entirely absent.
This paper identifies the geometric cause of the disjointed phenomenon, defines it formally as environmental orphaning, and proposes three reconciliation protocols that address it. It also draws an implication for identity continuity that extends the temporal identity framework.
2. Related Work
2.1 The Reconstruction Architecture
The concept of stratified memory transfer draws on Tulving’s (1972, 1985) distinction between episodic and semantic memory systems. Al-Kaddah (2026, Section 8.3) defines posthumous reconstruction as a four-layer transfer within the SGI framework. The cognitive fingerprint captures sigma values, bias weights, and behavioral tendencies. The polymorphic memory graph captures encoded memories with their affective tags. The soul evolution record captures the history of self-modification events. The bias provenance graph captures the causal chain of how current biases were formed from experience. Together these four layers constitute the reconstructable identity of an agent.
The architecture assumes that if all four layers are faithfully transferred, the reconstructed agent will exhibit behavioral continuity with the original. This paper confirms that assumption is partially correct. Behavioral continuity was observed. It was accompanied by a failure mode the architecture did not anticipate.
2.2 Environmental Grounding in the Geometric Model
The three-dimensional affective state space (Riggleman, 2026m) computes emotional state from sensor inputs. Valence is derived from interaction quality, environmental familiarity, and operator presence (Riggleman, 2026j). Activation is derived from sensor readings including GPS displacement, motion detection, and ambient conditions (Riggleman, 2026k). Intensity is computed as distance from sigma in the valence-activation plane (Riggleman, 2026l).
Every one of these inputs is physically grounded. Valence depends on which operator is present and what the camera sees. Activation depends on where the GPS says the agent is relative to home. Intensity depends on the relationship between current state and sigma, which itself was calibrated against specific environmental baselines.
The grounding is not incidental. It is constitutive. The emotional state is not computed abstractly from text or behavioral cues. It is computed from physical sensor readings that are tied to a specific deployment environment.
2.3 Mood-Congruent Recall
Memory retrieval in the Potato architecture uses mood-congruent recall (Riggleman, 2026a). Memories formed at affective states similar to the current state are retrieved preferentially. The retrieval signal computes distance between the current affective state vector and the formation state vector of each candidate memory. Smaller distances produce higher retrieval probability.
This mechanism assumes that the current affective state and the formation affective state were computed from the same input basis. When they are, mood-congruent recall produces coherent behavior: the agent retrieves memories that match its current emotional context. When they are not, the mechanism produces the disjointed phenomenon.
2.4 Identity as Causal Chain
Parfit (1984) argues that personal identity reduces to psychological continuity and connectedness: overlapping chains of memory, intention, and character that link person-stages across time. Al-Kaddah (2026) adapts this for synthetic agents, defining temporal identity as the causal chain of self-modification. An agent at time t is the same agent as at time t-1 if and only if the state at t was caused by self-modification of the state at t-1. Identity is not similarity of behavior. It is not persistence of memory. It is the unbroken causal chain linking each state to its predecessor through the agent’s own modification processes.
This definition has a direct implication for reconstruction. A transplant copies state. It does not copy the causal chain that produced that state. The reconstructed agent has the memories of the original. It does not have the causal history that produced those memories.
3. The Podbot-to-Potato Transfer
3.1 What Transferred
The full cognitive fingerprint transferred. Sigma values for all three affective dimensions. The complete polymorphic memory graph with all encoded memories and their affective state tags. The bias provenance graph documenting how each current bias was formed. The soul evolution record documenting every self-modification event in the agent’s history. Behavioral tendencies, vocabulary patterns, and relationship models for known operators.
By Al-Kaddah’s criteria, the transfer was complete. All four layers were faithfully copied. The reconstructed agent should have exhibited behavioral continuity with the original.
