Concept Exploration is an exploratory thinking approach aimed at discovering concepts that have not yet been fully recognized and possibilities that may have been overlooked by conventional classifications, rather than merely collecting information about already known subjects. It focuses on the characteristics, structures, relationships, and patterns of technologies, business opportunities, research themes, and social changes without being confined to established names or categories.
Conventional research generally begins by defining a subject and collecting information about it. For example, when investigating a particular technology, researchers may examine its market size, competitive environment, patents, and scientific developments. Such methods are effective when the research subject and objectives are clearly defined.
However, when exploring new business opportunities or R&D themes, the fundamental problem may be that we do not yet know what we should investigate. Emerging developments that do not fit established categories may also be overlooked when searches depend on familiar keywords.
Concept Exploration begins with this problem.
For example, several technological or business developments may share common characteristics or patterns of change that have not yet been recognized. Alternatively, by abstracting the functions and structures of a company’s existing technologies, it may be possible to discover similar problems or technologies in previously unfamiliar industries.
Concept Exploration is not simply about investigating things whose names we already know. It is about discovering what we should be looking for.
The term Concept Exploration is used here in a sense distinct from concept testing in marketing or concept inventories used to assess students’ understanding in education. This article explains the approach originally proposed by Mindware Research Institute in the 1990s under the Japanese name 概念調査 (formerly translated as Conceptual Investigation), and its subsequent technological development.
The Origins of Concept Exploration
The idea of Concept Exploration emerged from experience in planning advanced technology seminars during the 1980s and subsequently working in new product and business development.
Planning advanced technology seminars required identifying promising subjects across diverse fields, including information technology, semiconductors, biotechnology, materials, and energy.
However, it was impractical to acquire comprehensive specialist knowledge of every field before selecting seminar topics. Such an approach would have made it difficult to respond to emerging technological developments.
In practice, subjects that attracted considerable interest often shared characteristics that extended beyond their respective technological domains. These included the emergence of new technological possibilities, the removal of existing limitations, and the potential for changes in industrial structures.
The important capability was not necessarily knowing the name of a particular technology in advance. Rather, it was understanding the characteristics shared by promising subjects and finding new developments that matched those characteristics.
Continuous observation of technological and industrial developments also revealed that phenomena previously regarded as unrelated could sometimes be understood through a newly recognized common concept.
These experiences suggested the need for a form of research concerned not only with collecting established facts but also with identifying emerging concepts within ongoing technological and social changes.
This approach came to be called Conceptual Investigation (概念調査). In 1997–1998, a series of articles explored its theoretical foundations with reference to philosophy, sociology, psychology, cognitive science, and related fields.
At that time, practical technologies for implementing Concept Exploration as a computational analytical method were limited.
Subsequent developments in machine learning, including Self-Organizing Maps (SOM), natural language processing, and Large Language Models (LLMs), have made it possible to provide computational support for forms of exploration that originally depended heavily on human reasoning.
What Is a Concept?
A concept is more than a word or a label. It is a cognitive framework through which common characteristics of different objects or phenomena can be recognized and understood as belonging to a meaningful category.
Traditional logic distinguishes between the intension and extension of a concept.
The intension of a concept refers to the characteristics or conditions that define it. Its extension refers to the set of objects to which the concept applies.
Consider the concept of “materials suitable for high-temperature environments.”
Its intension might include the ability to maintain physical properties at specified temperatures, preserve sufficient mechanical strength, and remain chemically stable.
Its extension would include the various materials that satisfy those conditions.
Conventional searches often begin with known material names or established material categories.
However, even when the name of the required material is unknown, describing its necessary characteristics may make it possible to discover previously unfamiliar candidates.
The extension of a concept may also change as knowledge increases or technologies advance.
New objects may be discovered, or common characteristics may be identified among objects that were previously considered unrelated.
Such developments can lead to revisions of a concept’s intension or extension.
Concept Exploration is concerned with these processes of concept formation, transformation, and reorganization.
Discovering Emerging Concepts
One direction of Concept Exploration involves continuously observing technological and social developments to identify newly emerging concepts.
Consider changes in manufacturing businesses.
Traditionally, selling industrial equipment and maintaining that equipment were often treated as separate business activities.
However, as remote monitoring, data analytics, and predictive maintenance become more widespread, new business arrangements become possible in which companies provide not merely equipment but services that support its continuous operation.
In this case, it is important to examine more than the appearance of individual technologies or services.
The relationships among products, maintenance activities, data, contracts, and customers may themselves be changing.
Similar transformations may occur across different industries before a commonly accepted name emerges.
Recognizing their shared structures can reveal the formation of a broader concept that goes beyond individual news reports or product announcements.
However, the appearance of a new term does not necessarily indicate the emergence of a new concept.
Sometimes an existing technology or business practice is simply given a new name.
Conversely, a familiar term may continue to be used even though the underlying characteristics, relationships, or scope of what it represents have changed substantially.
Concept Exploration therefore examines not only the appearance and frequency of terms but also changes in the objects, characteristics, relationships, and applications associated with them.
Relationship with Horizon Scanning
The continuous observation of emerging developments gives Concept Exploration something in common with Horizon Scanning.
Horizon Scanning is an approach to identifying early indications of potentially important changes in technology, society, economics, policy, and other fields.
Developments that are currently small or fragmented may be recognized as Weak Signals if they suggest the possibility of more substantial future changes.
For example, similar technological or business initiatives may begin appearing independently in different industries.
These developments might represent the early stages of a broader transformation.
Concept Exploration seeks not only to detect such signals but also to understand the common characteristics and relationships that connect them.
Observing individual events and forming a new concept from those observations are different thinking activities.
For example, identifying several companies that have introduced similar services is one task.
Understanding how those services collectively indicate a transformation in the relationship between businesses and their customers is another.
This does not mean that Concept Exploration is inherently superior to Horizon Scanning.
Horizon Scanning includes established approaches for systematic information gathering, signal assessment, and analysis of future implications.
Concept Exploration can use similar information-gathering processes while placing particular emphasis on the formation and transformation of concepts and the discovery of common structures across different domains.
The Difference Between Factual Research and Concept Exploration
Conventional factual research generally defines the research subject and questions in advance, then collects information relevant to those conditions.
For example, researchers may investigate scientific papers, patents, products, companies, and markets associated with a particular technology.
This is effective when the subject is known and the objective is to establish or verify facts.
Concept Exploration begins at an earlier stage.
Rather than treating established names and categories as fixed, it examines what should be recognized as a meaningful subject of investigation.
| Aspect | Factual Research | Concept Exploration |
|---|---|---|
| Fundamental question | What is known about this subject? | What should be recognized and investigated? |
| Starting point | Known names, categories, and research questions | Characteristics, structures, relationships, and patterns of change |
| Primary focus | Information about known subjects | Emerging concepts, unknown objects, overlooked relationships |
| Typical activities | Searching, matching, measuring, verifying | Observing, comparing, abstracting, reconsidering classifications, exploring |
| Main results | Knowledge of established facts | New research subjects, conceptual structures, hypotheses |
However, factual research and Concept Exploration are not competing approaches.
When Concept Exploration identifies a new possibility, factual research is needed to determine whether it is supported by evidence.
Conversely, information obtained through factual research may reveal unexpected similarities or relationships that initiate a new round of Concept Exploration.
The two processes can therefore be repeated and combined.
Searching for Things Whose Names We Do Not Know
Another direction of Concept Exploration is to search for unknown objects based on their characteristics or structures, without knowing their names or technological domains in advance.
Consider an R&D problem in which a conventional technology stores a necessary substance at a particular location and releases it when required.
This approach has a limitation: once the stored substance has been consumed, the same location may not be able to respond to a subsequent event.
Instead of beginning a search with known material or product names, we can describe the required function more abstractly:
A mechanism that can repeatedly deliver a required substance to a location where it is needed.
This description is not limited to a particular material or industrial sector.
It may suggest investigating biological transport networks, fluid control systems, microfluidics, distributed supply systems, and other technologies that perform structurally similar functions.
The names of these technologies do not need to be specified in advance.
The search is guided by the required function or structure rather than by a predefined product category.
However, this does not mean that exploration can begin with absolutely no prior knowledge.
Some understanding of the problem, desired function, or relevant characteristics is necessary to guide the search.
Exploring the unknown does not mean searching without any clues. It means using clues that do not depend on knowing the target’s name or technological domain in advance.
Focusing on Structures, Relationships, and Patterns
When searching for similarities across technological domains, the same words or superficial characteristics may not appear in the relevant information.
For example, one field may use the term “repair material supply,” another may discuss “substance transport,” and a third may describe “resource redistribution.”
Although their terminology differs, these systems may share a relational structure:
Resources are stored in one location and repeatedly delivered through a network to locations where they are needed.
