The CIDOC CRM, and the Question of 'Complexity'

Knowledge Representation in Digital Systems

The CIDOC CRM

The CIDOC CRM standard is clear about its primary concern;

“The primary role of the CIDOC CRM is to enable the exchange and integration of information from heterogeneous sources for the reconstruction and interpretation of the past at a human scale, based on all kinds of material evidence, including texts, audio-visual material and oral tradition. It starts from, but is not limited to, the needs of museum documentation and research based on museum holdings.”

To see how we fall short let’s look at a famous object at the British Museum - the Rosetta Stone. It includes the following fields;

\(\begin{array}{l|l} \textbf{Database Field Label} & \textbf{Entry Stored in the BM Database} \\ \hline \text{Museum number} & \text{EA24} \\ \text{Title} & \text{Object: The Rosetta Stone} \\ \text{Excavator/Field collector} & \text{Pierre François Xavier Bouchard} \\ \text{Production date} & \text{196 BC} \\ \text{Findspot} & \text{Fort Saint Julien, Egypt} \end{array}\)

It uses a static data model which has changed very little over 30 years which immediately creates issues. Historically, in this case, there was no archaeological excavation which revealed the Rosetta Stone (as the language seems to suggest). The fields are ‘one size fits all’ - static attributes of the ‘object’ which are flattened. We infer (fill in the gaps) that there was some event or activity that happened in a location, on a date, by the person named. The database itself has no understanding of field labels - there is no semantic framework. This flattening means that integration with other databases is problematic since different systems have their own individual design with their own static data models, indicating heterogeneity, but creating fields which are local and closed. Synthesis based on isolated field values and subjective nouns instead of shared general concepts will always be flawed.

You don’t have to be a curator in the Egypt & Sudan department to know that Bouchard was not an archaeologist (he was a soldier and engineer), there was no archaeological excavation, and Bouchard was not the finder or collector, but simply the commanding officer of a detachment that came across the Rosetta Stone accidentally while fortifying an Egyptian building previously known as the Burj of Rashid (Tower of Rosetta). This record, at least the structured data part, is hiding a great deal. The only way that the database can provide context is through an unstructured, difficult to search, free text field which itself often lacks essential context.

The Rosetta Stone - (c) Trustees of the British Museum

In contrast the CIDOC CRM focusses on event or activity concepts which are universal across all CIDOC CRM knowledge graphs, because reality is exposed by human events and activities. For example, we could see what other event or activity occurred around the same space/time or different ones, for a comparative view of society at different levels and vantage points. The data representations are meaningful to both human and computer regardless of local knowledge. This is the process of unflattening the data that provides meaning across different disciplines and crucially, from the perspective of synthesis, allows the computer to interpret the data. This unflattening is what has unfortunately been labelled as complexity! This is a misdirection.

In CIDOC CRM terms we can stop thinking about a shopping list of fields and start with events and activities.

1. Works (activity) were performed by the French Soldiers to expand the fortifications (activity) of Fort Saint Julian ready for a battle (an event).

2. As part of those works the Rosetta Stone was accidentally discovered (activity) and removed (activity).

3. The timespan for these activities was July 1799

4. The group had a member (a relationship) - Pierre-Francois Bouchard who also participated in the activities, but in the role of the commanding officer, not someone who actually found the artifact.

In 1801 the British, with the Ottoman army, defeated the remaining French army in Alexandria. This is where the French stored the Egyptian antiquities they had captured. A treaty was subsequently signed but the British had to hunt down the Stone which had been hidden by the French General, Menou. Once located it was transported on a captured French frigate to Portsmouth and then presented to the British Museum by King George III. This of course talks to the legal ownership and provenance of the Rosetta Stone today and there has been a consistent campaign to return it to Egypt.

Here are some more fields from the same record.

\(\begin{array}{l|l} \textbf{Database Field Label} & \textbf{Entry Stored in the BM Database} \\ \hline \text{Associated names} & \text{Named in inscription: Ptolemy V Epiphanes} \\ \text{Acquisition name} & \text{Donated by: George III, King of the United Kingdom} \\ \text{Acquisition date} & \text{1802} \\ \text{Department} & \text{Egypt and Sudan} \\ \text{BM/Big number} & \text{EA24} \\ \text{Registration number} & \text{.24} \end{array}\)

In this part of the record the database records the acquisition which comes via King George III. It doesn’t, and can’t, adequately attempt to explain all the details of provenance which involve events and activities and their context. In fact the legal transfer to the British Empire, that the British Museum cites as justifying its presence in the Museum, is the 1801 Capitulation of Alexandria. Article 16 indeed transfers ownership. The document was signed by General-in-Chief Jacques-François Menou for the French Empire, Lieutenant-General John Hely-Hutchinson, and Admiral Lord Keith for the British. It was also signed on behalf of the Ottoman Empire by Kapudan Pasha Küçük Hüseyin Pasha (the Grand Admiral of the Ottoman Navy). There is no signature representing the Egyptian state - the original owners. At the time the Ottoman Empire held official state sovereignty but with little practical control. They were just paper rulers. It was a Egyptian elite, the Mamluks, who held local power but were deposed by the French in 1798, who then assumed control of Egypt and its assets, as was the custom of war. The Mamluks were therefore not part of the Capitulation agreement. There was no legally recognised Egyptian state.

Complexity or Richness

For these fields a simple relational database would spread fields across various tables only related by technical keys which allow queries across them. These table relationships have no semantics and events are also flattened. Creating these tables and keys requires database administration knowledge including a knowledge of ‘third normal form’ (3NF). The entire goal of 3NF is to eliminate data redundancy (unnecessary duplication) for technical efficiency - efficiency is prioritised over meaning. This focus means it is often difficult to understand the abstract fields at this data level.

The actual relational version of the British Museum database is much more complicated and impenetrable - and it is fixed. It took a considerable amount of time and skill to create it.

The British Museum collection system relational data model

In reality, although traditional databases are flat and reductive they often require highly complicated models. The technical complexity is designed to ensure consistency and predictability, particularly important in financial and transactional systems. To do this it trades human readability, flexibility and any formal semantics. Once installed it is hardly ever changed, or when it is, the project to do so is complex and expensive. It needs to construct user interfaces which also depend on the stability of the data making the systems more rigid. Regardless of a users knowledge of underlying technology it is this issue which prevents user interfaces that are flexible and which can expand models into new areas of information and knowledge. For the arts and humanities this is bad complexity!

In contrast graph based CIDOC CRM systems are far more intuitive and provide the basis for quality and good complex manipulation by users - subject experts. Rather than being hidden, the model can be exposed to give the user the opportunity to be processual and dynamic. It is very easy to convert the Museum’s flat fields to CIDOC CRM. Technical difficulties relate to the poor quality of the data in collection systems not the ontology. The layout is entirely different - it is an explicit and free flowing graph of concepts and relationships. Even for a straight mapping, the meaning is enhanced, although a straight conversion from a traditional database will still hide important information which is why a flexible, expandable graph is required to remedy the limits and constraints.

There are no complicated technical table connections trying to link abstract fields. In this system the database is far less complicated and remains coherent the larger the database gets. You can focus on one part of it and understand it. There is just a simple flow of concept - relationship and concept. An object was produced by an activity, the activity had a timespan, the object carries words (linguistic object), the words include a part which says “Ptolemy V Epiphanes”, and so on. You can expand and add as many relationships as you like - the process is consistent. What this means is that you can produce tools that allow this process to be performed by users who can expand the graph as they generate new knowledge!

When I upload just the diagram above to an AI engine it can lookup the CIDOC CRM references and interpret it. This is what Gemini, for example, says;

“1. The Classes (The Boxes)

The codes starting with “E” represent standard real-world entities (classes). I know exactly what kind of information belongs in them based on their definitions:

E19 Physical Object tells me we are dealing with a material thing you can touch.

• E21 Person tells me this is an individual human being with agency.

• E7 Activity tells me something explicitly happened—it represents an action bound by time and space.

2. The Properties (The Arrows)

The codes starting with “P” represent the semantic links (properties) that establish the exact relationship between the boxes. They dictate how the data must be read, always moving from the subject to the object:

• P12 occurred in the presence of explicitly connects the Finding the Stone activity to the physical Rosetta Stone itself.

• P11 participated in connects Bouchard to that same finding activity, defining his role as a historical actor who was physically there, rather than a passive label in a database column.

