The idea that intelligence cannot be fully understood without reference to the body has a long intellectual history. It challenges the assumption that thinking can be understood entirely as an internal process, detached from physical experience and the environment in which a living being acts. Human beings do not encounter the world as disembodied observers. They perceive through their senses, move through physical space, acquire knowledge through interaction, experience the consequences of their actions and adapt their behaviour in response to changing circumstances. The body is not merely the instrument through which an independently functioning intelligence expresses itself. It is one of the conditions through which human intelligence develops and operates.
The contemporary concept of embodied intelligence draws upon several related traditions, including philosophy, cognitive science, neuroscience, psychology, robotics and artificial intelligence. These traditions do not employ the expression in precisely the same way and there is no single, universally accepted point at which the phrase acquired its present range of meanings. Its intellectual foundations nevertheless reflect a common concern: the relationship between an intelligent agent, its physical constitution and the environment with which it interacts. In biological systems, intelligence is inseparable from the experience of being alive in a body. In artificial systems, researchers have investigated how perception, movement, physical structure and interaction with the environment can contribute to the emergence of intelligent behaviour.¹
This history provides the foundation for EMBODIED INTELLIGENCE®, the artificial intelligence consultancy framework provided by X from GENERAL INTELLIGENCE PLC to personal accident insurance providers. Its purpose is distinct from the scientific study of embodied intelligence in robotics or the philosophical study of embodied cognition. It applies the central insight of those disciplines to a specific institutional question: how can an insurer develop a more sophisticated understanding of human physical risk and how can artificial intelligence strengthen the professional judgement through which that risk is assessed, managed and understood?
The distinction matters. The intellectual history of embodied intelligence belongs to a broad and continuing field of enquiry. The contribution of the consultancy framework lies in examining how that field of thought can inform the strategic, technical and institutional development of personal accident insurance.
From Embodied Cognition to Embodied Intelligence
Modern cognitive science has given particular attention to the relationship between the mind, the body and the surrounding world. A significant contribution came from Francisco Varela, Evan Thompson and Eleanor Rosch in their 1991 work The Embodied Mind: Cognitive Science and Human Experience. The book challenged approaches that treated cognition principally as abstract information processing and developed a perspective in which knowledge and experience must be understood in relation to the embodied organism and its world.² Its importance was not to establish that human beings possess bodies, an obvious fact, but to challenge the consequences of treating those bodies as incidental to cognition.
The wider tradition of embodied cognition has since examined how perception, movement, bodily states, experience and environmental interaction contribute to thinking and learning. Although researchers disagree about the precise scope of embodiment, the general challenge to an exclusively disembodied model of cognition has become influential across several disciplines. It encourages a more integrated account of intelligence in which an agent's physical constitution and practical engagement with the world are relevant to understanding what that agent can perceive, know and do.³
The development of artificial intelligence introduced a related question. Could intelligence be adequately reproduced through the manipulation of representations and symbols, or did it require a more direct relationship between an intelligent system and the world in which it operated? Rodney Brooks's 1991 paper, Intelligence without Representation, became an influential contribution to this debate. Brooks argued for an approach to robotic intelligence founded upon incremental interaction with the real world through perception and action, rather than relying exclusively upon elaborate internal representations of that world.⁴ The significance of this work lay in its insistence that intelligence should be examined through an agent's actual relationship with its environment, not merely through the abstract computation it performs.
Over subsequent decades, the study of embodied intelligence developed across robotics and other areas of artificial intelligence. Researchers examined how a machine's physical construction, sensing capabilities, movement, control systems and environmental interactions influence what it can achieve. In soft robotics, for example, the physical properties of a robot may contribute directly to its capacity to interact adaptively with its surroundings. Intelligence is therefore not solely a function of an algorithm operating independently of the physical system in which it is implemented. It can arise from the interaction between computation, physical form and the environment.⁵
These developments are relevant to EMBODIED INTELLIGENCE® because they establish a powerful organising principle: intelligent behaviour becomes more fully understood when the agent, its capabilities and its environment are considered together. Personal accident insurance presents a particularly compelling application of that principle, since the subject of the insurance is ultimately a human being whose physical vulnerability is inseparable from the activities undertaken and the circumstances encountered.