Exploring this kind of similarity can draw on ideas from Structure-Mapping Theory.
In 1983, Dedre Gentner proposed Structure-Mapping Theory as a framework for understanding analogy.
The theory emphasizes the importance of correspondences between relational structures rather than merely matching superficial attributes of objects.
Two technologies may use different materials and components while sharing common relationships among their functional elements.
By abstracting those relationships and examining whether they can be applied in another domain, it may be possible to develop new technological hypotheses.
However, structural similarity does not prove that a technology can be transferred directly from one application to another.
Differences in physical conditions, performance requirements, scale, manufacturing methods, and costs may prevent an apparently similar mechanism from being used successfully.
Analogy in Concept Exploration is therefore a means of discovering possibilities that require investigation and experimental validation, rather than a demonstration of technical feasibility.
Two Starting Points for Concept Exploration
Concept Exploration can begin from two broadly different perspectives.
The first is to identify emerging concepts through observation of technological and social developments without specifying a particular target in advance.
The second is to begin with a problem or desired characteristics and search for unknown objects that match those characteristics.
| Discovering Emerging Concepts | Exploring Unknown Objects | |
|---|---|---|
| Starting point | Broad information and observations of change | Problems, requirements, functions, characteristics, patterns |
| Central question | What new developments are emerging? | Where can we find something with these characteristics? |
| Main approach | Continuous observation, comparison, classification, concept formation | Abstraction, analogy, cross-domain exploration |
| Typical results | New concepts, emerging signals, future hypotheses | Unknown technologies, applications, business opportunities, R&D hypotheses |
Although the starting points differ, the two approaches can reinforce each other.
For example, identifying an emerging concept through continuous observation may lead to a search for additional technologies or markets that share its characteristics.
Conversely, exploring unfamiliar technological domains may reveal previously unrecognized common structures or developments.
Concept Exploration can therefore be understood as an iterative process involving observation, concept formation, exploration, and verification rather than a fixed sequence of steps.
The Basic Concept Exploration Process
The specific procedure depends on the purpose and starting point of the investigation, but a general process can be described as follows.
- Define the exploration perspective. Determine whether the objective is to identify emerging concepts or discover unknown objects matching particular characteristics.
- Collect information broadly. Investigate technical literature, products, business activities, and social developments without relying exclusively on established classifications.
- Identify similarities and differences. Compare information to examine common characteristics, relationships, structures, and patterns of change.
- Form and abstract concepts. Express the identified similarities in a form that is not limited to particular products or technological domains.
- Use concepts to guide further exploration. Search for cases, technologies, applications, or problems that match the abstracted characteristics or structures.
- Verify candidates against evidence. Examine scientific papers, patents, technical documents, experimental results, and other sources to assess the validity of the hypotheses.
- Update the conceptual structure. Incorporate new information and revise the definitions of concepts or the direction of exploration when necessary.
An important principle is not to treat the initial concept definitions as permanent.
During exploration, previously overlooked characteristics may turn out to be important.
Objects initially considered unrelated may reveal meaningful similarities.
Conversely, objects initially classified under the same concept may need to be distinguished.
Concept Exploration is therefore an iterative process in which the conceptual framework itself can change.
Relationship with the KJ Method and Grounded Theory
Concept Exploration shares some characteristics with qualitative methods such as the KJ Method and Grounded Theory.
The KJ Method organizes diverse pieces of information, often recorded on cards, and examines their relationships to identify meaningful groups and overall structures.
Grounded Theory develops concepts and categories from empirical data and examines their relationships to construct theoretical explanations.
Both approaches can generate new classifications and interpretations rather than merely assigning information to predefined categories.
In this respect, they are related to Concept Exploration.
However, Concept Exploration is not another name for either method.
The KJ Method places particular emphasis on organizing information into meaningful structures.
Grounded Theory focuses on developing theory grounded in data.
Concept Exploration emphasizes the discovery of emerging concepts and previously unknown targets for further investigation.
These approaches can nevertheless be combined.
For example, information collected from diverse sources could be organized using an approach similar to the KJ Method.
The resulting structure might suggest a new concept, which could then be used to guide additional exploration.
Visualizing Conceptual Structures with Machine Learning
One direction in the computational development of Concept Exploration has been the use of machine learning to structure information.
During the 1990s, Self-Organizing Maps (SOM) were investigated as a means of visualizing similarities and distributions within large collections of information.
