The Crisis of Entity Resolution in Heterogeneous AI Environments: Data, Financial, and Semantic Dimensions

Daniel Beránek

doi:10.6084/m9.figshare.33420025

AI summary · DOI: 10.6084/m9.figshare.33420025 · Decision Sciences

The Crisis of Entity Resolution in Heterogeneous AI Environments: Data, Financial, and Semantic Dimensions

Daniel Beránek

Original paper Published on 02 Sep. 2026 Figshare preprint License: CC BY

How to read this summary? Illustrated guide

The essentials

The preprint synthesizes entity-resolution risks in heterogeneous AI systems and proposes DIO/ODI as a testable, not yet validated, framework for digital identity integrity.

Based on the abstract only: check the details in the original paper.

Question

How do entity-resolution failures across data, time, and context affect AI systems such as retrieval-augmented generation and multi-agent systems? Can a proposed digital-identity framework help conceptualize integrity across these settings?

Method

The paper synthesizes research on entity linking and matching, evolving knowledge bases, retrieval-augmented generation, and multi-agent LLM systems alongside public evidence on enterprise data quality and AI readiness. It organizes the discussion around four anomalies and presents DIO/ODI as a theoretical integrity hypothesis, not as an empirically validated solution.

Findings

Authors’ interpretation The paper identifies identity collapse under name ambiguity and popularity bias as one of four connected anomalies.
See the passage in the paper
It identifies four interconnected anomalies: identity collapse under name ambiguity and popularity bias; semantic fragmentation across data silos; temporal discontinuity when entities change while indexes remain static; and semantic vulnerability when retrieved context is incomplete, conflicting, or manipulated.
Authors’ interpretation The paper argues that entity-resolution failures in RAG and agentic systems can propagate across multiple stages.
See the passage in the paper
In RAG and agentic systems, these failures may propagate from indexing and retrieval through generation to subsequent tool calls and apparent verification.
Result DIO/ODI are proposed as a testable framework rather than a validated technology or ready-made remedy.
See the passage in the paper
DIO/ODI are not presented as empirically validated technology or as a ready-made solution to entity resolution, temporal linking, or RAG security.
Authors’ interpretation The proposed framework concerns coherent reconstruction of digital identity across multiple dimensions.
See the passage in the paper
They are proposed as a testable integrity hypothesis for the coherent reconstruction of digital identity across people, systems, languages, time, and contexts.

Limitations

The paper does not empirically validate DIO/ODI as a technology or solution to the identified problems.
See the passage in the paper
DIO/ODI are not presented as empirically validated technology or as a ready-made solution to entity resolution, temporal linking, or RAG security.
The supplied source is an abstract, so its full methods and supporting evidence cannot be assessed here.

Why it matters

The authors frame entity resolution as a cross-cutting integrity problem for heterogeneous AI, with failures that can affect retrieval, generation, and tool use. They propose DIO/ODI as a testable conceptual basis for examining coherent digital identity across systems and contexts.

Written by AI from the paper. Passages are checked word for word against the text we analysed; the rest may contain errors.

AI-generated keywords: Entity resolution; Digital identity; Retrieval-augmented generation; Knowledge graphs; AI systems