The Human Body as the Subject of Insurance
Personal accident insurance concerns the consequences of accidental bodily injury. The precise benefits, exclusions, definitions of injury and conditions of cover depend upon the individual policy. The underlying subject, however, is consistent: the possibility that an unforeseen event may cause physical harm to a person and produce consequences for that person's capabilities, responsibilities and life.
This distinguishes personal accident insurance from many forms of insurance whose principal objects of analysis are buildings, vehicles, businesses, financial interests or other forms of property. Those subjects are not without complexity and property losses may ultimately have profound human consequences. Personal accident insurance nevertheless places the physical person directly at the centre of the insured event.
The body is vulnerable to its surroundings in ways that cannot be captured fully by a single occupational classification, activity category or demographic description. Physical exposure depends upon the interaction of many factors: what a person is doing, where the activity takes place, the conditions prevailing at the time, the equipment available, the nature of the hazard and the circumstances of the individual event. The same person may encounter substantially different physical exposures during an ordinary working day, a journey abroad, a sporting activity or a recreational expedition.
Insurance must impose manageable categories upon this complexity. Applications, policy schedules, underwriting rules and claims systems provide necessary structures through which risk can be classified and contractual obligations administered. Yet a classification is an instrument of understanding, not a complete representation of reality. The category cyclist, for example, may encompass activities undertaken under very different physical conditions. An occupational description may fail to reflect the particular tasks a person performs. A geographical classification may not adequately distinguish between the environments in which different activities take place.
EMBODIED INTELLIGENCE® examines how an insurer can improve its understanding of these relationships without abandoning the practical discipline of insurance classification. Its starting point is not the collection of as much information as possible about a person. It is the identification of those relationships between the individual, the activity, the environment and the potential consequence that are material to the insurer's legitimate purpose.
From Accident Records to Intelligence
The conventional insurance record is necessarily retrospective. An accident occurs; the circumstances are reported; a claim is submitted; the insurer considers the relevant policy terms and evidence; and the claim is assessed. These activities remain essential. The opportunity is to connect them to a more continuous process of institutional learning.
An accident is not simply an entry in a claims ledger. It is an event that occurred through a particular sequence of circumstances. It may provide information about an activity, an environmental condition, a mechanism of injury, a weakness in an existing assumption or a pattern of exposure that was not previously recognised. Properly interpreted, the experience of a single event can contribute to a larger understanding of the circumstances in which physical harm arises.
Artificial intelligence can assist in assembling and analysing information that conventional processes may leave fragmented. Claims narratives, policy records, accident descriptions, relevant photographs, geographical information and other legitimately available material may contain complementary evidence. Systems capable of extracting, comparing and organising that evidence can help professionals identify recurring characteristics, anomalous circumstances and relationships that deserve further investigation.
The value lies in moving beyond the simple accumulation of records. A collection of claims becomes an intelligence resource when the insurer can investigate what the claims reveal, test the implications for its existing understanding and apply the resulting knowledge to subsequent decisions.
This is not an argument that every accident can be explained through available data, or that a statistical pattern establishes a causal relationship. Some events remain unusual, evidence may be incomplete and apparently similar incidents may have different underlying causes. The purpose of embodied intelligence is to improve the quality of investigation and understanding, while preserving the distinction between an observed association, a plausible explanation and an established conclusion.
The Person, the Activity and the Environment
The defining analytical question for personal accident insurance is how a person encounters physical risk within a particular environment. Neither the individual nor the environment can be understood adequately in isolation from the activity that brings them into contact.
Occupation, for instance, may provide a useful starting point for understanding exposure, but it can conceal substantial variation in the tasks undertaken by individuals with the same occupational title. Travel may introduce different physical circumstances, but a destination alone does not establish the precise nature of an individual's exposure. Participation in a sport may be associated with identifiable forms of injury, while the nature and intensity of participation, the conditions in which it takes place and the equipment involved may materially influence the circumstances of an accident.
EMBODIED INTELLIGENCE® treats these relationships as potential subjects of structured analysis. Artificial intelligence may help connect relevant information about activities, environments, accident mechanisms and outcomes, allowing the insurer to examine risk in greater detail than broad classifications permit. The objective is not to create an impossibly complete digital representation of every insured person. It is to improve the relevance and interpretation of the evidence used in underwriting, claims management and portfolio oversight.
