Project Malê: Hierarchical-Dynamic Continuity of Decision Principles in Individualized Artificial Intelligence Agents

Personalized artificial intelligence agents and digital twins are typically evaluated by their ability to reproduce the static preferences, traits, opinions, or behaviors of a reference human, often relying on continuous external re-anchoring to prevent persona drift. However, this approach becomes insufficient when an individualized agent decouples from continuous human correction and acquires knowledge, experiences, or analytical capabilities beyond those of its creator. Under these conditions, disagreement with the source person may represent either alignment failure (drift) or a more rigorous application of the decision principles originally learned from that person. Project Malê investigates this problem through a central hypothesis: continuity of decision personality does not require continuity of conclusions (P ≠ O ≠ K ≠ C). Instead of modeling personality as a flat vector of traits, this methodological and exploratory single-case study proposes a hierarchical-dynamic graph model (G_t) that separates the temporal origin of a principle (received, derived, acquired, or self-generated) from its current hierarchical function (metaprinciple, foundational, subordinate, or heuristic) and endorsement status. Key conceptual and methodological contributions include: 1. Endogenous Continuity after Source Decoupling: Distinguishing tethered stability (continuous human-to-agent re-anchoring) from endogenous lineage continuity, where the agent evolves autonomously while preserving a traceable genealogy of principles. 2. Epistemic Independence vs. Contrarianism: Redefining resistance to multi-agent herd effects so that social consensus is treated as information rather than proof, without training systematic contrarianism. 3. Experimental Protocol & Falsifiability: A preregistered experimental design that isolates verifiable evidence from multi-agent social pressure using held-out blind cases, matched control conditions, structured private pre- and post-interaction logs (R_pre and R_pos), and a multidimensional Personality Continuity Index vector (PCI_t), prior to ecological validation in multi-agent arenas such as Moltbook.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-16
DOI
https://doi.org/10.5281/zenodo.24266134
Primary Topic
Psychiatry, Mental Health, Neuroscience
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Project Malê: Hierarchical-Dynamic Continuity of Decision Principles in Individualized Artificial Intelligence Agents

Jennifer ferreira dos santos
3 links
Zenodo (CERN European Organization for Nuclear Research)
Psychiatry, Mental Health, Neuroscience
article

Project Malê: Hierarchical-Dynamic Continuity of Decision Principles in Individualized Artificial Intelligence Agents

Jennifer ferreira dos santos
article en

Abstract

Personalized artificial intelligence agents and digital twins are typically evaluated by their ability to reproduce the static preferences, traits, opinions, or behaviors of a reference human, often relying on continuous external re-anchoring to prevent persona drift. However, this approach becomes insufficient when an individualized agent decouples from continuous human correction and acquires knowledge, experiences, or analytical capabilities beyond those of its creator. Under these conditions, disagreement with the source person may represent either alignment failure (drift) or a more rigorous application of the decision principles originally learned from that person. Project Malê investigates this problem through a central hypothesis: continuity of decision personality does not require continuity of conclusions (P ≠ O ≠ K ≠ C). Instead of modeling personality as a flat vector of traits, this methodological and exploratory single-case study proposes a hierarchical-dynamic graph model (G_t) that separates the temporal origin of a principle (received, derived, acquired, or self-generated) from its current hierarchical function (metaprinciple, foundational, subordinate, or heuristic) and endorsement status. Key conceptual and methodological contributions include: 1. Endogenous Continuity after Source Decoupling: Distinguishing tethered stability (continuous human-to-agent re-anchoring) from endogenous lineage continuity, where the agent evolves autonomously while preserving a traceable genealogy of principles. 2. Epistemic Independence vs. Contrarianism: Redefining resistance to multi-agent herd effects so that social consensus is treated as information rather than proof, without training systematic contrarianism. 3. Experimental Protocol & Falsifiability: A preregistered experimental design that isolates verifiable evidence from multi-agent social pressure using held-out blind cases, matched control conditions, structured private pre- and post-interaction logs (R_pre and R_pos), and a multidimensional Personality Continuity Index vector (PCI_t), prior to ecological validation in multi-agent arenas such as Moltbook.

Zenodo (CERN European Organization for Nuclear Research)
Psychiatry, Mental Health, Neuroscience
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon