The Convergence of AI Governance and Information & Records Management

Reframing Ethical AI Through Records and Information Management

The Convergence of AI Governance and Information & Records Management

Introduction: Reframing Ethical AI Through Records and Information Management

As artificial intelligence (AI) technologies proliferate across public and private sectors, ethical governance has emerged as a central concern. Organisations and regulators are scrambling to develop responsible AI principles, accountability frameworks, and compliance protocols. But many of these challenges are not new. In fact, records and information management (RIM) professionals have long confronted the same ethical questions now being rebranded through an AI lens: how to ensure data quality, maintain transparency, protect privacy, and uphold public trust.

This article argues that ethical AI governance can and should draw from the established disciplines of RIM. Far from being left behind, records professionals hold a key to managing the risks and realising the benefits of AI. Many AI principles reflect longstanding archival values, and tools such as ontology and metadata management offer practical pathways to accountability, explainability, and compliance.

The Role of RIM for Ethical AI Governance

Traditionally records management practice has been designed for creating, maintaining, and ensuring the accessibility of information within a paper paradigm. Organisations now create and hold huge stores of digital information, from databases with millions of emails, to legacy systems and shared drives full of unstructured content, and we now need to appraise this information, classify it, and apply disposal authorities. However, it is no longer possible for people to sort through all this information manually. Therefore, we require automation to undertake records and information practice and Artificial Intelligence (AI) to underpin this approach at scale, underpinned by ontology-based information modelling.

As the records management profession is beginning to be disrupted by AI, impacting core functions from healthcare and education, through to core operations such as policy development to service delivery. The need for robust ethical frameworks and governance mechanisms becomes more critical. We have seen the rise of (often self-proclaimed) new role descriptions, such as “AI Ethicists”, and “AI Governance Experts”, and the concomitant new emerging industries. However, at its core, ethical AI and governance are just extensions of traditional information ethics and records management governance. The rise of these ‘new’ roles, I believe, is just a reflection of how poorly we have been undertaking this core infrastructural good practice for information and data management over the preceding couple of decades.

The article will explore how principles of data governance, societal expectations, legislative compliance, and other critical factors which underpin the ethical management of AI systems are just the existing under-employed current capabilities that records, and information management provides. It also looks at the critical role that ontology-based approaches can take in enabling good records and information management practice in AI business systems.

The emerging topic of 'ethical AI governance' and traditional records and information management practices share a core underpinning, reflecting their common foundations in the need for governance, ethical considerations, and regulatory compliance. The ISO 15489 Standard for Records Management describes the attributes of information as reliability, integrity, authenticity and usability. The following consideration set out how existing information management practice is really seeking the same outcomes as AI Ethics and Governance.

Information/Data Quality and Integrity

High-quality information is crucial for the success of AI business systems. Poor information and data can lead to inaccurate models, biased outcomes, and eroded trust. Managing information quality involves rigorous validation, cleansing (Retention & Disposal), and enrichment processes (Metadata Management), which are fundamental aspects of AI governance. Additionally, addressing information quality technical debt, i.e. the cost of deferred maintenance or suboptimal solutions, is vital for maintaining the integrity and performance of AI business systems. This requires ongoing investment in data infrastructure, records and information management policies and governance practices to prevent the accumulation of this information and data quality technical debt.

• Ethical AI Governance: Emphasises the need for high-quality, accurate, and unbiased data to ensure that AI systems produce reliable and fair outcomes.

• Records & Information Management: Focuses on maintaining the integrity and accuracy of records throughout their lifecycle to ensure reliable and trustworthy information.

Regulatory Compliance

Ethical AI frameworks, such as those from the EU, OECD, ISO 42000 series, and NIST/IEEE Frameworks, promote principles like fairness, accountability, transparency, and explainability. These legislative frameworks are evolving to address the challenges posed by AI. Regulations like the European Union's AI Act set stringent requirements for data protection, transparency, and accountability, and are built on existing principles of data governance, emphasising the need for lawful, fair, and transparent data processing. Compliance with these regulations necessitates a deep understanding of data ethics and governance, as organisations must navigate complex legal landscapes to ensure their AI systems adhere to regulatory standards.

So, regulatory conformance is not a new consideration, for example the first Information management legislative challenge was the introduction of the first Freedom of Information Act in Sweden in 1766.

• Ethical AI Governance: Requires adherence to laws and regulations concerning data protection, privacy, and the ethical use of AI, such as the EU GDPR and AI Acts.