SOM is a neural network method that represents similarity relationships in multidimensional data on a lower-dimensional map.
Subsequently, text embeddings and methods such as Growing Neural Gas (GNG) made it possible to construct models of conceptual similarity from large collections of text.
For example, technical documents can be converted into embedding vectors, and their distribution can be modeled using GNG or related approaches.
This can support the visualization of groups of semantically similar technological information and the relationships among those groups.
If information is collected continuously, changes in the resulting structures may help reveal emerging topics or new relationships among existing groups.
However, the appearance of a new cluster does not automatically establish that a new concept has emerged.
Changes in source coverage, terminology, embedding models, or analytical parameters may also affect cluster structure.
Likewise, documents that are semantically similar do not necessarily share the same technical principles or causal relationships.
Machine-learning-based concept maps should therefore be treated as tools for identifying candidates for exploration, not as proof of conceptual or technological relationships.
The Development of Concept Exploration Through LLMs
The emergence of Large Language Models has substantially changed the practical possibilities for Concept Exploration.
Traditionally, abstracting technological characteristics and searching for analogous cases in unfamiliar fields required the researcher to possess knowledge extending across many domains.
LLMs can now assist in extracting characteristics, functions, structures, and operating principles from technical documents and expressing them at different levels of abstraction.
These abstract descriptions can then be used to search for similar mechanisms in other technological domains.
For example, technical documentation for a device might contain functions such as:
- Detecting very small changes.
- Performing continuous monitoring with low power consumption.
- Responding locally when an abnormal condition occurs.
Each function can become a starting point for investigating whether analogous mechanisms exist in unrelated fields.
The names of the technologies to be discovered do not need to be specified in advance.
However, technologies and examples proposed by an LLM may not exist, and explanations of similarity may contain errors.
Search results must therefore be verified against original scientific papers, patents, product documentation, or other reliable evidence.
LLMs should not be regarded as systems that guarantee correct conclusions.
Their value in Concept Exploration lies in helping generate new directions for investigation and hypotheses that can subsequently be tested.
The Development of Cross-Domain Concept Exploration
Cross-Domain Concept Exploration is a more specific development of Concept Exploration directed toward discovering R&D themes and technological applications across different fields.
Conventional technology research typically begins by defining a technological domain and investigating relevant scientific papers, patents, products, and companies.
However, when searching for applications of an existing technology outside its established industry, the names and keywords associated with those applications may be unknown.
Cross-Domain Concept Exploration therefore begins with the characteristics, structures, functions, and operating principles of a technology rather than its established name.
For example, a company may extract functional patterns from its accumulated technological capabilities and investigate whether similar patterns exist in different industries.
The resulting candidates can be examined for their relationships to existing technologies, conditions of applicability, and technical feasibility.
This process can lead to hypotheses for new R&D themes or business applications.
Cross-Domain Concept Exploration gives a concrete procedural form to one important aspect of the original Concept Exploration approach: searching for unknown targets through structures and patterns without being restricted to familiar names or specialist domains.
However, Cross-Domain Concept Exploration does not encompass the entire original concept.
Concept Exploration also includes the broader objective of continuously observing newly emerging concepts and understanding how concepts form and change.
Cross-Domain Concept Exploration can therefore be understood as a specialized development of that broader approach for R&D and technological or business opportunity discovery.
Technical Implementation Through ConceptMiner
ConceptMiner represents an effort to provide computational support for Concept Exploration.
Its approach involves obtaining embeddings from documents and other information and modeling conceptual similarity structures using machine-learning techniques such as GNG.
The objective is to help organize large collections of information and provide a structured overview for exploration.
ConceptMiner R&D combines this kind of conceptual structure modeling with LLM-assisted cross-domain exploration.
For example, a knowledge base and concept map can be constructed from existing technical information.
Technical elements can then be extracted and used to generate exploration questions based on their functions, structures, and characteristic patterns.
These questions guide investigation into candidate technologies from different domains.
The resulting findings can be checked against supporting literature and organized into a broader concept map.
This creates an iterative process in which exploration begins with existing technical knowledge, extends into unfamiliar domains, and incorporates newly discovered information into an expanding knowledge base.
However, ConceptMiner should not be equated with Concept Exploration itself.
Concept Exploration is a methodological approach concerned with how people recognize what they know, what they do not know, and what they should investigate.
ConceptMiner is a technological system intended to implement and support selected parts of that approach.