• P129 carries and is about tell me the precise hierarchy of information: the physical stone (E19) carries the text (E33), and that text is intellectually about the pharaoh (E21).

By following these standardized paths (Subject > Property > Object), the diagram acts like a sentence diagram for history. Because the CIDOC CRM framework uses these exact same rules every time, anyone (or any semantic computer system) can read this graph and reconstruct the identical narrative without needing to guess the local context. I am using the official CIDOC CRM (Conceptual Reference Model) definitions and properties to interpret the diagram.”
Google Gemini (July 2026)

Consider this graph, which extends well beyond the museum record.

Again, here’s Gemini’s interpretation of the graph (as a diagram) on the ontology references

“The Ancient Origin: Political Crisis, Influence, and Creation (196 BC)

The top portion of the graph details the complex web of social, mental, and physical factors that led to the creation of the Rosetta Stone:

• The Catalyst (Socio-Political Crisis): There was a native insurgency (EO1 Situation Type: Native Insurgency).

• The Psychological Pressures (Inner Influence): This insurgency threatened IN33 Societal Influence (the political leverage of the general assembly of Egyptian priests) and triggered IN34 Inner Influence (Ptolemy V’s practical need to compromise with them to maintain stability). Both the societal threat and his need to compromise led to Ptolemy V avoiding further conflict (IP52 led to avoiding).

• The Power Negotiation (Communicative Influence): The Assembly of Priests (E74 Group) exerted political leverage to negotiate (IN32 Communicative Influence). This influence directly pressured King Ptolemy V (E21 Person) and was specifically induced by the Council of Memphis backing Ptolemy V (E5 Event), which occurred in 196 BC (E52 Time-Span). This council was contributed to by societal leverage and motivated by the concept of Pharaonic Divine Legitimacy (E28 Conceptual Object).

• The Creation: Under these multi-causal pressures (IP50 has influenced), Ptolemy V’s Macedonian Greek ethnic group (E74 Group) and the Egyptian Priesthood (E74 Group) collaborated. This resulted in the drafting of the decree (E65 Creation), which produced the physical text itself: E33 Linguistic Object: The Decree Text.

• The Physical Manifestation: This linguistic text was inscribed onto (P128i is carried by) the physical slab of granodiorite: E19 Physical Object: The Rosetta Stone.

2. The Modern Rediscovery: Fortification and Repurposing (1793–1799)

The middle-left portion of the graph jumps forward in time to depict the rediscovery of the stone by French forces:

• The Military Operation: In July 1793 (E52 Time-Span), the French Engineering Detachment (E74 Group), which included Lieutenant Pierre-François Bouchard (E21 Person), was carrying out the activity of building fortifications (E7 Activity: Building Fortifications).

• The Discovery: This construction work consisted of clearing debris (E7 Activity: Clearing Debris) at Fort Saint Julien in Egypt (E53 Place). The physical Rosetta Stone (E19 Physical Object) was present at this debris-clearing event.

3. Change of Context and Conceptual Use (4th Century BC vs. 18th Century AD)

The bottom portion of the graph tracks how the physical stone’s function and identity changed over time via formal attribute assignments:

• Public Monument (4th Century BC): During its original classical context (E52 Time-Span: 4th century BC), the physical stone’s presence (E93 Presence) was classified by an attribute assignment (E13 Attribute Assignment) as having the type (crm:P2_has_type) of a Public Monument (E55 Type).

• Building Material (18th Century): By the time the French uncovered it (E52 Time-Span: 18th century), its physical presence (E93 Presence) in the ruined fort was classified by another attribute assignment (E13 Attribute Assignment) as having the type (crm:P2_has_type) of mere Building Material (E55 Type).”

Google Gemini (July 2026)

Of course all of this can be presented in a suitable web environment. If the same process were applied to a relational data model, then although the tables names and field labels could be read, the essential nuance (the ‘good’ complexity) would be lost because a relational diagram, like the British Museum one above, is a map of technical and organisational logic, while the CIDOC CRM graph is a map of knowledge. The former had built in obsolescence, while the latter is future proofed.