• Records & Information Management: Involves compliance with various legal and regulatory requirements related to data retention, privacy, and access, ensuring that records are managed according to applicable laws.

Privacy and Data Protection

Privacy and data sovereignty are paramount in the age of AI. Ethical AI governance must prioritise the protection of individual privacy rights and comply with data sovereignty laws that dictate where data can be stored and processed. These concerns are deeply rooted in traditional information and data ethics, which emphasise the importance of safeguarding personal data and respecting jurisdictional boundaries. This is even more important in the New Zealand context in seeking to meet our obligations under Te Tirit o Waitangi, Māori Data Sovereignty, and managing information as taonga, as the New Zealand records and information management profession has wrestled with over the last fifty years or so,

• Ethical AI Governance: Prioritises the protection of individual privacy and personal data, and the rights and entitlements of minorities, to ensure that AI systems do not infringe on privacy or human rights.

• Records & Information Management: Includes safeguarding personal and sensitive information within records, preventing unauthorised access, and ensuring confidentiality for all stakeholders.

Public Trust, Transparency and Accountability

The societal implications of AI are manifold and multifaceted, affecting everything from social services to privacy rights. Public trust in AI systems hinges on their transparency, fairness, and accountability. Ethical AI governance ensures that these systems operate within societal norms and values, avoiding biases that could lead to discrimination or injustice. This mirrors the broader goals of information and data ethics, which aim to safeguard individual rights and promote equity in data handling practices. Information Management practice however seeks to embed these principles from the very fabric of how information is collected and managed, not as a post-creation guardrail.

• Ethical AI Governance: Demands transparency in AI decision-making processes and accountability for the outcomes produced by AI systems.

• Records & Information Management Requires clear documentation and traceability of records, ensuring that actions can be audited, and accountability to be maintained.

Stewardship and Governance

Information stewardship involves the responsible management and oversight of information assets. In the context of AI, this means ensuring that data is used ethically and in alignment with organisational values, and public expectations. Information Management maturity is critical for achieving this. Mature public sector agencies are often better equipped to implement comprehensive Information stewardship and governance strategies, incorporating ethical considerations into their AI development and deployment processes. This maturity is often reflected in well-defined Information management frameworks, for example as set out in the Public Records Act legislation, and privacy practices, and a culture of records and information management capability.

• Ethical AI Governance: Involves responsible management and oversight of AI data assets, ensuring ethical use and alignment with organisational values.

• Records & Information Management: Entails stewardship of records and information assets, ensuring they are managed responsibly and ethically throughout their lifecycle.

Lifecycle and Process Management

Effective information management, over time, is the backbone of ethical AI. It encompasses the policies, procedures, and standards that govern the collection, storage, use, and dissemination of data. High-quality data governance ensures that AI systems are trained on accurate, unbiased, and representative data sets, reducing the risk of harmful outcomes. Metadata management, training data registers, and rigorous data quality assessments are essential components of this process. These practices are not new; they are rooted in established data information management frameworks that prioritise data integrity and reliability. The Lifecycle Model was conceived by Phillip Coolidge Brooks and Emmett J. Leahy in the United States National Archives in the 1940s, so it’s almost a century old approach.

• Ethical AI Governance: Focuses on managing the lifecycle of data used in AI, from collection and storage to processing and disposal, ensuring ethical considerations are maintained at each stage.

• Records and Information Management: Involves managing the entire lifecycle of records, including creation, maintenance, use, and disposal, with a focus on ethical and compliant practices. The role of ontology is critical in taking the step towards managing records within continuum model paradigms.

Metadata Management, Ontology, and Taxonomy

Effective metadata management, ontology, and taxonomy are essential for organising and managing data in a way that supports ethical AI governance. Metadata provides context and provenance information, aiding in traceability and accountability. Ontologies and taxonomies help in structuring data, enabling more accurate and meaningful AI insights. These practices are foundational elements of Records and Information Management practice, ensuring that data is well-organised, understood, and utilised ethically. These considerations are also at the heart of how LLMs (Large language Models) are used and deployed, which are the basis for Generative AI.

• Ethical AI Governance: Utilises metadata to provide context, data lineage (provenance), and traceability for data used in AI systems, enhancing transparency and accountability.

• Records and Information Management: Relies on metadata to organise, classify, and manage records, supporting efficient retrieval and ensuring records are properly contextualised, and will increasingly rely on ontological modelling to embed, and automate, these concepts within AI Business Systems.

