Some of the most interesting innovations begin when someone recognizes an unexpected connection.

A problem in one industry resembles a problem in another.

A biological structure suggests an engineering mechanism.

A method developed for scientific research appears relevant to organizational decision-making.

A question originally raised about human–AI collaboration becomes interesting when examining autonomous machines.

The connection may be valuable.

But it may also be misleading.

Two situations can appear similar while depending on fundamentally different conditions.

A method may work in one environment and fail in another.

And a shared vocabulary does not establish that two systems possess equivalent capabilities.

This creates an important challenge for cross-domain innovation.

Research Question

How can we recognize useful relationships between different domains while preserving the distinctions that determine whether an idea, method or capability can actually transfer?

The objective is not simply to connect more ideas.

It is to discover which connections deserve investigation — and what would have to be true for them to become useful.

What is cross-domain innovation?

Cross-domain innovation involves applying ideas, mechanisms, methods or knowledge from one field to a problem or opportunity in another.

The source and target may belong to different scientific disciplines, industries, technologies or professional practices.

A manufacturing challenge might benefit from methods developed in logistics.

A materials engineer might investigate structures found in nature.

An organization might examine whether insights from distributed computer systems are useful when designing resilient workflows.

These possibilities do not require the domains to use the same vocabulary.

They require a plausible relationship between the problem being investigated and something understood elsewhere.

Cross-domain innovation overlaps with several established approaches.

Interdisciplinary research integrates knowledge and methods from different disciplines to investigate problems that cannot be adequately addressed within a single field.

Technology transfer can move inventions, methods or technical knowledge toward different organizations and applications.

Design by analogy uses a source problem or solution to inspire investigation of a target problem.

Knowledge recombination explores new combinations of existing knowledge components.

These approaches are related, but they are not interchangeable.

The US National Academies' 2005 report, Facilitating Interdisciplinary Research, describes interdisciplinary research as integration and synthesis across specialized knowledge areas rather than simply placing disciplines beside one another.

For our purposes, the most important question is not which label a connection receives.

It is what that connection actually allows us to investigate.

Why unfamiliar knowledge combinations matter

Innovation does not always require entirely new knowledge.

Existing discoveries may become useful when brought into an unfamiliar relationship.

A method known in one field may address a constraint in another.

Two previously separate technologies may become complementary.

A scientific result may suggest an application that its original authors did not consider.

However, unusual combinations have different levels of plausibility, uncertainty and usefulness.

Documented Observation

In their 2013 paper Atypical Combinations and Scientific Impact, Brian Uzzi and colleagues examined 17.9 million scientific papers. Their analysis associated high scientific impact with combinations that were predominantly conventional while also incorporating unusual relationships between previous bodies of work.

This finding offers a useful perspective on innovation.

Unexpected combinations can matter, but novelty alone is not sufficient.

The study analyzes patterns of scientific publications and citation impact. It does not demonstrate that every unusual combination causes a breakthrough or that the same statistical relationship applies directly to product development.

A second research tradition emphasizes uncertainty.

In his 2001 paper Recombinant Uncertainty in Technological Search, Lee Fleming examined technological invention through component recombination.

The research found greater variability associated with novel technological combinations, including the possibility of both unsuccessful outcomes and breakthroughs.

This is important for cross-domain innovation.

The same unfamiliarity that creates an opportunity for a surprising solution can also make the result difficult to predict.

A novel connection is a reason to investigate, not evidence of success.

A historical example: From plant burrs to fastening technology

One well-known example of cross-domain inspiration is the development of hook-and-loop fastening.

The source domain was biological.

The target domain was mechanical fastening.

The functional connection involved a simple physical interaction.

Documented Observation

The Smithsonian's account of biomimicry and invention describes how Swiss engineer George de Mestral drew inspiration from plant burrs adhering to clothing and animal fur. The resulting hook-and-loop fastening concept required subsequent engineering development before it became a practical manufactured product.

The distinction matters.

The invention did not require transferring an entire plant into a fastening system.

It required recognizing a useful mechanism: small hooks could engage with loops in another material.

That principle had to be translated into appropriate materials, dimensions, manufacturing techniques and repeatable behavior.

The source supplied inspiration.

The engineering work established a new implementation.

And the usefulness of that implementation depended on conditions different from those in the natural environment.

This example illustrates why cross-domain innovation involves more than recognizing resemblance.

A successful transfer identifies what can be preserved from the source and what must be redesigned for the target.

Even a historically successful analogy does not imply that other biological mechanisms will automatically transfer successfully into engineering.