Read further

Comprehensive Summary
Layman's Summary
Blog article
Original abstract

The preprint examines whether AI systems can reliably recognize and reconstruct the same entity—such as a person, organization, or concept—across sources, time periods, languages, and contexts. It argues that this entity-resolution problem deserves greater attention as AI systems combine large, heterogeneous datasets and increasingly rely on retrieval-augmented generation (RAG) and interacting language-model agents. The paper describes this shift as an “entity turn”: moving beyond isolated tokens, documents, and embeddings to consider whether systems preserve identity consistently. Rather than reporting a controlled experiment, the paper synthesizes existing research on entity linking and matching, evolving knowledge bases, RAG, and multi-agent LLM systems. It also draws on public enterprise evidence about data quality and AI readiness. The supplied abstract specifies no dataset, experimental protocol, or quantitative analysis, so the work should not be interpreted as a benchmark study with measured performance results. The paper identifies four related failure patterns: 1. **Identity collapse:** Ambiguous names or a bias toward popular entities can lead systems to conflate distinct entities or miss the intended one. 2. **Semantic fragmentation:** Information spread across data silos can leave systems with an incomplete or inconsistent understanding of an entity. 3. **Temporal discontinuity:** Entities change over time, but static indexes or knowledge resources may not reflect their current state. 4. **Semantic vulnerability:** Retrieved context may be incomplete, contradictory, or manipulated. These failures can compound in RAG and agentic systems. Errors introduced during indexing or retrieval may shape generated answers, influence subsequent tool calls, and create an appearance of verification without reliably reconstructing identity. The final section introduces Digital Identity Optimization (DIO) and the Ontology of Digital Identity (ODI) as a theoretical framework for governing digital-identity integrity. The abstract presents these ideas as a testable integrity hypothesis: a proposal for examining whether identity can be coherently reconstructed across people, systems, languages, times, and contexts. It does not claim that DIO or ODI is empirically validated or offers a ready-made solution to entity resolution, temporal linking, or RAG security. Overall, the paper argues that reliable entity resolution is a foundational challenge for heterogeneous AI systems, not merely a data-cleaning task. It recommends assessing how well systems preserve identity across changing, distributed, and potentially adversarial information environments, particularly when outputs trigger further actions. As represented in the supplied abstract, the paper’s contribution is primarily conceptual and synthetic: it organizes relevant concerns into four failure patterns and proposes a framework for future testing. The assessment is limited because only the title and abstract were available; the full argument, cited evidence, and detailed methods therefore cannot be evaluated. The supplied material reports no numerical results, effect sizes, or benchmark comparisons. DIO and ODI also remain unvalidated, and their practical effectiveness would require empirical evaluation.

Why AI Needs to Know Who—or What—it Is Talking About

AI systems increasingly draw on information scattered across databases, websites, documents, and languages. But bringing more information together does not automatically make it coherent. A system may encounter the same person under different names, confuse two organizations, or rely on a description that was accurate years ago but is now out of date.

In The Crisis of Entity Resolution in Heterogeneous AI Environments: Data, Financial, and Semantic Dimensions, Daniel Beránek argues that these problems deserve more attention. The paper’s central concern is entity resolution: whether an AI system can recognize and reconstruct the same entity across sources, time periods, languages, and contexts.

An “Entity Turn” in AI

Much AI research focuses on tokens, documents, or the representations used to compare them. Beránek calls for an “entity turn”: a shift toward asking whether systems preserve identity consistently as information moves between datasets and applications.

This matters especially as AI systems adopt retrieval-augmented generation (RAG), which brings external information into a model’s answers, and agentic workflows, where language-model systems interact with tools or with one another. In these settings, an error about identity may travel beyond a single response. Retrieved information can shape an answer, inform a tool call, and influence later steps.

The paper is a conceptual synthesis, not a controlled experiment. It brings together concerns from entity linking and matching, changing knowledge bases, RAG, and multi-agent systems, alongside public enterprise evidence about data quality and AI readiness. The available abstract reports no dataset, experimental protocol, quantitative analysis, or benchmark results. Its claims should therefore be read as an argument about a foundational challenge—not as measured proof of how often particular systems fail.

Four Ways Identity Can Break Down

Beránek organizes the problem into four related failure patterns:

Identity collapse: Ambiguous names or preference for well-known entities can cause a system to confuse different people or organizations—or fail to find the intended one.
Semantic fragmentation: When information is split across organizational or technical silos, a system may form an incomplete or inconsistent picture of an entity.
Temporal discontinuity: People, organizations, and other entities change. Static indexes or knowledge resources can preserve outdated details and miss those changes.
Semantic vulnerability: Retrieved information may be incomplete, contradictory, or manipulated, making it an unreliable basis for identifying or describing an entity.

These patterns can overlap. A fragmented record may also be stale; an ambiguous name may be paired with misleading context. In a RAG system, the resulting answer can sound well-supported even when the information retrieved does not establish that it refers to the right entity. In a multi-step system, that uncertain answer may then steer further searches or actions.

Why This Is More Than Data Cleaning

Entity resolution is sometimes treated as a back-end housekeeping task: standardize names, remove duplicates, and reconcile records before an AI system uses them. Beránek’s framing suggests a wider problem. Identity must remain coherent across different data sources, changing circumstances, languages, and uses—and it may need to withstand incomplete or adversarial information.