Data Retention and Disposal of AI Processes, and Outputs

Good data and information management practice, including retention and disposal schedules, is critical for ethical AI governance. These practices ensure that data is retained only as long as necessary and disposed of securely when no longer needed. This reduces the risk of data breaches and ensures compliance with legal and regulatory requirements. Records management is a well-established discipline emphasising the importance of managing data lifecycle processes ethically and responsibly. It is also critical to maintain records of automated decision-making, training data, algorithms, data processing and outputs for as long as they are required to be accessible for accountability and reproducibility purposes.

• Ethical AI Governance: Requires policies for the retention and ethical disposal of data used in AI systems to prevent misuse and ensure compliance with regulations. A need that is not currently being fully met.

• Records and Information Management: Involves implementing retention schedules and disposal policies to manage records' lifecycle, ensuring they are kept only as long as necessary and disposed of securely. Automation of these disposal policies at scale requires AI toolsets with underlying ontologically driven modelling for policy automation and auto-classification of these schedules.

  Automated Decision-Making and Risk Management

Automated decision-making systems powered by AI raise significant ethical concerns. These systems must be transparent, explainable, and accountable to ensure they do not perpetuate biases or make unfair decisions. Ethical AI governance involves implementing robust oversight mechanisms, including audit trails, decision logs, and impact assessments, good recordkeeping, and provenance (aka data lineage). These practices are extensions of established information management principles that prioritise accountability and transparency in data processing activities.

• Ethical AI Governance: Involves identifying and mitigating risks associated with AI systems, including biases, data breaches, and unethical outcomes.

• Records and Information Management: Includes assessing and managing risks related to records, such as loss, unauthorised access, and non-compliance with regulations.

Training and Awareness

Training promotes a culture of responsibility among AI practitioners. It ensures that those involved in AI development and deployment are conscious of the ethical dimensions of their work and are committed to upholding ethical principles. When developers and designers are trained in AI ethics, they are more likely to incorporate ethical considerations into the design and implementation phases of AI systems, leading to more ethically sound outcomes. Maintaining records of training and awareness of staff is key to providing evidence or good intent, and awareness of ethical practice in information design and handling is usually already addressed in organisation’s records and information management training.

• Ethical AI Governance: Requires training stakeholders on the ethical use of AI, ensuring they understand the implications and responsibilities associated with AI systems.

• Records and Information Management: Involves educating employees on proper records management practices, including compliance, privacy, and ethical considerations.

Why ontology is a critical enabler for Records & Information Management Practice

Using an ontology-based approach to records and information management when implementing good practices for ethical artificial intelligence is critical because it establishes a structured, semantic framework that ensures consistency, comprehensiveness, and interpretability across complex systems. Here are the key reasons:

Semantic Interoperability

AI systems often operate with diverse datasets and information sources. An ontology provides a shared vocabulary and standardised structure for data representation, enabling seamless integration across different systems and domains. It ensures that records from disparate sources are interpreted in a consistent and meaningful way by AI, avoiding misinterpretation or information/data silos.

Enhanced Contextual Understanding

Ontologies define the relationships between entities, attributes, and processes within records, offering rich contextual information for AI systems to understand and analyse information accurately, and apply classifications, policy and additional contextual metadata. This improves decision-making capabilities in AI, as the system can grasp the nuanced meanings and relationships inherent in the records and retain records of these contextual relationships, and trustworthy evidence of these decision-making processes.

Improved Data Governance and Regulatory/Legislative Compliance

Records management usually requires compliance with specific legal, regulatory, or organisational policies. Ontologies can encode rules and classifications that align with these requirements, making it easier for AI to automatically enforce compliance routines and policies, e.g. access permissions, privacy protocols, and legal holds/retention and disposal actions. This is particularly valuable in industries like healthcare, finance, and legal, where regulatory adherence is critical.

Efficient Records Classification and Retrieval

Ontology-based approaches allow AI to classify and retrieve records more effectively by leveraging both the hierarchical and relational structure of information and provide the basis for automation of these processes. This ensures that relevant records are surfaced based on their meaning and context rather than just keywords, enhancing search precision and efficiency for users, as well as access controls, de-duplication and version control.