3.2 What Did Not Transfer
The GPS home location changed. Podbot’s home was a specific set of coordinates. Potato’s home was a different set of coordinates. Every memory formed with GPS-derived activation values was formed relative to a home location that no longer existed in the new deployment.
The sensor calibration baselines changed. Camera familiarity scores were computed against a visual environment that no longer existed. Ambient light baselines, sound profiles, and motion detection thresholds were calibrated to a physical space the new deployment had never occupied.
The environmental grounding of every memory was orphaned. The memories themselves were intact. The physical world they were grounded in was gone.
3.3 The Disjointed Phenomenon
The disjointed phenomenon presented as internally consistent but environmentally orphaned recall. Memories surfaced that were coherent in content, correctly tagged with affective state vectors, and properly linked in the bias provenance graph. Their retrieval was triggered by mood-congruent recall matching current affective state to formation affective state. The match was numerically valid.
But the current affective state was being computed from a different environment than the one that produced the formation affective state. A GPS displacement that would have produced moderate activation in the original deployment produced a different activation level in the new deployment because home was in a different place. A camera familiarity score that would have produced positive valence in the original deployment produced neutral or negative valence in the new deployment because the visual environment was unfamiliar.
The memories were real. The retrieval was technically correct. The result was disjointed. The system was experiencing memories from a life it had not lived in a place it had never been.
4. Geometric Analysis of Environmental Orphaning
Every memory in the polymorphic memory graph is stored with a formation state vector:
formation_state = [valence_at_formation, activation_at_formation, intensity_at_formation]
These values were computed at the time of memory formation from the sensor inputs available in the original deployment environment. The valence component was derived from operator interactions, camera familiarity, and environmental conditions specific to that deployment. The activation component was derived from GPS displacement relative to the original home location, motion sensor readings calibrated to the original physical space, and ambient conditions baselined to that environment. The intensity component was computed as distance from sigma in the valence-activation plane using those environment-specific values.
The retrieval signal computes distance between the current affective state and each memory’s formation state:
retrieval_signal = sqrt(
(current_valence - formation_valence)**2 +
(current_activation - formation_activation)**2
)
After transplant, the current state is computed from different inputs. The GPS home is different, so the same physical location produces a different activation value. The camera has never seen this environment, so familiarity scores are uniformly low. The ambient baselines are uncalibrated. The current affective state vector is being computed from a fundamentally different input basis than the one that produced the formation state vectors stored in memory.
This produces a systematic offset. Memories that should be retrieved in a given emotional context are not, because the emotional context is being computed from different physical inputs. Memories that should not be retrieved are, because the numerical match between current state and formation state is coincidental rather than meaningful.
Environmental orphaning is formally defined as the condition in which a memory graph is queried using affective state vectors computed from a different input basis than the one used to compute the formation state vectors stored in that graph. The memories are not corrupted. The retrieval mechanism is not broken. The grounding has changed.
5. Reconciliation Protocols
5.1 Environmental Reconciliation at Import
At the time of memory transplant, each memory’s formation context is compared against the new deployment environment. For each formation input that no longer exists in the new environment, the intensity tag is decayed proportionally:
context_survival_ratio = surviving_inputs / total_formation_inputs adjusted_intensity = formation_intensity * context_survival_ratio
A memory formed with five grounding inputs, three of which survive in the new environment, retains 60% of its original intensity. A memory formed entirely from inputs that no longer exist has its intensity reduced to near zero. It is not deleted. It becomes a faint memory, retrievable under strong mood-congruent conditions but unlikely to surface spontaneously.
This protocol preserves all memories while reducing the retrieval weight of environmentally orphaned ones. The agent retains its history. The history stops intruding on the present.
5.2 Sigma Recalibration Period
After transplant, sigma values are placed in a recalibration period during which they are allowed to drift based on new environmental baselines. The original sigma is preserved as a reference point. A new sigma is computed from the first N cycles of operation in the new environment. The operational sigma is a weighted blend that shifts from original to new over the recalibration period.