That makes the issue important not only for answer quality but also for systems that act on their outputs. If a generated response triggers a follow-up search, a database update, or another tool call, an identity error can become part of a chain of decisions. The more a system relies on retrieved context and interacting components, the more useful it is to ask whether identity has actually been reconstructed rather than merely assumed.

DIO and ODI: A Proposal to Test

In its final section, the paper introduces Digital Identity Optimization (DIO) and the Ontology of Digital Identity (ODI) as a theoretical framework for governing digital-identity integrity. The abstract presents these ideas as a testable integrity hypothesis: a way to examine whether identity can be coherently reconstructed across people, systems, languages, times, and contexts.

That distinction is important. The available material does not establish that DIO or ODI has been empirically validated, nor does it present them as ready-made solutions to entity resolution, temporal linking, or RAG security. Their practical value would need to be assessed through future testing.

A Useful Agenda, with Evidence Still to Come

The paper’s contribution, as described in its abstract, is to gather related risks into a shared framework and argue that identity preservation should be treated as a core challenge for heterogeneous AI. It encourages researchers and system designers to assess how well AI maintains identity as information changes, spreads across sources, or becomes potentially adversarial—especially when outputs can trigger further actions.

There are limits to what can be concluded. Only the title and abstract were available for this assessment, so the full argument, cited evidence, and detailed methods cannot be evaluated. The supplied material reports no numerical findings or benchmark comparisons, and the proposed DIO and ODI framework remains unvalidated. Still, the paper poses a timely question for AI development: when a system retrieves and combines information, how do we know it has kept track of the same entity?

AbstractThis preprint examines the crisis of entity resolution in heterogeneous AI environments across data, financial, and semantic dimensions. It develops an entity turn in AI: a shift from isolated tokens, documents, and embeddings toward the question of whether systems can reliably identify and reconstruct the same entity across sources, time, languages, and contexts.The paper synthesizes research on entity linking, entity matching, temporally evolving knowledge bases, retrieval-augmented generation (RAG), and multi-agent LLM systems with public enterprise evidence on data quality and AI readiness. It identifies four interconnected anomalies: identity collapse under name ambiguity and popularity bias; semantic fragmentation across data silos; temporal discontinuity when entities change while indexes remain static; and semantic vulnerability when retrieved context is incomplete, conflicting, or manipulated. In RAG and agentic systems, these failures may propagate from indexing and retrieval through generation to subsequent tool calls and apparent verification.The final section examines Digital Identity Optimization (DIO) and the Ontology of Digital Identity (ODI) as an author-formulated theoretical framework for governing the integrity of digital identity. DIO/ODI are not presented as empirically validated technology or as a ready-made solution to entity resolution, temporal linking, or RAG security. They are proposed as a testable integrity hypothesis for the coherent reconstruction of digital identity across people, systems, languages, time, and contexts.Record and version Version 1.0.3 (Figshare); corresponds to Version 1.0.0 of the canonical Zenodo record.This Figshare item distributes the same preprint as the canonical Zenodo record. For archival reference and citation, use: Beránek, D. (2026). *The Crisis of Entity Resolution in Heterogeneous AI Environments: Data, Financial, and Semantic Dimensions*. Zenodo. https://doi.org/10.5281/zenodo.22257504A companion version of this manuscript is also available on SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7396398AI disclosureThis manuscript was developed through a human-led process of research design, source selection, comparative AI-assisted exploration, and substantive authorial revision. Manus AI was used as a synthesis and drafting instrument on the basis of the author's curated inputs and continuous revisions. Other AI-assisted outputs served only as provisional analytical material and were incorporated only after authorial selection, verification against cited sources where applicable, and further revision. Perplexity AI additionally supported the technical and administrative preparation of this record and its distribution, including citation formatting, metadata structuring, and platform-specific publication procedures across SSRN and Zenodo. The DIO/ODI framework and all critical intellectual decisions remain the sole work of the author, who takes full responsibility for the manuscript's content, interpretation, source use, conceptual framing, and conclusions.

Created on 01 Nov. 2026

Was this synthesis useful?