Lifecycle Management and Provenance Tracking

Managing the lifecycle of records (e.g., creation, use, retention, and disposal) is a cornerstone of records management. Ontologies can provide explicit definitions and workflows for these processes, enabling AI to automate and monitor them. Ontology-based frameworks also facilitate the tracking of provenance, ensuring transparency and accountability in how records are used and modified by AI systems. This is critical for implementing records and information management practice in AI Business Systems. As we move towards a new era of Continuum thinking and move beyond d the linear and unidirectional lifecycle model the ability to capture provenance in complex muti-entity relationships becomes more challenging and even more crucial to track and capture provenance metadata effectively and retain it overtime as part of the record context.

Facilitating Explainable AI (XAI)

Explainability is critical for AI systems, particularly in high-stakes applications. An ontology-based records management approach makes it easier to trace and explain AI decisions by linking them back to well-structured, semantically rich data sources. This improves trust in AI systems and enables stakeholders to audit decisions effectively, and retain trustworthy, accurate authentic and usable records of these business processes.

Essentially, we cannot have accountability. Governance, audit, or ethical/explainable AI without the ability to consult the records of the inputs, outputs, algorithms, training data, training methods, the determinations and subsequent actions. Records of the process of quality of the underlying data ns the logic and reasoning engine performance and are critical to understanding the final results of an AI query, or process, and detecting any potential bias, or inaccuracy.

Scalability and Adaptability

AI systems often need to adapt to new domains, data types, or regulatory environments. Ontologies can be extended or modified to incorporate new knowledge without disrupting existing structures. This makes the records management framework scalable and future-proof, accommodating growth in AI capabilities and data complexity, and allows extensibility across geographical spaces, languages and cultural frameworks at a global scale.

Addressing Ethical Concerns

Ethical AI usage requires transparency, accountability, and the minimisation of bias. Ontologies can be used to define ethical guidelines and constraints directly within the records management framework and information architecture, guiding AI systems to operate within these boundaries, and retaining records for accountability and evidential purposes.

Enterprise Business Information Standardisation

By adopting an ontology-based approach, organisations can create a standardised foundation for records management, reducing inconsistencies and ensuring uniform application of AI processes across departments, and regions. Ontology-based approach to records and information management also provide a basis for interoperability across organisations and can provide the consistent metadata schema and data classification for both cross-organisational information systems and data-lakes.

This also allows for an approach of having a centralised data dictionary, ‘business taxonomy’, or term-store. Therefore, if a new item is added, a term is removed or altered then this can be applied globally across the data ecosystem not just locally within a particular silo, single instance, or subset of the over data enterprise information and data ecosystem.

The Role of Knowledge Graphs in Ontology-Based Records Management

Knowledge graphs play a critical role in implementing ontology-based records management for AI. They act as a practical, dynamic representation of the ontology, enabling AI systems to interact with, infer from, and visualise the structured relationships between records and associated data.

Dynamic Data Representation

Knowledge graphs serve as a flexible and dynamic way to represent the relationships and hierarchies defined in the ontology. They organise information into nodes (entities) and edges (relationships), making it easy for AI to traverse and query complex data structures. This provides a consistent basis for records management automation implementation, and maturity improvement.

This dynamic nature allows for continuous updates and integration of new data, supporting evolving organisational needs, and ongoing records and information management maturity and capability.

Facilitating Complex Querying and Insights

Knowledge graphs enable AI systems to perform complex, context-aware queries, such as identifying relationships between records or extracting trends across datasets. This is particularly useful for applying records and information policy attribution and value classification.

An ontology-based approach to records management enables AI business systems to return results with a clear, structured, and semantically rich understanding of records, enhancing interoperability, compliance, contextual accuracy, and scalability. It ensures that AI systems can effectively manage records while supporting ethical and explainable practices when queries are inputted. This structured methodology is indispensable for achieving robust, reliable, and future-ready AI implementations in records management, particularly when using a query-based approach to apply records management policies, such as retention and disposal for example.

Ontology as the ultimate underpinning design model for AI

An ontology for data and information modelling leverages concepts like epistemology, tautology, and semantics to enhance the structure, meaning, and relationships of data. Here's how these philosophical and linguistic concepts play a role:

Epistemology (The Study of Knowledge):

In the context of data and information modelling, epistemology relates to how knowledge is structured, acquired, and validated within the system at a semantic level. Ontologies use epistemological principles to ensure that data is not just stored, but that its meaning and sources are well-understood. This involves:

• Knowledge Representation: Defining the relationships between entities, and concepts to represent knowledge in a way that both humans and machines can understand. This helps systems differentiate between "what we know" and "how we know it."

• Assertions and Inference: Ontologies allow for logical reasoning, where new knowledge can be inferred from existing data. Epistemology helps ensure that the data structures support the derivation of valid conclusions and decision-making.

