[Advanced] Recall Mechanisms

Two distinct recall methods that enable both immediate contextual access and targeted historical analysis for clinical intelligence

Recall mechanisms are fundamental to Amigo's functional clinical intelligence, where L3 (the global user model) provides memory at the right interpretation, precision, and depth to power knowledge application and reasoning. The system employs two distinct approaches to memory access within the unified Memory-Knowledge-Reasoning (M-K-R) framework:

Overview

The dual recall mechanism ensures optimal performance by balancing immediate access to comprehensive context with the ability to perform deep historical analysis when needed.

Rare Recontextualization (Adds Latency)

Triggered by: Detection of genuinely new context that requires historical perspective beyond what's available in L3 (constantly held in memory).

1

Dual Anchoring Process

L3 serves as anchor point for both expansion query generation and historical interpretation

2

Query Generation

Expansion queries generated with respect to current L3 state against present live session context

3

Historical Retrieval

System reasons over past n L0 sessions plus relevant sessions identified through expansion targets

4

Anchored Interpretation

Query answering from historical L0 is anchored against L3, merging past raw events with present global understanding across time

5

Temporal Coherence

Historical events understood through lens of complete current patient understanding (L3) rather than isolated past context

Why It's Rare: L3 provides comprehensive coverage of all functionally important dimensions at the desired depth, precision, and interpretation. L3's constant availability means recontextualization is only needed when genuinely new contextual relationships emerge that require deep historical analysis.

When recontextualization is needed, the professional identity framework ensures targeted, clinically relevant expansion rather than broad searches.

Why Dual Anchoring Prevents Information Loss

The dual anchoring mechanism solves a critical problem: how to retrieve and interpret historical information without losing context at boundaries.

Query generation with L3 anchoring: When the system needs historical context, it generates expansion queries conditioned on L3's current functional dimensions. This ensures queries target outcome-relevant information rather than generic searches. The query "tell me about medication adherence" becomes "retrieve adherence patterns related to stress-medication timing interaction" because L3 knows this dimension matters for this patient.

Historical interpretation with L3 anchoring: Retrieved L0 conversations get interpreted through L3's current understanding. A patient saying "I felt stressed at work" three months ago gets understood in context of the now-discovered stress-medication dimension, not as isolated past event.

Boundary loss prevention: Without anchoring, merging past and present context loses information where they meet. L3 anchoring maintains coherence by ensuring both retrieval (what to get) and interpretation (what it means) use the same dimensional framework. This is the same principle as L2→L3 merging: balance finding new patterns with preserving current understanding.

L3 Constant Access (90-95% of Cases)

Triggered by: All standard clinical interactions where L3 (the global user model) provides immediate access to required context.

Core Capabilities

• L3 remains constantly in memory during live sessions

• All functionally important dimensions available at desired depth, precision, and interpretation

• Professional identity guides interpretation at every level without additional retrieval

Clinical Application

Achieving Functional Clinical Intelligence

This approach achieves functional clinical intelligence because L3 provides memory at the precise interpretation depth required for clinical knowledge application and reasoning with immediate availability.

Unified Context

The unified context enables high-quality reasoning because memory, knowledge application, and reasoning operate on consistently interpreted information without retrieval latency.

Information Determination

Contextual Coherence

The Complete Architecture

This architecture ensures both:

• ✅ Immediate access to comprehensive patient context (L3 constant availability)

• ✅ Targeted historical analysis when genuinely new contextual relationships emerge (rare recontextualization)

Creating functional clinical intelligence optimized for medical performance within the unified M-K-R framework.

The system creates multiple interconnected feedback loops between global patient understanding and local processing, where professional identity-driven interpretation prevents clinical misinterpretation at every level. This transforms medical AI from simple information retrieval into true clinical intelligence that maintains perfect contextual awareness across all patient interactions.