Limitations and the Need for Verification
Although Concept Exploration emphasizes searching beyond established categories and specialist domains, it has several limitations that must be recognized.
First, all exploration begins from some perspective or set of assumptions. Searching for unknown objects does not mean conducting a search without any prior knowledge. The important task is to recognize the assumptions guiding exploration and revise them when necessary.
Second, abstraction can remove important differences. Overemphasizing similarities among technologies may produce analogies that do not hold under actual technical conditions.
Third, a discovered object or relationship is not necessarily new. It may be unfamiliar to the researcher while already being widely known elsewhere.
Fourth, LLMs and machine-learning models are influenced by the information sources and modeling methods used. Their results may reflect biases in coverage, representation, or classification.
Fifth, discovering a promising hypothesis does not establish its feasibility. Technical investigation, market research, simulation, and experimentation may still be necessary.
For these reasons, Concept Exploration should distinguish clearly between ideas generated through exploration and facts supported by evidence.
When a hypothesis is found to be invalid, that result should be incorporated into subsequent exploration.
Concept Exploration in ThinkNavi
ThinkNavi regards Concept Exploration as a thinking approach for recognizing the limitations of existing knowledge and classifications and discovering emerging concepts and possibilities.
Concept Exploration is not a single analytical procedure that can be completed through a fixed sequence of operations.
Methods such as the KJ Method and qualitative analysis can support the formation of new concepts.
Systems Thinking, Structure-Mapping Theory, and analogy can help examine relationships and structural similarities.
Horizon Scanning and Scenario Analysis can support the investigation of emerging developments and possible futures.
Once new business opportunities or R&D hypotheses have been identified, SWOT and TOWS Analysis, Morphological Analysis, and Multi-Criteria Decision Analysis can help organize, combine, and evaluate them.
The thinking methods introduced through ThinkNavi perform different analytical roles.
Among them, Concept Exploration occupies a distinctive position: it concerns the discovery and revision of the subjects and questions that other methods subsequently analyze.
Advances in AI have made it easier to search, summarize, and classify enormous amounts of information.
However, the ability to answer an established question accurately is not the same as the ability to discover a question that has not yet been considered.
No matter how much information is collected, important possibilities may remain overlooked if the investigation never moves beyond its original categories and assumptions.
The significance of Concept Exploration lies not only in collecting more facts, but in reconsidering the conceptual framework that determines which facts, objects, and possibilities should be investigated in the first place.
Observing the emergence of new concepts and discovering unknown objects through structures and patterns are two different approaches to this broader objective.
LLMs and machine learning provide powerful new technical means of supporting such exploration.
Yet the fundamental requirement remains a human willingness to question established knowledge and classifications and to recognize that important possibilities may exist beyond what is currently known.
Primary English Term: Concept Exploration
Related Expressions: Emerging Concept Monitoring, Cross-Domain Concept Exploration
Related Thinking Methods:
- Horizon Scanning | Discovering Early Signals of Change and Exploring Future Possibilities
- The KJ Method | Building Meaningful Structures from Fragmented Information
- Grounded Theory,
- Structure-Mapping Theory | Understanding Relational Correspondences and Generating New Inferences Across Domains
- Analogy — Connecting Knowledge Across Domains to Generate New Ideas and Hypotheses
- Systems Thinking — Understanding Interdependence, Feedback, and Change Over Time
- Morphological Analysis | Systematically Exploring New Possibilities Through Combinations of Elements
- Scenario Analysis | Exploring Multiple Futures and Developing Strategies Under Uncertainty
- Cross-Domain Concept Exploration
References and Further Reading
- Mindware Research Institute, What Is Conceptual Investigation? (Japanese)
- Dedre Gentner, “Structure-Mapping: A Theoretical Framework for Analogy,” Cognitive Science, Vol. 7, No. 2, pp. 155–170, 1983. Paper
- Barney G. Glaser and Anselm L. Strauss, The Discovery of Grounded Theory: Strategies for Qualitative Research, Aldine, 1967.
- Jiro Kawakita, Hassōhō: Sōzōsei Kaihatsu no Tame ni (The KJ Method: Developing Creativity), Chūō Kōronsha, 1967.
- Teuvo Kohonen, Self-Organizing Maps, Springer, 1995.
- OECD — Strategic Foresight
- OECD — Tools for Agility: Actionable Strategic Intelligence and Policy Experimentation
- ConceptMiner R&D — Cross-Domain R&D Concept Exploration