• Trustworthiness: Epistemology ensures that the provenance of data (where it comes from and how it was derived) is clear, making it possible to evaluate the credibility of the information.

Tautology (Logical Truth or Redundancy):

Tautology in ontology ensures that relationships and definitions are logically consistent and non-redundant. This is important for:

• Consistency in Definitions: Ontologies must avoid tautological definitions (where terms are defined using themselves), ensuring clarity and avoiding circular reasoning in data relationships.

• Logical Integrity: While tautology in natural language is often seen as redundant, in formal logic, it can ensure that certain statements are always true (e.g., a "dog is a dog"). In ontology design, tautological rules help maintain the logical integrity of the system, ensuring that any entity is correctly classified without contradiction.

• Minimising Redundancy: A well-constructed ontology minimises tautological or redundant information to avoid inefficiencies and confusion, ensuring that each concept is defined uniquely and with purpose.

Semantics (Meaning of Concepts and Relationships):

Semantics is central to ontologies because the primary goal is to model the meaning of data and information. It focuses on how concepts, terms, and relationships are defined and understood. Ontologies employ semantic principles in several ways:

• Defining Terms and Concepts: Semantics ensures that each term or concept within the ontology has a clear, agreed-upon meaning. This is critical for enabling shared understanding between different systems or people.

• Relating Entities: Semantics in ontology modelling defines how entities are related (e.g., part-whole, type-instance, cause-effect relationships). This allows systems to interpret the meaning behind data, rather than treating it as isolated facts.

• Disambiguation: Semantics helps to differentiate between homonyms (words with multiple meanings) and synonyms (different words with the same meaning), ensuring that the data model captures the correct meaning in context.

• Formal Semantics: Ontologies use formal logic to define the relationships between entities, allowing systems to reason over data and derive new meanings or facts based on the modelled relationships.

Integration of Epistemology, Tautology, and Semantics in Ontology Design:

• Epistemology shapes how knowledge is represented, validated, and organised within the ontology, ensuring that it reflects the way knowledge is structured in the real world.

• Tautology ensures logical consistency within the ontology, avoiding circular definitions and maintaining a robust framework where entities and relationships are logically sound and non-redundant.

• Semantics gives meaning to the concepts and relationships in the ontology, allowing for effective communication between systems and enabling machines to interpret the data in a human-like way.

The Value of Ontology for AI: Enhancing Information, Data Management, and Tailoring AI to Cultural Contexts

In the realm of Artificial Intelligence, the importance of data structure and representation cannot be overstated. Ontologies, which provide a structured framework for organising information, offer significant benefits in both data management and AI applications. Ontologies describe the relationships and concepts within a domain, enabling machines to understand, process, and reason with information in a meaningful way. While ontology has long been recognised for its value in information and data management, its utility in AI particularly in enhancing generative AI (GenAI) across cultural contexts and fine-tuning large language models (LLMs). We see this challenge with different Generative Pre-trained Transformers (GPTs) across cultures e.g. Mistral from the EU, DeepSeek from China, or ChatGPT from the USA. The underlying cultural perspectives, socio-political perspectives and societal mores increasingly presents an emerging frontier from understanding information context across cultural realms (see below).

Traditional Benefits: Ontology in Data and Information Management

Before delving into its role in AI, it’s essential to understand ontology’s foundational value in data and information management. Ontologies improve the consistency, quality, and discoverability of data by creating a shared semantic layer across systems and teams. This results in:

• Improved Data Quality: Ontologies establish clear definitions for concepts, ensuring that data across systems is consistently structured and semantically aligned. This directly contributes to higher data quality, as it eliminates ambiguity and misinterpretation.

• Enhanced Interoperability: By using a common conceptual framework, ontology supports seamless data integration across diverse systems. This is particularly valuable in large organisations where information often resides in silos.

• Efficient Data Governance: Ontologies assist in data governance by clearly defining the roles, rules, and relationships of data elements, making it easier to manage access, compliance, and security.

• Better Knowledge Discovery: With an ontology in place, organisations can leverage advanced search and reasoning capabilities. Users can query data in more sophisticated ways, discovering relationships and insights that may not be apparent through traditional keyword-based search.

These traditional benefits provide a strong foundation for information and data management. But the value of ontology in AI applications goes beyond this, particularly when it comes to tailoring AI models for cultural sensitivity and organisational context.