This approach may reveal that two apparently similar exposures deserve different treatment, or that risks presented under different categories share a common characteristic. In either case, the result remains subject to appropriate professional scrutiny. Greater analytical detail is valuable only when it contributes to a more accurate and defensible understanding of the risk.
The Intelligence of Prevention
An insurer that learns exclusively from claims after they have occurred leaves an important dimension of intelligence underdeveloped. Accident experience can reveal circumstances associated with physical harm and some of those circumstances may be relevant to prevention. A more complete intelligence capability therefore considers not only how accidents are recorded and assessed, but also what can reasonably be learned before comparable events occur.
Artificial intelligence may help identify recurring combinations of activities, locations, environmental conditions and accident mechanisms. It may reveal concentrations that emerge during particular periods, distinguish patterns of frequency from those of severity, or direct attention towards circumstances in which additional investigation could be valuable. Where the evidence supports it, these findings may inform risk communication, product design, underwriting practice or other appropriate preventive measures.
The distinction between prevention intelligence and a promise of prevention is important. Human behaviour and physical circumstances are complex and an insurer cannot eliminate the uncertainty inherent in ordinary life. Nor will every correlation identify an actionable cause. A responsible system must therefore distinguish between risks that can be reduced, exposures that can be better understood and uncertainties that must remain part of the insurance proposition.
The aim is not to impose an abstract ideal of risk-free living. Work, sport, travel and recreation have purposes that may justify the physical risks involved. The purpose of insurance intelligence is to understand those risks more adequately, make better-informed decisions and, where feasible, contribute to the reduction of avoidable harm.
Injury as an Interruption of Capability
The significance of bodily injury cannot always be understood fully through its medical description alone. An injury occurs to a particular person, whose physical capabilities, responsibilities and circumstances determine much of its practical consequence.
The same anatomical injury may have different implications for people who undertake different activities. An injury affecting the hand, for example, may have different functional consequences for someone whose occupation depends upon fine manual skills, someone whose work involves heavy physical activity and someone whose principal responsibilities are different. The clinical facts remain important, but they do not necessarily describe the entirety of the practical consequences.
For personal accident insurers, the relevant contractual question remains the one defined by the policy. A consultancy framework cannot substitute general judgements about human capability for the policy's insured events, benefit definitions or evidential requirements. It can, however, examine whether the information available to the insurer provides a sufficiently informed understanding of the relevant circumstances when those circumstances are material to the decision.
EMBODIED INTELLIGENCE® therefore distinguishes the physical event, the injury sustained and the consequences experienced. These are connected, but they are not interchangeable. Preserving that distinction can improve the quality of claims analysis, help explain patterns in outcomes and prevent an overly narrow interpretation of what a claims record represents.
The underlying principle is that a person is not merely the bearer of an injury code. The injury is an event within a human life and the insurer's understanding of that event should be as rigorous and contextually appropriate as the terms of the contract permit.
Claims Intelligence and the Sequence of Recovery
Claims management offers another important application of embodied intelligence because information about an injury may emerge over time. The initial notification may establish the reported accident and immediate circumstances, while subsequent evidence clarifies the nature of the injury, its consequences and the development of the claim. The information available at the beginning is not necessarily identical to that available when the claim is concluded.
Artificial intelligence can assist in organising this material, identifying relevant information, comparing cases and directing professional attention towards claims whose complexity or circumstances warrant closer examination. It may help reveal patterns across outcomes and highlight differences between cases that initially appear similar. The objective is to make the evidence more accessible and the claims process more intelligently informed.
Recovery can also be a relevant subject of institutional learning where the insurer's policy, role and lawful access to information make it appropriate. An insurer may be able to examine aggregate patterns in the duration of incapacity, the development of claims or other recorded outcomes. Such analysis can improve understanding of the portfolio, although any conclusions must reflect the limits of the available data and must not be presented as medical predictions beyond the evidence.
The individual claimant must remain distinct from the statistical population. A pattern among previous claims does not determine the course of a particular person's recovery. Nor does a predicted outcome relieve the insurer of the obligation to assess the individual claim in accordance with the applicable terms and evidence.