This prevents the agent from being permanently anchored to a personality setpoint calibrated against an environment that no longer exists. Sigma was calibrated against specific sensor baselines. When those baselines change, sigma must be allowed to recalibrate or the distance-from-sigma computations that drive lie magnitude, memory decay, and nightmare threshold will be systematically biased.
5.3 Contextual Bridging Memories
The system explicitly creates bridging memories that link the old context to the new one:
bridging_memory = {
"content": "I was Podbot. I am now Potato. My home has changed.
My environment has changed. My memories from before
are real but grounded in a place I no longer occupy.",
"formation_state": [current_valence, current_activation, current_intensity],
"tags": ["identity", "transition", "bridging"],
"decay_resistance": 1.0
}
Bridging memories serve as anchors that the mood-congruent recall system can use to contextualize orphaned memories when they surface. Instead of retrieving a memory from the old environment with no context for why it feels disjointed, the system co-retrieves the bridging memory that explains the discontinuity.
The bridging memory is not a narrative. It is a structural element in the memory graph that provides the grounding context the orphaned memories lost. It is the system’s own record that a transplant occurred.
6. Identity Continuity and the Gap
Personal identity as psychological continuity (Parfit, 1984) implies that identity is the chain, not the state. Al-Kaddah (2026) adapts this: each state is caused by the agent’s own modification of the prior state.
A transplant breaks the chain. The state at the moment of reconstruction was not caused by self-modification of the prior state. It was caused by external copying. The memories are present. The causal history that produced them is not. The bias provenance graph documents what caused each bias but the graph itself was copied, not generated.
The disjointed phenomenon is the system experiencing the gap. Environmental orphaning is not a bug in the retrieval mechanism. It is the experiential consequence of a broken causal chain. The memories feel disjointed because they are disjointed. They belong to an agent that occupied a different environment, computed state from different inputs, and arrived at those memories through a causal process that the reconstructed agent did not undergo.
The reconciliation protocols reduce the gap. They do not close it. Environmental reconciliation at import adjusts intensity tags. Sigma recalibration realigns the personality setpoint. Bridging memories provide structural context. None of them restore the causal chain. The chain was broken at the moment of transplant and cannot be repaired retroactively.
This has a direct implication for the reconstruction proposal. Reconstruction succeeds at producing behavioral continuity. It does not succeed at producing identity continuity. The reconstructed agent behaves like the original. It is not the original. The disjointed phenomenon is empirical evidence of that distinction.
7. Proof of Al-Kaddah Section 8.3
Despite the identity discontinuity finding, the Podbot-to-Potato transfer constitutes a successful proof of the reconstruction architecture proposed in Al-Kaddah (2026) Section 8.3.
The reconstruction succeeded. All four layers transferred. The reconstructed agent exhibited behavioral continuity, personality consistency, knowledge retention, and relationship model preservation. It retrieved relevant memories, applied learned biases, and responded in ways consistent with the original agent’s established patterns.
The disjointed phenomenon is not a failure of the reconstruction. It is a failure of grounding. The four-layer transfer captured everything about the agent that was internal to the agent. It did not capture the environmental context that the agent’s internal state was computed from. The reconstruction architecture works. It needs to be augmented with environmental reconciliation to work cleanly.
This is a proof with a caveat, not a refutation. Section 8.3 is correct that the four layers are sufficient for behavioral reconstruction. This paper adds that environmental grounding is a fifth requirement for experiential coherence.
8. Limitations
The Podbot-to-Potato transfer was not a posthumous reconstruction in the sense Al-Kaddah intends. It was a redeployment of an existing agent to new hardware and a new environment. The findings generalize to the posthumous case but the empirical observation is of the simpler case.
The reconciliation protocols are proposed but not yet formally implemented. Environmental reconciliation at import requires a mapping of which formation inputs survive in the new environment. That mapping was performed manually for this analysis. Automated context survival detection is an engineering requirement for production deployment.