This structured data explains the information contained in the free text field of the British Museum record. With only the data and a copy of the ontology references, a narrative is produced that is clear, meaningful and computer interpretable. Although generative AI cannot understand language, the RDF data is rich and precise, and the conceptual descriptions in the references clear enough, for an AI engine to explain the graph eliminating hallucinations, in a way it wouldn’t be able to for the relational model above. It is far from perfect, of course, but the point is that it can read the data like text, albeit using probabilistic algorithms, but which are both human curated and systematic. Instead of a collection of isolated data facts this is a rich data narrative. It includes some of the historical conditions which led to the creation of the Rosetta Stone, and the power balance between the Greek Pharaoh and the Egyptian priesthood.

Unlike the text field it is part of the structured data that can be queried precisely and compared with other data. Moreover, the huge difference is the ability to synthesise with other graphs created by others, and be constantly expanded and updated. Tools are available that make this more intuitive and which makes use of its semantic qualities.

In fulfilling this objective it provides a quality, highly semantic (meaningful) and scientifically precise way of representing, not just information, but knowledge. The difference is that information, like that placed in a traditional tabular database, is simply the organisation of atomic facts. Knowledge is something altogether different and encompasses argument, interpretation, wisdom, and explanation. Overall the greater the context, the richer the representations, the greater the points of synthesis that can be established. The narrower the context, and the more that we rely on specific categories, the less likely and meaningful the synthesis.

Processual Thinking

Knowledge synthesis relates to a significant and continuing goal of museum curators since the 1968 conference at the Metropolitan Museum called, ‘Computers and their potential applications in museums ’. As information and knowledge have become more important, interdisciplinary, and vast, so this goal gained momentum across the historical sciences. For example, the historian Eric Hobsbawm said,

“...the wider the range of human activities which is accepted as the legitimate concern of the historian, the more clearly understood the necessity of establishing systematic connections between them, the greater the difficulty of achieving a synthesis. This is, naturally, far more than a technical problem of presentation, yet it is that also.”

Hobsbawm, Eric. On History. Abacus, 1998.

Computers in the 1960s created a new impetus for implementing syntheses through computer networks but the importance of knowledge synthesis has ancient origins. In particular a form of processual thinking (interconnected and dialectic) was foundational to understanding the world in Ancient China and was practised by Laozi and through the Daoist tradition. In Chinese philosophy reality is understood as a dynamic web of interconnected processes. Some aspects of this were captured in Ancient Greece, by philosophers such as Heraclitus, particularly the dynamic nature of a world always in flux, and even in some parts of Aristotle’s work in which the whole was more than the parts.

Processual thinkers in the medieval period were seen as the link between medieval mysticism and modern science. Notably, in the arts and humanities we can point, for example, to the Renaissance polymath and artist Leonardo da Vinci - an important figure in the history of processual thinking. He understood that knowing any discipline well requires an understanding of its connections with everything else. This understanding is now firmly part of modern arts and humanities and can be seen in the work of artists like Olafur Eliasson, or in the music of David Byrne, two highly processual thinkers. Yet the eighteenth century painter Joshua Reynolds, the first president of the Royal Academy, steadfastly adhered to the metaphysics of ‘Mechanical Enlightenment’ in his quest for the definitive universal rules of art. It is Reynolds rather than Da Vinci who represents our positivist reductive computerised world today. The strength of this legacy means that technologists make a feature of eliminating complexity from digital data systems, thus blocking synthesis and hiding human experience - experience which is naturally and essentially complex.

The likes of Leibniz, Hegel, and Marx made significant contributions to understanding the world as systems of relationships. However, the Western philosophical tradition has been dominated by an intrinsic, analytic, static, substance based approach, and the quest for establishing immutable rules that govern everything in the universe. In modern times this has been underpinned by the highly influential work of René Descartes and Isaac Newton who painted a mechanical world.

In the nearly 60 years since the conference we have not come close to achieving a meaningful digital synthesis. The reasons are various and therefore complex but traditional database technology is highly complicit in this, not just in the reductive way it represents the world, but how it reinforces a particular mode of computational thinking. The first priority of computers in the 1960s was to support the commercial sector, burdened by increased bureaucracy which was impeding growth. Corporate instrumentation required representational qualities that were deliberately simplistic and specifically focussed. The adoption of the same technology in the humanities has been limiting, preventing subject and technical experts from collaborating and fully exploring developing use cases due to constraints around cost of ownership. However, the database market and the option that it now provides opens up new possibilities.