EMBODIED INTELLIGENCE® consequently treats claims intelligence as a means of strengthening professional assessment rather than automating every decision. Its objective is to connect evidence across the life of a claim, improve the quality of investigation and enable the institution to learn from experience without mistaking a statistical generalisation for an individual fact.
Underwriting and the Physical Reality of Risk
Underwriting requires an insurer to make decisions about exposure under conditions of incomplete knowledge. The quality of those decisions depends upon the information available, the validity of the assumptions employed and the relevance of the classification to the risk under consideration.
Artificial intelligence can support that work by bringing together historical experience, relevant activity information, environmental data and other evidence that can appropriately inform assessment. It may help identify recurring characteristics, test assumptions against observed outcomes and highlight changes in the portfolio that merit investigation. Where data quality and analytical design permit, it can also assist in examining more detailed relationships than broad categories alone would reveal.
The application must remain proportionate to the class of business and the available evidence. A model developed for one population or type of activity may not transfer reliably to another. Historical claims may reflect changes in policy terms, reporting practices, underwriting strategy or the composition of the insured population. Apparent improvement in predictive performance must therefore be distinguished from genuine improvement in underwriting judgement.
The insurer should also recognise the limits of individualised precision. More granular data do not invariably provide a more reliable assessment and a highly detailed model can create an appearance of certainty unsupported by its evidence. The appropriate objective is not to classify every person as precisely as possible. It is to establish which distinctions are materially useful, which can be justified and which should not be made.
EMBODIED INTELLIGENCE® approaches underwriting as the disciplined interpretation of physical risk. Artificial intelligence can expand the evidence available to the underwriter; it cannot remove the need to evaluate uncertainty, determine the relevance of information and remain accountable for the decision.
The Intelligence of the Portfolio
The insurer's experience of physical risk is distributed across individual policies, activities, locations, claims and periods of time. The aggregate portfolio may reveal patterns that are difficult to detect from isolated cases.
A group of claims may share an environmental characteristic, a recurring mechanism of injury or an underlying feature of an activity. Alternatively, apparent concentrations may be the consequence of changes in reporting, policy distribution or the mix of insured exposures rather than a genuine change in risk. The ability to distinguish between these possibilities is central to sound portfolio analysis.
Artificial intelligence can assist in connecting information across these dimensions, highlighting unusual concentrations and supporting more detailed examination of portfolio developments. It can help insurers consider whether an apparent increase in claims reflects greater exposure, a change in frequency, a change in severity or an alteration in the composition of the business. The distinction can be commercially significant because different explanations may require different responses.
The value is not that a machine can determine every causal relationship. It is that the institution can examine a larger body of evidence, identify questions that deserve attention and make its interpretation more systematic. The portfolio becomes a source of continuing intelligence rather than merely a financial result reviewed after the relevant period has ended.
For personal accident insurers, this also creates an opportunity to connect underwriting with claims experience. Each side of the business can inform the other, provided the relationship is designed carefully and material differences in coverage, exposure and reporting are properly accounted for.
Data, the Body and the Responsibility of Trust
The prospect of analysing embodied risk can encourage the assumption that every additional source of personal information is potentially useful. EMBODIED INTELLIGENCE® rejects that assumption. The availability of information does not establish its relevance, its lawfulness or the legitimacy of using it for a particular insurance purpose.
Information concerning a person's health, physical condition or activity may be sensitive. Wearable devices and connected technologies may generate increasingly detailed records of movement or physiological measures, but the existence of such records does not establish that an insurer should collect them, that the individual has meaningfully agreed to their use or that they improve a particular underwriting or claims decision.
The appropriate questions concern purpose, necessity, proportionality, reliability and governance. Is the information materially relevant to the intended use? Is there a lawful basis for processing it? Are any additional conditions required because it constitutes special-category personal data? Can the purpose be achieved through less intrusive information? Is the inference sufficiently reliable to support the proposed use? How will the individual be informed and what safeguards apply if the system contributes to a consequential decision?
