Skip to content
#explainable ai Open access

The Report Card of 500 Institutions: Why Modern Systems Have Files but No Memory

Sep 2026 · Zenodo (CERN European Organization for Nuclear Research)

Abstract

Institutional Intelligence, Living Experiments & Traceable Accountability The Report Card of 500 Institutions: Why Modern Systems Have Files but No Memory Publication Number: #143Corpus: Longitudinal Institutional Intelligence Research Corpus (LIIRC)Resource Type: Publication / Working Paper / Institutional BenchmarkLicense: Creative Commons Attribution 4.0 International (CC BY 4.0) Keywords: Institutional Intelligence, Longitudinal Data, Traceability, Citizen Auditing, Administrative Intelligence, Institutional Learning, AI Governance, Institutional Memory, Cross-Event Correlation, Accountability Abstract For months, signals, notices, requests, and documented communications were introduced into a large and diverse administrative ecosystem. Each interaction generated observable traces: delivery confirmations, acknowledgments, tracking numbers, case identifiers, automated responses, substantive responses, routing events, follow-ups, and, in some cases, observable changes in subsequent behavior. At the 500-institution scale, a central research question emerged: Does the accumulation of institutional records produce institutional learning? A single institutional interaction can often be explained by local procedural circumstances. A longitudinal corpus allows a different level of observation. By comparing repeated signals across hundreds of institutional endpoints and over extended periods, the research examines whether previously recorded events become connected to later events and whether such connections produce observable changes. The central distinction developed in this publication is: Recording is not necessarily remembering.Remembering is not necessarily connecting.Connecting is not necessarily learning.Learning is not necessarily behavioral adaptation. The research therefore does not attempt to infer hidden institutional states. Instead, it establishes an observable framework for examining the transition from information receipt to contextual understanding, cross-event correlation, institutional learning, and subsequent action. The result is a longitudinal research model for examining institutional intelligence through traceable external evidence. 1. The Great Illusion: Receive ≠ Understand Modern institutions are highly capable of detecting incoming information. They generate: delivery confirmations, ticket numbers, case codes, acknowledgment messages, automated replies, routing notifications, archival records, and procedural closure notices. These mechanisms demonstrate that information has entered an administrative system. However: Receipt does not establish comprehension. A message reaching an institutional endpoint demonstrates transmission and, under appropriate evidence, successful receipt. It does not independently demonstrate that the information was understood in its broader context, connected with previous events, or incorporated into subsequent institutional reasoning. This distinction forms the first methodological boundary of the research. Observable proposition Receive = evidence that an information transfer occurred. Non-equivalent proposition Understand ≠ automatically established by successful receipt. The research therefore separates the existence of an information trace from claims about what an institution internally understood. 2. The Procedural Escape Hatch: Response ≠ Learning Administrative systems frequently operate through a recognizable sequence: Receive → Classify → Respond → Close This sequence can be highly efficient from a procedural perspective. A response may contain: a case reference, a policy citation, a jurisdictional explanation, a referral, a standard template, a request for additional information, or a formal closure. The existence of such a response establishes that an institutional process occurred. It does not, by itself, establish institutional learning. A central research distinction is therefore: A response is an event. Learning is a change across events. Institutional learning becomes observable only when a later interaction demonstrates a meaningful relationship with previously documented information. 3. The Missing Link: Institution ≠ Institutional Destination Large organizations frequently consist of multiple administrative layers, departments, systems, databases, legal units, technical teams, regional offices, and external interfaces. Consequently, sending information to an institution does not necessarily mean that the information reaches every organizational component capable of acting upon it. A signal may enter through one administrative endpoint while a related event exists elsewhere. The research therefore distinguishes: Institution from Institutional Destination and records the observable path of information whenever that path can be established. The key question is not simply: "Was the institution contacted?" but: "What traceable path, if any, connected the signal to a subsequent institutional event?" 4. The Repeated Signal Test Repeated signals provide an important longitudinal observation mechanism. Where appropriate and ethically permissible, repeated documented notifications can be examined across time: Event₁ → Event₂ → Event₃ → … → Eventₙ The research asks whether repetition produces an observable change in: routing, acknowledgment, contextual reference, response content, escalation, case linkage, procedural treatment, or subsequent institutional behavior. The fundamental research question is: Does repeated exposure to a traceable signal produce an observable change in subsequent institutional behavior? Importantly, repetition alone does not prove that an institution failed to learn. It provides an opportunity to observe whether evidence of learning or adaptation becomes externally visible. 5. The 500-Institution Administrative Report Card The 500-institution corpus provides a comparative longitudinal environment for examining institutional behavior. The dimensions below are analytical categories rather than claims of direct access to internal institutional states. Dimension Observable Interpretation Information Detection Evidence that an incoming signal was received or entered an intake mechanism Procedural Response Evidence of an acknowledgment, reply, classification, or procedural action Archiving & Recording Presence of a reference, case number, record, or other trace Context Awareness Explicit recognition of relevant surrounding context Routing Visibility Observable evidence of transfer, escalation, or departmental routing Cross-Event Correlation Explicit linkage between separate events Institutional Learning Observable change following previously documented signals Behavioral Adaptation Observable modification in subsequent institutional behavior The purpose of this table is not to assign an absolute institutional score. It establishes a common observation vocabulary through which longitudinal institutional behavior can be compared. 6. The Missing Dimension: Time and Relationship A database can contain Event A. A database can contain Event B. The existence of both records does not establish that the institution recognizes: Event A → Event B as elements of a common longitudinal sequence. This distinction is fundamental. The research therefore identifies the following progression: Information → Record → Context → Chronology → Relationship → Significance The critical loss may occur between the existence of individual records and the recognition of their relationship. This is why institutional memory cannot be evaluated solely by asking whether records exist. A more meaningful question is: Can previously recorded information demonstrably influence the interpretation or handling of subsequent events? 7. Institutional Intelligence as an Observable Process For the purposes of this corpus, institutional intelligence is not defined as intelligence in the psychological sense. It is treated as an observable organizational capacity to transform accumulated information into progressively more context-aware and traceable subsequent action. The research model is: Receive → Understand → Connect → Remember → Learn → Act Each stage represents a distinct analytical question. Receive Did the signal reach an identifiable institutional endpoint? Understand Is there observable evidence that the relevant subject matter was recognized? Connect Was the event explicitly related to previous or parallel events? Remember Was previously documented information incorporated into a subsequent interaction? Learn Did subsequent institutional behavior demonstrate an observable change associated with prior information? Act Did the institution subsequently modify, escalate, redirect, resolve, or otherwise alter its behavior? The critical methodological transition is therefore: Recording → Connecting 8. Operational Definition of Institutional Learning The research does not claim direct access to institutional consciousness, internal deliberation, or hidden databases. For this corpus: Institutional learning is operationally defined as an observable change in subsequent institutional behavior that can be temporally associated with previously documented signals or events. Potential observable indicators include: explicit reference to an earlier event; linkage of previously separate case identifiers; altered routing; changed response content; escalation to another responsible unit; modification of procedural treatment; acknowledgement of accumulated context; documented policy or process modification; or another externally observable change that follows an established sequence of prior signals. The presence of such an indicator does not prove the complete internal mechanism that produced it. It establishes an observable longitudinal relationship. 9. The Traceabil

