Giving stateless agents a sense of time: A reconsolidating episodic–semantic memory architecture

Stateless agents may retrieve past interactions yet lack an operational sense of time: external memory makes the past available, but does not represent how experience loses accessibility, returns to relevance, or changes meaning. This paper defines that capacity as an episodic–semantic lifecycle governed by importance, elapsed time, recall and evidence. Pragma, developed by the author, stores sessions as episodes whose narratives stay outside the rewrite target, and consolidates recurring regularities into source-linked beliefs. Deterministic salience and decay regulate accessibility; recall and evidence trigger versioned reconsolidation without overwriting recorded events. The architecture is examined in 229 executions spanning four scenarios, four model families and ten executions per condition. The deterministic layer behaves identically in every family: salience, reinforcement and decay reproduce exactly, no belief is admitted without two distinct sources, and the tracked episode returned to the active zone in all 45 executions where it became dormant. The judgements it consumes do not: the same estimate varies by an order of magnitude across families, and that variation alone can decide whether a designed transition occurs. Removing the lifecycle answers the same memory probes equally well at the store sizes these scenarios reach, which delimits where the difference can be sought. Code, stimuli, metadata and traces make every reported transition inspectable. The contribution is an operational definition of temporal continuity, an architecture implementing it, and evidence separating the properties that hold by construction from those that depend on the model. It is not a validation of reliability, semantic correctness or comparative performance.

Authors

Institutions

Publication Details

Journal
Knowledge-Based Systems
Published
2026-10-13
DOI
https://doi.org/10.1016/j.knosys.2026.117193
Primary Topic
AI-based Problem Solving and Planning
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Giving stateless agents a sense of time: A reconsolidating episodic–semantic memory architecture

Gian Maria Santi
Knowledge-Based Systems
AI-based Problem Solving and Planning
article

Giving stateless agents a sense of time: A reconsolidating episodic–semantic memory architecture

Gian Maria Santi
article en

Abstract

Stateless agents may retrieve past interactions yet lack an operational sense of time: external memory makes the past available, but does not represent how experience loses accessibility, returns to relevance, or changes meaning. This paper defines that capacity as an episodic–semantic lifecycle governed by importance, elapsed time, recall and evidence. Pragma, developed by the author, stores sessions as episodes whose narratives stay outside the rewrite target, and consolidates recurring regularities into source-linked beliefs. Deterministic salience and decay regulate accessibility; recall and evidence trigger versioned reconsolidation without overwriting recorded events. The architecture is examined in 229 executions spanning four scenarios, four model families and ten executions per condition. The deterministic layer behaves identically in every family: salience, reinforcement and decay reproduce exactly, no belief is admitted without two distinct sources, and the tracked episode returned to the active zone in all 45 executions where it became dormant. The judgements it consumes do not: the same estimate varies by an order of magnitude across families, and that variation alone can decide whether a designed transition occurs. Removing the lifecycle answers the same memory probes equally well at the store sizes these scenarios reach, which delimits where the difference can be sought. Code, stimuli, metadata and traces make every reported transition inspectable. The contribution is an operational definition of temporal continuity, an architecture implementing it, and evidence separating the properties that hold by construction from those that depend on the model. It is not a validation of reliability, semantic correctness or comparative performance.

Knowledge-Based SystemsVol. 353
University of Bologna (IT)
Openalex Percentile: Top 13%
AI-based Problem Solving and Planning
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.