Ontology’s Role in GenAI Across Cultures

One of the most significant challenges for generative AI models is operating effectively across diverse cultural contexts. GenAI models, such as those based on large language models, are typically trained on vast datasets that reflect global trends, often biased toward dominant languages and cultural norms. Ontologies can help address this by creating culturally specific frameworks that define key concepts, practices, and language use. They may allow conceptual understanding to be mapped across culturally specific concepts; this is however an emerging capability with the AI industry.

For example, an ontology tailored to a particular cultural context can teach an AI system the nuances of communication, norms, and values that are unique to that culture. By embedding these into the AI model’s reasoning structure, GenAI systems can generate responses, content, or insights that are culturally aware and contextually appropriate. This is particularly valuable in applications like customer support, content creation, and global-scale decision-making.

Ontologies also help to avoid misunderstandings that arise from cultural differences. For instance, while an AI might generate text that seems accurate from a global perspective, it could inadvertently offend or alienate users in specific cultural settings due to differing social or historical contexts. With ontology-driven guidance, AI systems can navigate these complexities more effectively, ensuring inclusivity and respect for diversity.

Customising Generically Pre-Trained Transformers and LLMs

Large language models like ChatGPT-4o are trained on enormous, generalised datasets that allow them to perform well in a wide range of tasks. However, for organisations to derive the most value from these models, they need to be fine-tuned to specific contexts—both organisationally and culturally.

This is where ontology comes in. By developing an organisational ontology that reflects the specific terminology, processes, and relationships within a business, organisations can customise generically pre-trained models to better align with their own knowledge structures and operational needs.

Consider a science-based organisation with a highly specialised domain of knowledge. The general-purpose language model will lack the precise understanding of the technical language and interdependencies within that domain. However, by incorporating the organisation’s ontology, the AI can better grasp these nuances, leading to more relevant outputs and more accurate decision support.

In addition, ontology-driven fine-tuning can help organisations embed their internal policies, values, and mission-critical objectives into AI models. This ensures that any AI-driven automation or decision-making processes are aligned with the organisation’s strategic priorities, minimising risks and enhancing efficiency.

4. Practical Implications for AI Development and Use

By leveraging ontology, organisations can achieve a more tailored, intelligent use of AI. Some practical implications include:

• Enhanced Personalisation: Ontology enables LLMs to generate more relevant, contextually aligned outputs, whether the goal is customer engagement, internal communication, or strategic analysis.

• Improved Collaboration Across Teams: Ontology can serve as a shared knowledge framework, helping cross-functional teams communicate more effectively and work together on AI initiatives with a common understanding.

• Ethical AI: Ontology helps organisations develop AI systems that are more culturally sensitive, ethical, and aligned with their values. This is particularly important in avoiding biases and ensuring fair, inclusive AI applications.

• Faster Time-to-Value: With an ontology in place, organisations can accelerate the fine-tuning of AI models. The time and effort needed to adapt a generic LLM to an organisation’s needs are significantly reduced, allowing businesses to extract value from AI more quickly.

Ontology’s Strategic Value for AI

The value of ontology for AI goes beyond traditional data and information management benefits. Ontologies allow organisations to customise AI models like GPTs and LLMs, making them more relevant and effective in specific cultural and organisational contexts. By leveraging ontology, businesses can unlock the full potential of AI, ensuring that these systems are not only intelligent but also culturally aware, contextually accurate, and strategically aligned. This capability will be increasingly important as AI becomes more pervasive in global operations, decision-making, and innovation.

Why a Fully Relational Ontology Approach is Essential for Information and Data Management

In today’s rapidly evolving digital landscape, the ways in which organisations manage information and data can determine their operational success. While traditional hierarchical, taxonomical models have long served as a basis for data management, they are increasingly proving insufficient to handle the complexity and dynamism of modern data environments. Instead, a fully relational ontology approach offers the flexibility, interconnectedness, and adaptability needed for more effective information and data management.

This shift from a rigid, top-down taxonomy to a dynamic, relational ontology is driven by several key factors. The need for richer contextual understanding, deeper interrelationships between data elements, and the ability to evolve with emerging technologies like AI makes relational ontology far more valuable.

Limitations of a Taxonomical, Hierarchical Approach

A taxonomical approach to data organisation arranges information into fixed, predefined categories in a top-down hierarchy. While this may work for simple classification systems, it falls short in more complex domains where relationships between entities are multifaceted and constantly changing.

Some major limitations include:

• Rigidity: Taxonomies are typically static. Once data is categorised, it’s difficult to adapt to new information, relationships, or changes in the way data is used or understood.

