How Artificial Intelligence Will Affect Classification Systems
Artificial intelligence (AI) could transform how we classify, organise, connect and retrieve knowledge. Its influence could extend across libraries, education, science, social media, research, business and almost every other field.
The most important change is this:
Classification systems could evolve from relatively static structures, designed and maintained by humans, into interconnected systems that use AI to understand, organise and relate knowledge.
To understand this transformation, we need to examine how classification works today, what AI can change, and what the future might look like.
1. From Traditional Classification to AI-Powered Classification
Traditional classification systems organise information according to established rules and categories. AI can help classify information by interpreting its content, context and relationships.
In a traditional system, people define the categories and determine where subjects belong. When new subjects emerge, the classification may need to be revised.
AI-assisted classification introduces additional capabilities. AI can analyse documents, identify their subjects, suggest appropriate categories and recognise relationships between concepts.
More advanced systems could also identify emerging fields of knowledge, suggest new categories and recommend changes to existing classifications.
For example, consider a traditional subject hierarchy:
• Science
• Natural sciences
• Social sciences
• Technology
• Computer science
• Artificial intelligence
• Machine learning
• Natural language processing
• Computer vision
This structure is useful because it places subjects within broader fields.
However, knowledge is more complicated than a single hierarchy suggests.
Artificial intelligence is connected to computer science, mathematics, neuroscience, linguistics, philosophy, medicine, economics and law.
Where should artificial intelligence belong?
A traditional classification system might place it primarily under computer science. However, AI is also relevant to numerous other disciplines.
AI can help identify these connections and represent them alongside the main hierarchy.
The fundamental shift is from classifying information according to where it belongs to also representing how it relates to other knowledge.
2. From Fixed Hierarchies to Interconnected Knowledge
A hierarchy organises knowledge into broader and narrower categories. A knowledge graph, by contrast, represents concepts and the relationships between them.
For example:
• Artificial intelligence uses machine learning.
• Machine learning draws upon statistics.
• Artificial intelligence is applied in medicine.
• AI in medicine raises questions about ethics and regulation.
• Computer vision enables the analysis of medical images.
These relationships create a network of interconnected concepts.
AI can help identify relationships by analysing research papers, books, websites and other sources of information.
It could also help distinguish between different types of relationships, such as:
• A subject belonging to a broader discipline.
• One concept being a component of another.
• A technology being used in a particular application.
• Two disciplines sharing methods or concepts.
• A new field emerging from several established disciplines.
This would allow classification systems to represent knowledge more flexibly.
Instead of forcing every subject into one position, a system could preserve its main classification while connecting it to other relevant fields.
3. Connecting Different Classification Systems
One of the most significant possibilities is that AI could help connect classification systems that currently operate independently.
Different institutions have developed classification systems to meet different needs.
For example:
• Dewey Decimal Classification organises library materials by subject.
• Library of Congress Subject Headings provides standardised subject headings for describing information resources.
• Wikidata represents identifiable concepts and their relationships in a structured knowledge base.
• The ACM Computing Classification System organises topics within computing research.
• The International Patent Classification organises patents according to their technical subject matter.
These systems differ in terminology, structure and purpose.
AI could help identify equivalent or related concepts across them.
Consider the subject of artificial intelligence.
A social media user might write #AI. A researcher might use the term artificial intelligence. A library catalogue might use an authorised subject heading, while a computing research database might assign a specialist classification.
AI could analyse the terminology and descriptions to determine which records concern the same subject.
It could also recognise relationships between different concepts without treating them as identical.
For example, artificial intelligence, machine learning and generative AI are related, but they are not interchangeable terms.
Connecting classification systems would make it easier to discover relevant information across different databases, institutions and platforms.
However, connecting classifications does not mean merging them into one system. Their differences may be important because each system serves a particular purpose.
4. Developing a More Comprehensive Classification of Human Knowledge
AI could help improve how we represent the relationships between academic disciplines.
Human knowledge is commonly divided into broad areas such as the natural sciences, formal sciences, social sciences, humanities and applied disciplines.
These include subjects such as:
Natural sciences
• Physics
• Chemistry
• Biology
• Astronomy
• Earth sciences
Formal sciences
• Mathematics
• Logic
• Statistics
• Computer science
Social sciences
• Psychology
• Sociology
• Economics
• Political science
• Anthropology
Humanities
• Philosophy
• History
• Literature
• Linguistics
• Religious studies
Applied disciplines
• Engineering
• Medicine
• Education
• Law
• Architecture
These divisions are useful, but they are not universally agreed upon or entirely consistent.
Computer science, for example, includes mathematical theory, engineering, cognitive science and the study of human behaviour.
AI could help identify these overlaps and represent them explicitly.
It could also analyse large collections of research to identify:
• Emerging interdisciplinary fields.
• Concepts that appear under different names in different disciplines.
• Connections between previously separate research areas.
• Gaps and inconsistencies between classification systems.
• Subjects that may be developing into distinct academic fields.
Could AI discover new subjects?
AI could identify clusters of research and recurring concepts that suggest an emerging field.
For example, research combining machine learning, biology and materials science might develop into a distinct research area.
AI could detect this development and recommend a new classification category.
However, detecting a cluster of related research does not automatically establish a new academic discipline. Researchers and institutions would still need to determine whether the proposed field is meaningful and useful.
AI could therefore help discover and organise emerging areas of knowledge, while people would continue to play an important role in deciding how those areas should be recognised.
5. How AI Could Change Hashtags on Social Media
This possibility is particularly relevant to Facebook, LinkedIn, Instagram and X.
Today, hashtags are largely informal labels. Users can invent new hashtags, use different hashtags for the same subject, or use the same hashtag to refer to different things.
Consider these examples:
• #AI
• #ArtificialIntelligence
• #MachineLearning
• #GenerativeAI
• #AIEthics
These hashtags refer to related subjects, but they do not all mean the same thing.
An AI-powered classification system could interpret the content of a post and identify its subjects independently of the hashtags selected by the author.
Imagine someone publishes the following statement:
Artificial intelligence is changing how hospitals analyse medical images to identify diseases.
Even without hashtags, AI might identify several relevant topics:
• Artificial intelligence.
• Medical imaging.
• Healthcare.
• Disease diagnosis.
• Computer vision.
It could then connect the post to established subject classifications and related knowledge.
In such a system, hashtags would become one of several signals used to identify a post’s subject.
Other signals might include the text, context, linked sources and relationships between concepts.
This could make it easier to discover relevant content even when people use different terminology.
For example, someone searching for artificial intelligence in healthcare might find useful material tagged with #AI, #MedicalImaging, #HealthcareTechnology or no hashtags at all.
This is a potential application of AI, rather than a description of a universal classification system currently shared by all social media platforms.
6. Making Classification Systems More Dynamic
Traditional classification systems generally change through formal revision processes. This helps maintain consistency, but it can take time to incorporate new developments.
AI creates the possibility of identifying changes in knowledge more quickly.
Consider the development of artificial intelligence itself.
As the field has evolved, areas such as machine learning, deep learning and generative AI have become increasingly prominent.
An AI-assisted classification system could monitor research publications, technical terminology and emerging relationships to identify developments that may require new categories.
It could recommend:
• New subject categories.
• Revised definitions.
• Additional relationships between disciplines.
• Connections between older and newer terminology.
• Changes to the boundaries between established fields.
However, constant change could make a classification system unstable.
A new term might become fashionable without representing a genuinely new subject. Similarly, an emerging research trend might disappear before it becomes an established field.
A reliable system would therefore distinguish between proposed classifications and officially approved classifications.
AI could identify possible changes, while transparent review procedures would determine which changes should be adopted.
7. Helping Resolve Ambiguity Between Subjects
Classification is not simply a technical problem. Many subjects genuinely belong to multiple disciplines, and experts may disagree about where a subject should be placed.
Consider consciousness.
It can be studied through:
• Neuroscience.
• Psychology.
• Philosophy.
• Cognitive science.
• Artificial intelligence.
• Biology.
Should consciousness be classified under philosophy, neuroscience or cognitive science?
There is no single answer that is appropriate for every purpose.
A library, a university department and a scientific database might legitimately classify the same work differently.
AI could help identify these different perspectives and explain the relationships between them.
It could also represent competing interpretations rather than forcing every subject into one category.
This matters because classification systems influence what people find, how they understand relationships and which perspectives receive attention.
Classification is not merely about organising information. It also influences how people navigate and interpret knowledge.
8. The Risks of AI-Powered Classification
Although AI offers substantial opportunities, it also introduces risks.
Bias
AI systems learn from existing information. If their training data underrepresents particular languages, cultures or intellectual traditions, their classifications may reproduce those imbalances.
For example, a system trained mainly on English-language academic literature might inadequately represent research and concepts developed in other languages.
Incorrect relationships
AI might identify two concepts as closely related because they frequently appear together, even when their relationship is superficial or misleading.
The fact that two subjects occur together does not necessarily mean that one explains, supports or causes the other.
Changing meanings
Words and hashtags can change meaning across different communities and contexts.
A system that classifies information without considering context may assign the wrong subject.
Loss of specialist distinctions
A broad AI-generated category might combine concepts that specialists deliberately distinguish.
This could make a classification easier to navigate but less accurate for research.
Control over knowledge
If a small number of companies control the systems used to classify online information, their decisions could influence what people discover and how knowledge is represented.
Important classification systems should therefore support transparency, human review, documented changes and the ability to challenge incorrect classifications.
9. What a Future AI-Powered Classification System Could Look Like
One promising approach would combine several layers rather than replace all existing classification systems with a single hierarchy.
First: Preserve the original information.
The system retains the original document, post, research paper or other resource.
Second: Understand its content.
AI analyses the language and identifies the subjects discussed.
Third: Identify the concepts.
The system determines which specific concepts, entities and topics are being referenced.
Fourth: Connect the concepts to standard terminology.
Different names and expressions are linked to established concepts where appropriate.
Fifth: Apply formal classifications.
The system connects the information to relevant academic, library, scientific or professional classification schemes.
Sixth: Represent relationships.
A knowledge graph connects the concepts to related subjects, applications, methods and disciplines.
Seventh: Maintain oversight.
People can review classifications, correct mistakes, document changes and resolve disagreements.
Technologies such as SKOS and RDF, developed by the World Wide Web Consortium, can help represent concepts and relationships in machine-readable formats.
AI could assist with interpreting information and identifying relationships, while these standards could help different systems exchange and understand the resulting data.
The outcome would not necessarily be one universal classification. It could instead be a network of compatible classification systems that retain their individual purposes.
10. Could AI Create a Universal Classification of Knowledge?
AI could help build a comprehensive, interconnected representation of human knowledge. However, creating a single universally accepted classification would be difficult.
There are three main reasons.
First, knowledge is constantly changing.
New discoveries, technologies and intellectual developments create new subjects and alter our understanding of existing ones.
Second, disciplines have different purposes.
A library classification, a scientific taxonomy and an academic subject hierarchy may need to distinguish concepts differently.
Third, classification involves human judgement.
The way knowledge is divided often reflects cultural perspectives, institutional traditions and intellectual decisions, not simply objective relationships between facts.
For these reasons, a more realistic goal may be a shared knowledge infrastructure that connects existing classification systems while preserving their differences.
Such an infrastructure could give each concept:
• A stable identifier.
• Multiple names and synonyms.
• Connections to broader and narrower subjects.
• Relationships to several academic disciplines.
• Links to relevant research and evidence.
• A record of changes to its definition and classification.
AI could help maintain this network and allow people to navigate it.
The result could be a much more connected representation of human knowledge without requiring every institution to adopt exactly the same classification.
Conclusion: The Most Important Change
AI could transform classification in three major ways.
First, it could automate classification. Information could be analysed and assigned to appropriate categories more efficiently.
Second, it could connect existing classification systems. Subjects could be linked across academic disciplines, libraries, research databases and social media platforms.
Third, it could help develop new ways of representing knowledge. Instead of relying exclusively on fixed hierarchies, classification systems could represent multiple perspectives, interdisciplinary connections and emerging fields.
The most important change may therefore not be the replacement of existing classification systems, but the ability to connect them and make them more responsive to how knowledge develops.
Three Possible Directions for the Future
There are three distinct approaches worth considering:
1. AI-powered classification: AI assigns information to categories within existing systems.
2. AI-connected classification: AI maps concepts across multiple established systems, allowing them to work together.
3. AI-developed classification: AI helps create and maintain a broader conceptual framework that represents relationships across human knowledge.
The first is already a practical application of AI. The second is achievable in specific settings, although comprehensive interoperability remains challenging. The third is a much more ambitious research and governance problem.
The central question is whether we should use AI merely to improve existing classification systems, or use it to help develop an entirely new way of organising human knowledge.