What exactly moves between domains?

The word transfer can describe several very different things.

A researcher may transfer a question.

An engineer may borrow a design principle.

A team may apply a method under new operating conditions.

A manufacturer may implement a technology in a different industry.

An organization may attempt to reproduce an established capability in a new environment.

These are not equivalent outcomes.

A question can transfer before a solution does

Suppose researchers in two fields encounter systems that depend on multiple interacting components.

One group asks how a capability remains available when its supporting environment changes.

The other group may find the same question useful.

That does not establish that the systems have identical dependencies, behaviors or operating constraints.

The question has transferred.

The answer has not.

A mechanism requires functional correspondence

Two systems may use different terms for the same functional role.

Conversely, they may use identical terms for functions that are not equivalent.

A useful cross-domain investigation should therefore ask what each relevant component or relationship actually does.

Which conditions make the mechanism work?

Which interactions are essential?

Which aspects belong to the source environment rather than the general mechanism?

Without that distinction, researchers may confuse similarity in description with similarity in function.

A method requires target-specific qualification

An evaluation method may work reliably in one context but become inappropriate elsewhere.

Measurements, thresholds, assumptions and failure consequences may differ.

A method does not become valid in a new domain simply because its terminology appears applicable.

A capability requires more than a description

A written explanation of how a system works is not necessarily sufficient to reproduce what the system can do.

The capability may depend on materials, calibration, tacit knowledge, specialized equipment, organizational coordination, training data, permissions or environmental conditions.

Preserving an idea does not automatically preserve the ability to execute it.

That becomes particularly important when comparing AI systems, organizations and physical machines.

Where cross-domain analogies can fail

A misleading analogy can be persuasive precisely because its language appears familiar.

Terms such as autonomy, memory, resilience, adaptation, learning and capability appear in many disciplines.

But their operational meanings can differ greatly.

For example, autonomy in a robotic system may concern the ability to execute specified tasks under technical and environmental constraints.

Human autonomy also involves questions of agency, judgment, responsibility and social conditions.

These uses cannot be treated as equivalent merely because they share a word.

Similarly, memory in a software system is not automatically comparable to human remembering.

And an engineering system that can be restored from a backup has not necessarily demonstrated the regeneration of a human or organizational capability.

A disciplined investigation must preserve these differences.

The purpose of analogy is to reveal a possible line of inquiry, not to erase what makes the target domain distinctive.

A CRE One research example: Human–AI systems and Physical AI

One research direction within CRE One has been to examine whether questions arising from persistent Human–AI systems remain useful when applied to Physical AI.

This is an exploratory cross-domain comparison, not a claim that the architectures are equivalent or that the proposed transfers have been experimentally validated.

The starting question concerns capability under changing conditions.

In a persistent Human–AI working environment, a useful outcome may depend on more than the language model itself.

Relevant conditions can include the human participant, accumulated research context, external evidence, tools, permissions, established working practices and the relationships among them.

If some of these conditions change, the ability to reproduce an earlier result may change as well.

This raises questions such as:

  • Where does the relevant capability actually reside?
  • Which people, tools or environmental conditions are necessary?
  • What evidence demonstrates that the capability remains available?
  • Can the capability be reconstructed after an important dependency changes?
  • What would need to be preserved to make later reproduction possible?

Those questions can then be taken into a very different domain.

Physical AI as a comparison domain

Consider a robotic system performing an industrial assembly task.

Its ability to complete that task may depend on sensors, actuators, software, calibration, component geometry, operating conditions, safety requirements and available human supervision.

The exact dependencies vary by system and application.

Now ask a related question.

If the robot's tools, operating environment, configuration or supporting services change, what evidence would demonstrate that it can still perform the intended task?

The question is structurally related to capability continuity in a Human–AI workflow.

But the answer requires different evidence.

In robotics, physical performance, environmental variation, actuation, reliability and safety may be essential.

In a Human–AI knowledge workflow, the relevant tests may instead concern whether evidence can be retrieved, reasoning can be challenged, decisions remain properly authorized and useful work can still be reproduced.

The domains share a possible investigative question.

They do not thereby share an identical capability model.

Why target-domain evidence is indispensable

Documented Observation

In their 2016 publication Test Methods for Robot Agility in Manufacturing, NIST researchers Anthony J. Downs, William S. Harrison and Craig I. Schlenoff described methods and metrics for assessing industrial robot agility under changing conditions, including evaluation in simulation and real environments.

This establishes an important target-domain consideration.