Score: 0

Ask questions about this paper to our AI assistant

How to chat with this paper? Illustrated guide

You can also chat with multiple papers at once here.

AI assistant instructions?

Tools for this paper

Explain it differently 1 free rewrite left today

Save to a collection Group papers in collections to compare them, chat with them and export them. Continue with Google Create a free account Log in

Export BibTeX · RIS (Zotero, Mendeley, EndNote) · Obsidian (Markdown) · Notion (Markdown)

Advanced tools Premium

For researchers who follow a field closely: citation alerts, journal-club slides, version comparison and Zotero export. Discover Premium

Zotero

Citation alerts Citation alerts guide

Who works on it? The people and institutions who publish on its topic

Journal club Premium+ Slides to present it Propose for the journal club

Talk about it in public Premium+ Press kit: release, questions, hype check Press kit guide

Papers around this one

See them on a map

Ranked by how close their title and abstract are to this paper’s. Select some to read them together.

Closest papers

Very close Builds on this paper
From Entity Resolution Anomalies to Reconstruction Failure Modes: A Taxonomic Note Following Minimum Viable Digital Identity Optimization (DIO)
Daniel Beránek · 2026 · Zenodo (CERN European Organization for Nuclear Research) preprint · AI summary available

The taxonomic note refines the paper’s entity-resolution anomalies into four reconstruction-failure domains and additional failure types.

In short
The note organizes reconstruction failure modes into four diagnostic domains and distinguishes realized failures from susceptibility to destabilization.
Very close Same method, another problem
Digital Identity Optimization (DIO): A Conceptual Framework for Entity-Based Visibility in the Age of AI-Mediated Search
Daniel Beránek · 2026 · Zenodo (CERN European Organization for Nuclear Research) preprint · AI summary available

Digital Identity Optimization applies the DIO framework to entity-based visibility in AI-mediated search, while the paper proposes it as an integrity hypothesis for identity reconstruction.

In short
The paper proposes Digital Identity Optimization (DIO), a conceptual framework for making entities consistently identifiable and recommendable across human and AI-mediated systems.
Close Builds on this paper
Constraining the Commit Boundary: A DIO/ODI-Informed Authorization Guardrail for Agentic Entity Resolution
Daniel Beránek · 2026 · Artificial Intelligence · AI summary available

The guardrail uses DIO/ODI and reconstruction-failure distinctions to control identity-changing commits in agentic entity resolution.

In short
The paper proposes a DIO/ODI-informed authorization boundary that permits persistent entity-resolution mutations only with claim-specific evidence, valid scope, and current state and policy versions.
Close Same method, another problem
Manifest of Digital Identity Optimization (DIO) and Ontology of Digital Identity (ODI): Canonical Multilingual Edition
Daniel Beránek · 2026 · Zenodo (CERN European Organization for Nuclear Research) · AI summary available

The canonical manifesto presents DIO and ODI as frameworks for representing and maintaining digital identity, which the paper discusses as a theoretical integrity hypothesis.

In short
The abstract presents DIO and ODI as a multilingual framework for conceptualizing and maintaining digital identity, but reports no empirical evaluation.
Close Broader context
The AI Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search
Yutaro Yamada, Robert Tjarko Lange, Cong Lu et al. · 2025 · Artificial Intelligence · preprint, not peer reviewed · cited 499 times · AI summary available

The AI Scientist-v2 uses agentic tree search for automated scientific discovery, whereas the paper examines how entity-resolution failures can propagate through multi-agent AI systems.

Abstract
AI is increasingly playing a pivotal role in transforming how scientific discoveries are made. We introduce The AI Scientist-v2, an end-to-end agentic system capable of producing the first entirely AI generated peer-review-accepted workshop paper. This system iteratively formulates scientific hypotheses, designs and executes experiments, analyzes and visualizes data, and autonomously authors scientific manuscripts. Compared to its predecessor (v1, Lu et al., 2024 arXiv:2408.06292), The AI Scientist-v2 eliminates the reliance on human-authored code templates, generalizes effectively across diverse machine learning domains, and leverages a novel progressive agentic tree-search methodology mana…
Close Broader context
Deconstructing Human-AI Collaboration: Agency, Interaction, and Adaptation
Steffen Holter, Mennatallah El-Assady · 2024 · Human-Computer Interaction · cited 70 times · AI summary available

The paper proposes dimensions for describing human-AI collaboration, while the paper examines identity integrity in AI systems including multi-agent settings.