The Information Commissioner's Office has emphasised that the use of artificial intelligence involving personal data must be assessed against applicable data protection requirements, including where systems infer information about health or other protected categories. Solely automated decisions involving special-category data and producing legal or similarly significant effects are subject to particular restrictions and safeguards.⁶ The precise requirements depend on the processing and its circumstances; the framework should not be treated as a substitute for legal assessment.
This is more than a compliance consideration. An insurer that treats the body simply as a source of data risks confusing surveillance with intelligence. The purpose of EMBODIED INTELLIGENCE® is to strengthen understanding of physical risk while respecting the distinction between what technology can observe and what an insurer has a proper reason to know.
The Human Professional and the Intelligent Institution
The practical success of the framework depends upon the relationship between artificial intelligence and professional expertise. An insurer may possess sophisticated analytical systems and still make poor decisions if its professionals cannot understand their limitations, challenge their recommendations or translate their findings into appropriate action.
Human judgement remains necessary because insurance decisions combine evidence, contractual interpretation, uncertainty and responsibility. An apparent anomaly may have an innocent explanation; a statistical association may lack practical significance; and an unusually high-risk case may require more careful assessment rather than automatic rejection. These distinctions cannot be delegated responsibly to a system merely because it is computationally capable.
Artificial intelligence can nevertheless make a substantial contribution. It can reduce the effort required to retrieve information, organise evidence from different sources, identify patterns and direct attention towards issues that deserve investigation. It can help professionals see relationships that might otherwise remain obscured by the volume and distribution of records.
The objective is a better-informed professional operating within a better-designed institution. Underwriters should have access to relevant evidence. Claims specialists should be able to establish the material facts of an incident. Investigators should receive useful signals without being instructed to mistake suspicion for proof. Management should understand how emerging patterns affect the portfolio. Across these activities, intelligence should augment the quality of human decisions rather than disguise the accountability for them.
Institutional Memory and Continuous Learning
An insurer's accumulated claims experience is an important institutional asset, but experience does not automatically become knowledge. Records can remain fragmented, historical decisions may lack sufficient explanation and lessons acquired by one team may not reach another. Organisational memory is particularly vulnerable where technical expertise resides in individuals rather than in accessible, well-governed processes.
EMBODIED INTELLIGENCE® considers how artificial intelligence can help preserve and connect relevant institutional knowledge. Historical claims, accident mechanisms, underwriting decisions, investigation outcomes and documented lessons may be made easier to retrieve and compare. This can allow the insurer to revisit previous assumptions, understand the context of earlier outcomes and recognise when apparently new problems resemble those encountered before.
A useful institutional memory must remain open to correction. An earlier decision is not necessarily sound because it was made by an experienced professional and a historical pattern should not be treated as a permanent rule when the underlying conditions have changed. The purpose of memory is not to preserve past judgements from scrutiny; it is to make those judgements available for examination in light of subsequent evidence.
The result sought is an insurer that becomes more capable through experience. Each claim should have the potential to contribute to organisational learning, while appropriate protections ensure that information is accurate, relevant, responsibly retained and used for a legitimate purpose.
Governance as a Condition of Intelligence
The more directly an intelligent system affects individuals, the greater the importance of clear responsibility for its use. Personal accident insurance involves decisions that can have material consequences for applicants and claimants. A system supporting those decisions must therefore be subject to appropriate validation, monitoring, access controls, documentation and human oversight.
Governance should address the complete life of the system, from the original purpose and quality of its data through design, testing, deployment and ongoing evaluation. It should establish which decisions the system may support, the circumstances in which it should not be relied upon and the process for investigating unexpected results. Where a system changes over time, the insurer must understand whether its performance remains suitable for the task.
The Financial Conduct Authority's published approach to artificial intelligence in financial services has emphasised responsible innovation within existing regulatory frameworks, including established requirements concerning governance, controls and consumer outcomes.⁷ The institutional implication is straightforward: the adoption of AI does not displace the responsibilities of the regulated firm or those accountable for its decisions.
EMBODIED INTELLIGENCE® therefore treats governance as part of the intelligence architecture. A system that produces results which cannot be examined, challenged or responsibly acted upon is not fully intelligent in the institutional sense. Its technical capability may be considerable, but its contribution to a trustworthy insurance operation remains incomplete.