Robotic agility must be evaluated in relation to the system's task and operating conditions.

A conceptual analogy derived from another discipline cannot substitute for those tests.

Likewise, demonstrating a robotic capability does not automatically establish that the same concept meaningfully describes an individual's judgment or an organization's ability to work with AI.

The purpose of our comparison is to discover which questions remain useful and which assumptions fail when the domain changes.

From resemblance to a testable transfer hypothesis

A cross-domain investigation benefits from distinguishing several questions.

First, what is the actual problem in the target domain?

Second, which source-domain mechanism or method appears relevant?

Third, what would have to remain functionally equivalent for the proposed transfer to make sense?

Fourth, which source-domain conditions are absent or different in the target?

Fifth, what evidence would demonstrate success or failure under the target's own requirements?

These questions can turn a vague resemblance into a researchable proposal.

Working Hypothesis

Cross-domain investigations may become more useful when they explicitly reconstruct the dependencies and validity conditions of a proposed analogy, rather than relying primarily on similarities in vocabulary or surface behavior.

This is a CRE One working hypothesis.

It has not been established as a universal rule of innovation.

Its value would need to be examined through actual investigations, including cases in which an apparently promising analogy fails.

Such failures are informative.

They may reveal hidden dependencies, invalid assumptions or differences that were not visible in the original comparison.

A practical way to investigate cross-domain opportunities

A useful investigation can proceed without pretending that every analogy has the same structure.

1. Define the target problem

What needs to change or improve?

What evidence would make an answer useful?

What are the relevant technical, organizational, social, legal or safety constraints?

2. Identify a candidate source

Which development, method or principle from another domain might address the target problem?

Why was it considered relevant?

Where did the information come from?

3. Reconstruct the source mechanism

What does the source approach actually achieve?

Under what conditions?

Which components, dependencies or relationships make it possible?

What has been demonstrated, and what remains an interpretation?

4. Test the proposed functional correspondence

Which aspects of the mechanism can plausibly operate in the target environment?

Which require adaptation?

Which assumptions do not survive the transition?

Where might identical terminology conceal different functions?

5. Examine target-specific evidence

What must be measured or demonstrated in the target domain?

Are the test conditions representative?

Who has authority to qualify an operational claim?

What would count as failure or insufficient evidence?

6. Preserve unsuccessful and unresolved paths

If a transfer fails, record the reason.

Was the underlying mechanism inappropriate?

Were necessary conditions missing?

Was the comparison based on misleading evidence?

Or is the answer simply unknown?

Preserving these distinctions can prevent a later investigation from repeating the same unsupported assumptions.

This sequence is a practical research synthesis, not a new mandatory standard for all innovation projects.

Cross-domain discovery is not the same as cross-domain validation

Recognizing a relationship can expand the investigation space.

It can reveal questions that would otherwise remain outside the researcher's original field of attention.

But the transition from an interesting relationship to a useful application requires further work.

A source-domain observation must not silently become a target-domain fact.

A design analogy must not silently become a validated implementation.

And a promising hypothesis must not become a capability claim merely because several technologies appear compatible.

This matters especially where the consequences of failure are substantial.

Technical, clinical, industrial, organizational and public-policy decisions may require different forms of evidence and different authorities.

Cross-domain inquiry can help identify possibilities.

It does not automatically authorize action.

Why preserved context can matter

Cross-domain opportunities can emerge long after the original source material was published.

A technology may become relevant because another industry's constraints change.

An established method may gain a new application when compatible tools become available.

A research question may become meaningful in a discipline whose practitioners were not aware of its original formulation.

The source material may have existed for years.

What changed was the surrounding context.

This is one reason why we are interested in observing relationships across sources and time.

An investigation that preserves source provenance, temporal context and structural relationships may make it easier to revisit previously unrelated developments.

But the observational process must remain bounded.

A relationship identified from a limited collection of sources cannot establish what has happened throughout the wider world.

And a structurally recognizable connection does not determine its meaning.

The role of AI in cross-domain innovation

Existing AI systems can already help researchers explore unfamiliar disciplines, compare terminology, search for analogies and generate hypotheses.

Research databases, knowledge graphs, scientific search engines and domain-specialized tools provide additional capabilities.

Therefore, producing a plausible connection between two subjects is not by itself proof of a distinctive new AI capability.

A useful research process must examine the evidence underlying the connection.

The researcher needs to understand which aspects come from documented sources, which are inferred and which remain hypothetical.

An AI system may help formulate candidate transfers.