The vigorous stripping of nuance, context and conflicting perspectives contained in social, cultural and historical knowledge through data has been achieved by a tacit consent on the basis of cognitive overload. This is born out by the many data aggregation projects over the years which have failed to provide sufficient benefit beyond a simplistic search hub, reference site or administrative resource. However, when information and knowledge is presented well the issue of cognitive overload disappears - overload is a symptom of poor quality, not richness. Those creating flat digital resources don’t see the problem. They prioritise efficiency, scalability, and standardisation over knowledge. This wasn’t what Hobsbawm was talking about when he looked to technology to solve the problem of historical synthesis.

The significance of the CIDOC CRM is that it provides a means, with some new skills, to achieve the vision by combining a quality cognitive framework with flexibility. The reasons why museums need to return to the path of synthesis include the growing lack of quality and intelligent ‘complexity’ which currently set it apart from the arts and heritage sector that it is meant to support. It includes the need to represent inclusivity and diversity - a museum cannot represent its collection properly without context and this is obviously lacking in the digital sphere despite the increased reliance on data networks.

The CIDOC CRM, unlike traditional databases, does not, in its native graph form, ‘simply’ organise information in a flattened form (a collection of nouns), but supports the representation of knowledge using relationships (the use of verbs). This representation, which we call modelling (not data modelling) is something we all do in the historical disciplines, but in our heads and often with the help of free form mediums, like paper, paint, a word processor, or a whiteboard, in our own individual subjective way. It is complexity managed using our own internal vocabularies, but when structured for use in computer systems we need to learn a specific set of concepts - a type of cognitive ‘complexity’ but aimed at quality and clear representation. But the CIDOC CRM is very far from the complexity of say, learning a foreign language, or programming code. There are some simple rules to flexibly use concepts that have been, in the case of CIDOC CRM, derived from real practice. Learning to use this semantic framework is simply an extension of what arts and humanities subject experts already do. As such it becomes a simple question of what benefits it provides - linked to the design of appropriate ‘psychological tools’.

A psychological tool is one which helps us interact with the world. It includes symbolic tools like language and logic, but also tangible aids like maps, index cards, canvas and brushes, spreadsheets, and so on. They all influence and change the structure of human thought, some better than others. For example, our brains do not think in terms of tables of rows and columns or definitive categorising nouns. Yet the relational database is used to squeeze ‘complex’ multi-dimensional patterns of thinking into a ‘simple’ two dimensional grid and this creates limits to the extent of its representational capabilities.

They are not actually perceived by their creators as psychological tools with the ability to impact negatively on thinking, but as neutral technical tools. Therefore the friction they create when attempting to build humanistic knowledge is ignored. The fact that tools are not neutral is a recurring theme of critique, whether that be Heidegger’s ‘enframing’, in which technology converts our vision of the world to an inventory of raw resources to be utilised, or McLuhan’s landmark, ‘The Medium is the Massage’, which charts the representation of history erroneously, as uniform and linear; or more overtly the control that technology has over us, as argued by Donna Haraway in the ‘Cyborg Manifesto’, which argues that we are incorporated into the machine and therefore we lose our critical insight of it. The Californian magazine of the 1980s, ‘Processed World’, asked the question, “Are you doing the processing ...or are you being processed?

The Opportunity for Heritage IT Departments

An example of the negative effects of traditional psychological tools is the standard collection management system. It is an accepted part of the museum infrastructure despite its obvious flaws. They operate as security and tracking inventories rather than engines of human knowledge - but are published for public communication. The key factors for modern museums in terms of digital collection systems is;

1. Is the object where it says it is.

2. Do we have legal title to the object.

3. What is its condition and value.

4. Can we identify the object.

The term ‘Tombstone data’ is applied readily without any sense of embarrassment.

However, museums create rich content for books, exhibitions, discrete projects and web pages. They have become content generators under a traditional editorial function, but museums have never really tackled underlying architectural issues which today include, context, decolonisation, diversity and inclusion. Museums don’t recognise the issue with collection databases because they take a classical view that data is neutral and limited, and if needed they correct particular issues by using the traditional but inadequate fillers called ‘free text fields’ to balance things up. Museum administrators don’t understand (but curators absolutely do - (Griffiths, Antony. ‘Collections Online: The Experience of the British Museum’. Master Drawings 48, no. 3 (2010): 356–67.) - that the combination of mechanism, structure and intrinsic content create a deep seated problem which is far from neutral and which contributes to an architectural weakness. Their only form of deflection is the lack of general interest generated by the lifeless information itself. There will surely become a point where the ubiquitous museum collection online becomes a redundant museum reference.