Abstract
As full AI-based automation remains out of reach in most real-world applications, the focus has instead shifted to leveraging the strengths of both human and AI agents, creating effective collaborative systems. The rapid advances in this area have yielded increasingly more complex systems and frameworks, while the nuance of their characterization has gotten more vague. Similarly, the existing conceptual models no longer capture the elaborate processes of these systems nor describe the entire scope of their collaboration paradigms. In this paper, we propose a new unified set of dimensions through which to analyze and describe human-AI systems. Our conceptual model is centered around three hig…
Close Review
Agentic AI for Scientific Discovery: A Survey of Progress, Challenges, and Future Directions
Mourad Gridach, Jay Nanavati, Khaldoun Zine El Abidine et al. · 2025 · Computation and Language · preprint, not peer reviewed · cited 127 times · AI summary available

The survey reviews agentic AI systems and challenges in scientific discovery, while the paper focuses on entity-resolution failures in RAG and multi-agent systems.

Abstract
The integration of Agentic AI into scientific discovery marks a new frontier in research automation. These AI systems, capable of reasoning, planning, and autonomous decision-making, are transforming how scientists perform literature review, generate hypotheses, conduct experiments, and analyze results. This survey provides a comprehensive overview of Agentic AI for scientific discovery, categorizing existing systems and tools, and highlighting recent progress across fields such as chemistry, biology, and materials science. We discuss key evaluation metrics, implementation frameworks, and commonly used datasets to offer a detailed understanding of the current state of the field. Finally, we …
Related Broader context
Decentralized Governance of Autonomous AI Agents
Tomer Jordi Chaffer, Charles von Goins II, Bayo Okusanya et al. · 2024 · Artificial Intelligence · cited 24 times · AI summary available

ETHOS proposes decentralized governance for autonomous agents, whereas the paper considers identity integrity and failure propagation in agentic systems.

Abstract
Autonomous AI agents present transformative opportunities and significant governance challenges. Existing frameworks, such as the EU AI Act and the NIST AI Risk Management Framework, fall short of addressing the complexities of these agents, which are capable of independent decision-making, learning, and adaptation. To bridge these gaps, we propose the ETHOS (Ethical Technology and Holistic Oversight System) framework, a decentralized governance (DeGov) model leveraging Web3 technologies, including blockchain, smart contracts, and decentralized autonomous organizations (DAOs). ETHOS establishes a global registry for AI agents, enabling dynamic risk classification, proportional oversight, and…

Go further

How are these papers found?

We compare the title and abstract of this paper with those of every paper summarized on this site, and with the papers Semantic Scholar recommends, those it cites and those citing it (for a much-cited paper, its most recent citations and, from OpenAlex, the most cited ones). Among the prior work and the follow-ups, the papers that use each other in depth, a method it uses and the much-cited follow-ups come a little higher. The levels (very close, close, related) come from the similarity of the texts, not from a judgment of quality; * marks a paper without an abstract, compared by its title. The sentence under a paper is written by an AI from the two abstracts: check it in the papers. A paper marked not relevant by three readers leaves the list.

Updated on 1 November 2026.

Disclaimer: The AI-based summarization tool and virtual assistant provided on this website may not always provide accurate and complete summaries or responses. We encourage you to carefully review and evaluate the generated content to ensure its quality and relevance to your needs.

The Crisis of Entity Resolution in Heterogeneous AI Environments: Data, Financial, and Semantic Dimensions

The essentials

Question

Method

Findings

Limitations

Why it matters

Read further

Definitions

Why AI Needs to Know Who—or What—it Is Talking About

An “Entity Turn” in AI

Four Ways Identity Can Break Down

Why This Is More Than Data Cleaning

DIO and ODI: A Proposal to Test

A Useful Agenda, with Evidence Still to Come

Ask questions about this paper to our AI assistant

Tools for this paper

Advanced tools Premium

Papers around this one

Closest papers

Go further

Cite this paper (BibTeX)