• Loss of Context: Hierarchical taxonomies often force data into narrow categories, which can result in a loss of the rich, contextual relationships that exist between entities.

• Over-simplification: Real-world data is rarely as clean and straightforward as a taxonomy suggests. Complex entities often belong to multiple categories or have relationships that span beyond simple parent-child structures.

For example, in a taxonomy, an organisation might classify all its employees by role. This simple structure, however, would fail to capture other important relationships, such as collaborations between teams, expertise areas across departments, or cross-functional project involvement. This lack of nuance creates a disconnected understanding of the organisational landscape.

The Power of a Fully Relational Ontology Approach

A fully relational ontology, by contrast, models data in a way that reflects the actual complexity and interconnectedness of real-world entities. Rather than adhering to a rigid hierarchy, it allows for a flexible, web-like structure where data points are related through multiple, often dynamic, associations.

The key advantages of this approach include:

• Flexibility and Scalability: A relational ontology can evolve with the organisation. As new data sources or relationships emerge, the ontology can easily adapt without the need to reclassify everything under a rigid hierarchy.

• Contextual Richness: By capturing relationships between entities in multiple dimensions, a relational ontology offers a much richer representation of information. This allows for more nuanced data discovery, analysis, and decision-making.

• Real-World Alignment: In a fully relational model, entities are not forced into one-size-fits-all categories. Instead, each entity is understood in the context of its various relationships—whether those are temporal, geographical, functional, or thematic. This mirrors the complexity of real-world systems and interactions.

Why Relational Ontology is Essential for Modern Information and Data Management

For organisations managing large volumes of data, a fully relational ontology approach is not just preferable—it’s essential. Here’s why:

• Handling Complexity: Modern enterprises deal with intricate data ecosystems that include structured and unstructured data, multimedia assets, customer interactions, and IoT-generated information. A relational ontology can manage these complex interconnections, enabling the discovery of insights that a taxonomical model might miss.

• Enabling Semantic Search and AI: Relational ontologies are at the heart of semantic search engines and AI applications. Because they capture relationships between data points, they enable more intelligent querying, making it easier to locate relevant information and generate insights. Ontology-driven AI systems can use this relational web to understand context, reason over data, and deliver more accurate predictions or recommendations.

• Enhanced Interoperability: A fully relational ontology provides a shared conceptual framework that allows data from different systems to interact seamlessly. This is especially important in large organisations or multi-party collaborations where data silos can hamper efficiency. Relational ontology fosters interoperability by enabling data from various domains to connect in meaningful ways.

• Dynamic Data Governance: A relational approach makes it easier to manage data governance in a world of ever-evolving privacy regulations, security requirements, and business needs. Instead of manually updating hierarchical categories and policies, organisations can automate governance processes based on the relationships between data elements. For example, if a new regulation affects data tied to a specific process, a relational ontology can instantly update governance protocols across all related data.

Why Not a Taxonomy? The Case Against Hierarchical Models in Modern Systems

Although taxonomies have served well in simpler, more static environments, they are no longer sufficient in today’s dynamic and interconnected data ecosystems.

• Poor Adaptation to Change: In a taxonomical structure, adding new data or understanding new relationships often requires reworking entire hierarchies. For example, when new technologies or business processes emerge, hierarchies struggle to accommodate these shifts without undergoing significant restructuring.

• Inflexible for AI and Machine Learning: AI systems thrive on understanding patterns, relationships, and context. Hierarchical taxonomies limit these capabilities by constraining data into rigid categories, making it harder for AI to detect the subtleties and complex interconnections inherent in relational data.

• Fragmentation and Redundancy: Hierarchical systems often lead to data fragmentation, where similar data is classified in different parts of the taxonomy without clear connections. This can lead to redundancy and inefficiencies, as information that could be linked together remains siloed.

The Strategic Value of Relational Ontologies for Future-Proofing Organisations

Adopting a fully relational ontology approach not only enhances current data management practices but also future-proofs organisations for emerging technologies and challenges. Here’s how:

• Tailoring AI Systems to Organisational Contexts: Relational ontologies enable AI systems to be fine-tuned to the unique processes, workflows, and culture of an organisation. By embedding specific relationships and knowledge structures, organisations can create AI models that are contextually aware and highly relevant.

• Supporting Generative AI (GenAI): A relational ontology is crucial for enabling GenAI systems to generate contextually and culturally appropriate outputs. By embedding complex relationships and cultural nuances into the model, organisations can ensure that GenAI generates more accurate, meaningful content that aligns with their specific needs.