Representing the epistemology and agency of different cultures is not the specialism of database administrators which requires an expansion that museums have yet to explore. It is not acceptable just to represent different indigenous cultures with ‘mystical beliefs’ rather than explore different but valuable perspectives on reality which elude Western analytic thinking. This can be done by representing real activities and events and explaining the purpose of these events, which can be synthesised across cultures providing a balanced approach. Good example includes issues of ecology, our relationship to nature and human relations, which seem to have been sidelined in countries where computer technology reigns. In this digital space museums are clearly struggling with their stated missions.

This struggle is evidenced by a common approach that museum IT departments use because of the lack of resources, organisational fragmentation and a previous lack of R&D into graph databases. This is called the “like for like” doctrine, designed to maintain a status quo and consistency across technology upgrades. It is a common pattern which is focussed on infrastructure maintainability without the overhead of looking at data which has accumulated content problems over time. It was common when I worked in the British Museum IT Department, and is common in other museums. It prevents the digital representation of new thinking and evolving knowledge. ‘Like for like’, is a symptom of the inability of museums to expand their current architectural constraints.

Traditional collection systems, because they address legal compliance issues are more permanent than their curators files, notebooks and accumulated knowledge. A former Keeper at the British Museum even commented that we would be better just publishing the images given the quality and nature of the data. They are content to take part in external research projects, but they rarely result in any change to their internal systems.

Because of the way that collection systems are procured, as administrative systems, implementing semantics is seen as a technological upgrade cost, rather than an essential philosophical decolonisation of their data and a responsibility to their professional and non-professional audiences. Until software developers build tools that treat curators as ‘knowledge creators’ rather than data-entry clerks, Edward Fry’s 1970 warning, reviewing the Metropolitan conference, of simply transferring index card type data (the basis of many museum collection systems) to digital collection systems, will remain true, in that,

“its inclusion on a museum network would…be of dubious value.”

There is a now a great opportunity for heritage IT departments to align their systems with organisational subject experts in a collaboration which reforms their data architecture such that it prioritises knowledge, rather than perpetuates the status quo. Graph systems are now being implemented at four times the rate of relational database systems across all sectors because of the urgent demands for flexibility and integration in many modern use cases. This flexibility allows new knowledge to be entered into systems as it is created in everyday knowledge processes. This is also driven by the growth of AI which can more readily and precisely interpret semantic data. Graph systems (underpinned by quality ontologies like the CIDOC CRM) can provide rich and evolving content which is efficiently produced through connecting the systems to everyday knowledge generation. They provide benefits internally and externally in a way that can’t be matched by traditional systems.

Why does it matter?

Increasingly data has become the medium though which people perceive and mistake as reality. It is pervasive in that it flows through social media and digital news outlets. It influences search engines, the algorithms of which are controlled by large tech companies. It is used in artificial intelligence tools that provide responses that reflect a conventional view of history as a linear path. Walter Benjamin famously appropriated Paul Klee’s painting, Angelus Novus, renaming it the Angel of History. As the Angel looks to the past instead of seeing,

“the appearance of a chain of events, he sees one single catastrophe, which unceasingly piles rubble on top of rubble and hurls it before his feet”

On the Concept of History, Walter Benjamin

The aim of the Angel is to freeze a moment in time - a dialectic image - and to dwell or linger in order to “awaken the dead and to piece together what has been smashed.” Instead the Angel is forced back first into the future unable to resist the force of homogenous time. The attempt to halt a moment of time involved the creation of a dialectic image, a psychological tool which captures a snapshot in time for the purposes of placing the past against the present. The challenge is to design the appropriate psychological tools which allow this rich complexity to be accessed and used effectively.

Museums and other public knowledge organisations which claim to have an educational mission should be continually sorting through the rubble of history from the vantage point of the present and continually disrupting the linear view in all the channels of communication is uses, including data. It is a social responsibility of all historians and knowledge organisations to review and expand history in the light of new questions in the present. A museum cannot claim to be an educational organisation if it communicates information that hides the context, significance and relevance of it collections.