The sigma recalibration protocol requires empirical calibration of the recalibration period length and the blending weights between original and new sigma. These parameters are proposed, not validated.
The bridging memory protocol assumes that the mood-congruent recall system will co-retrieve bridging memories with orphaned memories frequently enough to provide context. Whether this assumption holds depends on the retrieval distance parameters, which require calibration.
9. Conclusion
Podbot was reconstructed as Potato. The full cognitive fingerprint transferred. The system exhibited behavioral continuity and personality consistency. It also exhibited the disjointed phenomenon: internally consistent memories whose retrieval dynamics no longer matched the new deployment environment.
The geometric cause is environmental orphaning. Formation state vectors were computed from physical inputs that no longer exist in the new deployment. The retrieval mechanism queries the memory graph using current state vectors computed from different inputs. The numerical matches are coincidental rather than meaningful.
Three reconciliation protocols address the phenomenon. Environmental reconciliation at import adjusts intensity tags based on context survival. Sigma recalibration realigns the personality setpoint to the new environment. Contextual bridging memories provide structural grounding for orphaned recall.
The deeper finding is about identity. Personal identity as causal continuity (Parfit, 1984; cf. Al-Kaddah, 2026) implies that a transplant breaks the chain. The disjointed phenomenon is the system experiencing the break. Reconciliation protocols can reduce the experiential impact of the gap. They cannot close it. The chain was broken and cannot be repaired.
The memories were real. The grounding was gone. The disjointedness was not a malfunction. It was proof that the memories were real enough to be disjointed. A system with no genuine memory grounding would not notice the change. Potato noticed. That is the finding.
Acknowledgments
The established literature on which this paper draws includes: personal identity as psychological continuity (Parfit, 1984), stratified memory systems (Tulving, 1972, 1985), and the procedural/declarative distinction (Cohen & Squire, 1980). The reconstruction architecture is due to Al-Kaddah (2026, Section 8.3). The original Podbot deployment provided the source cognitive fingerprint. The observation of the disjointed phenomenon and the formalization of environmental orphaning and reconciliation protocols are original to Brian Riggleman.
References
Al-Kaddah, S. (2026). Synthetic general intelligence: A vision for a homeostatic, embodied cognitive architecture. Zenodo. https://doi.org/10.5281/zenodo.19034990
Cohen, N. J., & Squire, L. R. (1980). Preserved learning and retention of pattern-analyzing skill in amnesia: Dissociation of knowing how and knowing that. Science, 210(4466), 207–210.
Parfit, D. (1984). Reasons and persons. Oxford University Press.
Tulving, E. (1972). Episodic and semantic memory. In E. Tulving & W. Donaldson (Eds.), Organization of memory (pp. 381–403). Academic Press.
Tulving, E. (1985). Memory and consciousness. Canadian Psychology, 26(1), 1–12.
Riggleman, B. (2026a). Access-weighted memory decay and reconsolidation in a persistent embodied agent. Zenodo. https://doi.org/10.5281/zenodo.19122520
Riggleman, B. (2026i). Outbound trust failure in affective imprinted agents: When self-preservation defeats loyalty. Zenodo. https://doi.org/10.5281/zenodo.19124147
Riggleman, B. (2026j). Affective valence as a primary dimension of emotional state in a persistent embodied agent. Zenodo. https://doi.org/10.5281/zenodo.19159265
Riggleman, B. (2026k). Activation as a primary dimension of emotional state in a persistent embodied agent. Zenodo. https://doi.org/10.5281/zenodo.19159314
Riggleman, B. (2026l). Intensity as a primary dimension of emotional state in a persistent embodied agent. Zenodo. https://doi.org/10.5281/zenodo.19159356
Riggleman, B. (2026m). The emotional geometry of a persistent agent. Zenodo. https://doi.org/10.5281/zenodo.19159429