It should not turn an attractive analogy into a claim of verified applicability without target-specific evidence.

For many problems, the most valuable result may be discovering that an apparent connection does not survive examination.

Rejecting a misleading analogy can be as useful as identifying a promising one.

How Cross-Domain Innovation relates to Technology Intelligence and Horizon Scanning

Our Technology Intelligence research examines how changing technical evidence can justify reconsidering earlier decisions.

Horizon Scanning focuses on emerging developments whose significance remains uncertain.

Cross-Domain Innovation asks a different question.

What happens when an observation, mechanism or method from one field appears relevant to a problem in another?

Technology Intelligence may help establish what has changed.

Horizon Scanning may draw attention to an unfamiliar development.

Cross-Domain Innovation examines whether a meaningful relationship can be investigated across the boundary between fields.

These approaches can reinforce one another.

But none establishes successful transfer without appropriate evidence.

Knowledge valorisation and the creation of value

The practical relevance of cross-domain innovation extends beyond technical invention.

It includes research collaboration, public-sector applications, organizational learning and the creation of social value from existing knowledge.

The Council of the European Union's 2022 Recommendation on knowledge valorisation explicitly addresses connecting knowledge across areas and sectors and turning research results and know-how into outcomes that benefit society.

This is a useful reminder that innovation is not limited to patentable hardware or commercially available software.

A transferred question may lead to a better policy investigation.

A borrowed research method may expose a weakness in an existing evaluation.

A new collaboration may reveal knowledge that neither participant could fully apply alone.

However, transfer and value creation remain separate matters.

The fact that knowledge moved between domains does not establish that the receiving domain benefited.

That benefit still has to be demonstrated.

Where EUCLID and SENION enter the picture

Within CRE One, EUCLID investigates the structural observation of bounded information environments and how observable relationships develop over time.

One research possibility is that such observations could support investigations crossing established disciplinary or application boundaries.

But the existence of a relationship between sources does not establish that a functional mechanism can be transferred.

Nor does it establish that the relationship is useful to a particular person or organization.

SENION is being developed as an AI research extension intended to make selected observational capabilities accessible within compatible existing AI environments.

Product Direction

SENION is being developed to help people and their chosen AI investigate potentially relevant relationships across knowledge domains through inspectable observational context, while preserving the distinction between a proposed connection, a qualified transfer and demonstrated user value.

This describes a product-development direction, not a commercially available or validated automated cross-domain innovation capability.

The relevant value test concerns the complete researcher–AI combination.

Does the extension contribute information or an investigative opportunity that the researcher and their existing AI tools would otherwise have missed?

Can the new connection be examined with sufficient provenance and context?

Does it help formulate a useful next investigation?

And does that improvement justify the additional integration complexity and effort?

Research relevance alone does not establish product value.

The scientific questions remain valuable even if a particular extension does not ultimately implement them.

How would we know a transfer was useful?

A useful cross-domain investigation should not be judged only by the number of analogies it produces.

An attractive idea is not necessarily a practical solution.

Relevant outcomes might include:

  • A newly identified and defensible target-domain hypothesis.
  • A tested mechanism that remains useful under target conditions.
  • An earlier assumption shown to be invalid.
  • A source-domain limitation that rules out an unsuitable transfer.
  • A reproducible application with clearly documented operating boundaries.
  • A better-supported decision about whether further development is justified.

A transfer may also fail or remain unresolved.

The investigation should preserve those outcomes rather than treating every cross-domain connection as a success.

This distinction helps keep the research open to correction.

It also prevents novelty from becoming a substitute for evidence.

The boundary is part of the discovery

Cross-domain innovation invites us to look beyond familiar disciplinary boundaries.

That openness can be productive.

But the boundary itself contains important information.

The conditions under which an idea works in one domain may explain why it cannot simply be reproduced in another.

A difference in terminology may conceal functional similarity.

A similarity in terminology may conceal fundamentally different mechanisms.

And an unexpected disagreement between two domains may expose an assumption that neither would have examined independently.

The most valuable outcome is therefore not always a transferred solution.

Sometimes it is a better understanding of what can and cannot be transferred.

Cross-domain innovation begins with recognizing a connection. It becomes meaningful when we investigate whether the connection survives the change of context.

Continue exploring

If you are investigating a question that crosses industry or disciplinary boundaries, start a conversation.

Research context and primary sources

The following publications provide established scientific, historical and methodological context. They are not presented as evidence that EUCLID or SENION has already implemented or validated the proposed capabilities.