• Improved Decision-Making: With a relational ontology, decision-makers can query data from multiple perspectives, uncover hidden relationships, and make informed choices that reflect the full complexity of the situation.

The Case for a Fully Relational Ontology

In the modern era of information and data management, a fully relational ontology approach offers unparalleled benefits over traditional taxonomical models. It provides the flexibility, contextual depth, and adaptability required to navigate today’s data complexity. By moving away from hierarchical structures, organisations can unlock richer insights, enhance interoperability, support AI and machine learning initiatives, and future-proof their data strategies in an ever-changing landscape.

A relational ontology is not just a better way to organise data, it’s a strategic necessity for organisations seeking to maximise the value of their information assets in the age of AI and beyond.

Conclusions

In essence, both ethical AI governance and traditional records and information management practices are concerned with the responsible and ethical handling of data. They share common goals of ensuring data quality, protecting privacy, maintaining compliance, and fostering transparency and accountability. As AI continues to evolve, integrating these established records and information management principles into AI governance frameworks will be crucial for promoting ethical and responsible AI use. Ethical AI and governance requirements are not novel concepts but rather an extension of traditional records and information management, ethics and governance. By drawing on these established principles of Information Management, privacy, and accountability, organisations can develop and deploy AI systems that align with societal values, comply with legal requirements, and promote equitable outcomes. As AI continues to evolve, the importance of robust ethical frameworks and governance mechanisms will only grow, ensuring that AI technologies are developed and used in ways that benefit society. So, it’s time to re-invest in your records and information management organisational practice and capability.

References

1. Archives New Zealand. (2023). Atamai hangahanga me ngā mauhanga tūmatanui

2. Artificial intelligence and public and local authority records. https://www.archives.govt.nz/manage-information/how-to-manage-your-information/implementation/artificial-intelligence-and-public-records

3. Floridi, L. (2010–2022). Various works on the ethics of information and the philosophy of the infosphere. [See e.g., The Ethics of Information (2013); The Logic of Information (2019).]

4. Floridi, L., & Cowls, J. (2019). A unified framework of five principles for AI in society. Harvard Data Science Review, 1(1). https://doi.org/10.1162/99608f92.8cd550d1

5. Freedom of the Press Act of 1766, Swedish legislation regarded as the world’s first law supporting the freedom of the press and freedom of information. https://www.britannica.com/topic/Freedom-of-the-Press-Act-of-1766

6. General Data Protection Regulation (GDPR) https://gdpr.eu/what-is-gdpr/

7. International Organization for Standardization. (2016). ISO 15489-1:2016 — Information and documentation – Records management – Part 1: Concepts and principles. https://www.iso.org/standard/62542.html

8. International Organization for Standardization. (2017). ISO 23081-1:2017 — Information and documentation – Metadata for records – Part 1: Principles. https://www.iso.org/standard/67904.html

9. National Archives (UK). (2021). Managing the risk of legacy and emerging technologies: Artificial Intelligence.

https://www.nationalarchives.gov.uk

10. National Archives and Records Administration (NARA). (n.d.). Guidance on managing records in cloud computing environments and with artificial intelligence.

https://www.archives.gov

11. Philip C Brooks, "What Records Shall We Preserve?": in which he introduced the concept of the "life history" of records, a concept later generally called the "life cycle of records." Brooks 1940, see also Brooks, Philip Coolidge. “The Selection of Records for Preservation.” American Archivist 3, no. 4 (October 1940): 221–234. https://doi.org/10.17723/aarc.3.4.u77415458gu22n65."

12. Public Records Act 2005 https://www.legislation.govt.nz/act/public/2005/0040/latest/DLM345529.html

13. Research Data Alliance International Indigenous Data Sovereignty Interest Group. (2019). CARE Principles for Indigenous Data Governance. https://www.gida-global.org/care

14. Schwartz, J. M., & Cook, T. (2002). Archives, records, and power: The making of modern memory. Archival Science, 2(1–2), 1–19. https://doi.org/10.1007/BF02435636

15. Te Mana Raraunga – Māori Data Sovereignty Network. (n.d.). Principles of Māori Data Sovereignty. https://www.temanararaunga.maori.nz

16. The AI Act (Regulation (EU) 2024/1689 https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai#:~:text=The%20AI%20Act%20(Regulation%20(EU,regarding%20specific%20uses%20of%20AI