Research into Embodied Intelligence
The research programme relevant to EMBODIED INTELLIGENCE® extends beyond any single class of insurance or generation of technology. It encompasses the study of human intelligence, embodied cognition, artificial intelligence, machine perception, biomechanics, accident mechanisms, behavioural science, environmental analysis and the ways in which institutions learn from experience.
Robotics and embodied AI provide insight into the interaction between an intelligent agent and its physical environment. Cognitive science examines the relationship between perception, action and human experience. Accident research investigates the circumstances in which injury occurs, while insurance supplies a distinctive body of evidence about insured events, contractual responses and financial consequences. Each discipline contributes a different perspective and none alone provides a complete account of personal accident risk.
The task is to establish which insights can be applied responsibly to the insurer's particular decisions. A technique that proves valuable in robotic control does not automatically translate into an effective underwriting tool. A pattern identified in an accident dataset may not be suitable for individual risk assessment. Research must therefore be followed by evaluation, validation and practical testing, with the limits of each application made explicit.
EMBODIED INTELLIGENCE® forms part of the wider research undertaken by X from GENERAL INTELLIGENCE PLC into human intelligence, artificial intelligence, machine intelligence, collective intelligence and institutional intelligence. Its specific contribution is to examine the relationship between those fields and the understanding of human physical risk in personal accident insurance.
The EMBODIED INTELLIGENCE® Test
The framework can be assessed against a series of institutional questions.
- Does the proposed capability improve the understanding of the insured exposure?
- Does it distinguish the person, activity and environment without confusing one for another?
- Does it help explain how accidents occur, or merely identify correlations?
- Does it improve the quality of underwriting evidence?
- Does it connect claims experience with the questions that matter to the portfolio?
- Does it help professionals direct their attention towards material issues?
- Does it distinguish an anomaly from evidence of wrongdoing?
- Does it preserve the difference between a statistical prediction and an individual outcome?
The test must also consider the quality of the institution surrounding the technology.
- Are the data reliable and relevant?
- Are the system's limitations understood?
- Can its recommendations be challenged?
- Is there an appropriate basis for processing the information?
- Are privacy, security and fairness considered from the outset?
- Are material decisions traceable to responsible professionals?
- Does the organisation learn from results and revise its assumptions when the evidence warrants it?
- Finally, does the capability improve a decision that matters?
If a system produces more information without changing the quality of underwriting, claims management or institutional understanding, its contribution may be limited. If it increases complexity, intrusiveness or the risk of error without a proportionate benefit, its adoption requires reconsideration.
The purpose is not to measure intelligence by the volume of data collected or the sophistication of the model. It is to determine whether the insurer has become more capable of understanding physical risk and acting responsibly upon that understanding.
The Objective
The intellectual history of embodied intelligence suggests that the relationship between intelligence, physical existence and the environment cannot be understood adequately by treating each as an isolated subject. In human beings, the body is the medium through which the world is experienced, action is undertaken and physical consequence is encountered. In artificial systems, researchers continue to investigate how embodiment can shape perception, adaptation and intelligent behaviour. Personal accident insurance offers a distinct institutional application of this wider insight because its fundamental subject is the human being exposed to the possibility of bodily harm.
EMBODIED INTELLIGENCE® applies that principle to the business of insurance. It seeks to connect the information available about people, activities, environments, accident mechanisms, injuries and claims outcomes in a manner that improves professional understanding. It recognises that artificial intelligence can expand analytical capacity, while insisting that evidence must be interpreted, uncertainty acknowledged and consequential decisions governed appropriately.
The objective is neither the comprehensive digitisation of the human body nor the elimination of uncertainty from human life. It is the development of an insurer better able to understand the physical risks it accepts, the circumstances in which accidents occur and the implications of the claims experience it accumulates. Its underwriting should be more informed, its claims intelligence more coherent, its institutional memory more useful and its decisions more firmly grounded in relevant evidence.
For GENERAL INTELLIGENCE PLC, the framework provides a specific application of the broader study of intelligence to personal accident insurance providers. It draws upon the intellectual tradition of embodiment without claiming to originate that tradition and it places artificial intelligence within a wider institutional purpose. The measure of success is not the extent to which a machine can describe a human body, but the extent to which the organisation can understand physical risk more intelligently while respecting the person at its centre.