View source

Similar papers

#artificial intelligence Conference Open access Apr 2020

ECCOLA - a Method for Implementing Ethically Aligned AI Systems

The method, ECCOLA, is presented, which aims at making the high-level AI ethics principles more practical, making it possible for developers to more easily implement them in practice.

Ville Vakkuri, Kai-Kristian Kemell, P. Abrahamsson · 64 citations · ⚡6
#computer vision Review Apr 2024

AI-powered Code Review with LLMs: Early Results

The goal is to not only refine the accuracy of the LLM-based tool but also to underscore its potential in streamlining the software development lifecycle through proactive code improvement and education.

Z. Rasheed, Malik Abdul Sami, Muhammad Waseem et al. · 62 citations · ⚡3
#computer vision Open access Mar 2024

LLM-based agents for automating the enhancement of user story quality: An early report

The use of large language models to automatically improve the user story quality in Austrian Post Group IT agile teams is explored, with a reference model for an Autonomous LLM-based Agent System developed and implemented at the company.

Zheying Zhang, M. Rayhan, Tomas Herda et al. · 48 citations · ⚡4
#computer vision Review Mar 2024

System for systematic literature review using multiple AI agents: Concept and an empirical evaluation

This paper introduces a novel multi-AI-agent system designed to fully automate SLRs, and demonstrates how it substantially reduces the time and effort traditionally required for SLRs while maintaining comprehensiveness and precision.

Abdul Malik Sami, Z. Rasheed, Kai-Kristian Kemell et al. · 44 citations · ⚡2
#computer vision Feb 2024

Can Large Language Models Serve as Data Analysts? A Multi-Agent Assisted Approach for Qualitative Data Analysis

The proposed LLM-based multi-agent system automates qualitative data analysis process, creating opportunities for researchers and practitioners, and future improvements focus on enhancing multilingual performance and integrating continuous expert feedback.

Z. Rasheed, Muhammad Waseem, Aakash Ahmad et al. · 41 citations
#artificial intelligence Conference Open access Jun 2018

The Key Concepts of Ethics of Artificial Intelligence

It is suggested that the focus on finding keywords is the first step in guiding and providing direction for future research in the AI ethics field.

Ville Vakkuri, P. Abrahamsson · 39 citations · ⚡2

Related blog posts

Microsoft Research Blog Oct 7, 2026

Agent Lightning v1.0: A 3,500-Line Lightweight Agentic RL Framework for Training Agents with Real Harnesses

Training AI agents with reinforcement learning can be challenging because their tools, context, and decision-making are managed by complex frameworks. Agent Lightning connects existing agents to RL training, making it easier to improve them without rebuilding them. The post Agent Lightning v1.0: A 3,500-Line Lightweight Agentic RL Framework for Training Agents with Real Harnesses appeared first on Microsoft Research.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.