Fact-checked by Grok 2 weeks ago

Model

A model is a simplified representation of a real-world system, object, process, or phenomenon, constructed to facilitate understanding, prediction, and analysis of complex realities that are difficult to observe or manipulate directly.^[1]^[2] In scientific and mathematical contexts, models serve as tools for approximating target systems, enabling researchers to test hypotheses, simulate behaviors, and communicate ideas without relying solely on empirical data.^[3]^[4] Scientific models can be categorized into several primary types, each suited to different investigative needs. Physical models are tangible replicas, such as scale models of bridges in engineering or anatomical models in biology, which allow direct manipulation to study structural integrity or functional dynamics.^[5]^[6] Conceptual models provide abstract frameworks, often in the form of diagrams, flowcharts, or verbal descriptions, to illustrate relationships and mechanisms, like the water cycle diagram representing hydrological processes.^[7]^[8] Mathematical models, including equations and simulations, quantify variables and predict outcomes, as seen in the Lotka-Volterra equations modeling predator-prey interactions in ecology.^[2] Hybrid approaches, combining these elements—such as computational simulations integrating physical and mathematical components—are increasingly common in fields like climate science and engineering.^[6]^[9] The development and use of models follow a structured process integral to the scientific method: observing phenomena, formulating hypotheses, deriving predictions, and refining the model through iterative testing against data.^[9]^[10] Models are essential for advancing knowledge because they bridge the gap between theory and observation, allowing scientists to explore inaccessible scenarios, such as atomic structures or planetary formations, while highlighting limitations like assumptions or simplifications that require validation.^[11]^[12] In engineering and applied sciences, models optimize designs, assess risks, and inform policy, underscoring their role in innovation and problem-solving across disciplines.^[6]^[13]

Definition and Fundamentals

Core Definition

A model is a simplified representation of a system, process, or phenomenon that captures its essential features while omitting less relevant details to aid in understanding, analysis, or prediction.^[14] This representation shares important characteristics with its real-world counterpart, enabling systematic description and inference about the target.^[14] Models can take tangible forms, such as physical replicas, or intangible forms, like mathematical equations or conceptual diagrams, depending on the context of use.^[15] Central to models are the principles of abstraction and simplification, which involve selectively focusing on key elements to make complex realities more manageable.^[15] Their design is inherently purpose-driven, tailored to specific goals such as visualization or forecasting, rather than aiming for complete fidelity to the original system.^[16] Unlike prototypes, which serve as functional, often full-scale tests of a design's operability in engineering or product development, models emphasize representational accuracy over practical implementation.^[17] Models differ from theories in that the latter provide broad explanatory frameworks grounded in a system of assumptions about underlying principles, whereas models function as practical tools for application, often derived from or testing those theories.^[18] Theories aim to explain why phenomena occur across wide contexts, while models focus on moderate-scope predictions or descriptions.^[19] Common purposes of models include representation for visualization, as in diagrams illustrating atomic structures; prediction for forecasting outcomes, such as climate projections; and explanation for elucidating causal relationships, like epidemiological simulations of disease spread.^[8]^[20]

Historical Development

The concept of modeling has ancient origins, with evidence of scale models employed in architectural planning during the construction of the Egyptian pyramids around 2600 BCE. These physical representations allowed builders to test structural designs and proportions before full-scale implementation, demonstrating an early recognition of models as tools for prediction and verification in engineering.^[21] In astronomy, the Ptolemaic geocentric model, developed by Claudius Ptolemy in the 2nd century CE, represented a sophisticated mathematical framework positing Earth at the universe's center, with celestial bodies moving in epicycles to account for observed planetary motions. This model dominated astronomical thought for over a millennium, illustrating the shift toward abstract representations of natural phenomena.^[22] The Scientific Revolution in the 17th century marked a pivotal advancement in modeling, exemplified by Galileo's inclined plane experiments, which served as controlled physical models to investigate motion and acceleration under gravity. By rolling balls down inclines of varying angles, Galileo diluted gravitational effects to measurable speeds, enabling precise timing and leading to the formulation of uniform acceleration laws that challenged Aristotelian physics.^[23] Building on this empirical foundation, Isaac Newton's gravitational model in the late 17th century provided a universal mathematical description of attraction between masses, as articulated in his Philosophiæ Naturalis Principia Mathematica (1687), establishing gravity as a predictive force law integral to classical physics./09%3A_Gravity/9.02%3A_Newtons_Universal_Theory_of_Gravity) In the 19th and early 20th centuries, statistical modeling emerged as a cornerstone of quantitative analysis, pioneered by Karl Pearson's introduction of correlation coefficients and the chi-square test around 1900, which enabled the assessment of relationships in biological and social data through probabilistic frameworks. Ronald Fisher extended these foundations in the 1920s with methods like analysis of variance and maximum likelihood estimation, formalizing experimental design and inference in agricultural and genetic studies, thus solidifying statistics as a modeling discipline for handling uncertainty.^[24] Post-World War II, the advent of computational modeling accelerated with the completion of ENIAC in 1945, the first general-purpose electronic digital computer, which facilitated complex simulations in ballistics and weather prediction, transitioning models from manual calculations to automated processes.^[25] Key theoretical milestones further refined modeling paradigms, including Richard B. Braithwaite's 1953 publication Scientific Explanation: A Study of the Function of Theory, Probability and Law in Science, which formalized the role of models in deductive-nomological explanations, emphasizing their representational and explanatory power in scientific inference.^[26] In the 1970s, UNESCO recognized the interdisciplinary importance of modeling within systems science, supporting global initiatives that integrated dynamic simulations for addressing complex socio-economic and environmental challenges.^[27] Contemporary developments since the 2010s have integrated artificial intelligence and machine learning into modeling, particularly through neural network-based architectures for predictive tasks, enabling data-driven approximations of nonlinear systems in fields like climate forecasting and healthcare diagnostics.^[28] These advancements build on earlier computational foundations, allowing models to learn patterns from vast datasets and adapt dynamically, marking a shift toward interdisciplinary applications that blend empirical observation with algorithmic sophistication.

Classification of Models

Physical Models

Physical models are tangible, three-dimensional representations of real-world objects or systems, constructed from physical materials to replicate key attributes such as shape, proportions, and sometimes material properties.^[6] These scale models, often reduced in size while maintaining geometric similarity, serve as concrete analogs for testing and analysis in fields like engineering and architecture, allowing direct observation of physical behaviors that might be impractical or unsafe to study at full scale.^[29] For instance, models of buildings or vehicles enable designers to assess spatial relationships and structural forms without the constraints of actual construction.^[30] Construction of physical models typically involves selecting appropriate materials to achieve durability, precision, and realism, with common choices including wood (such as balsa for lightweight frames), plastics for molded components, and 3D-printed polymers for complex geometries.^[31] Techniques range from traditional handcrafting and carving to modern additive manufacturing, where 3D printing facilitates rapid prototyping of intricate designs from digital files.^[32] A fundamental aspect is adherence to scaling laws, which ensure proportional fidelity: if a linear scale factor k is applied (where k < 1 for reduced models), linear dimensions scale by k, surface areas by k^2, and volumes by k^3, influencing properties like weight and fluid interactions.^[33] In applications, physical models are widely used for aerodynamic testing in wind tunnels, where scaled aircraft or vehicle replicas simulate airflow to evaluate lift, drag, and stability under controlled conditions.^[34] Similarly, bridge prototypes assess structural integrity by subjecting them to simulated loads and environmental forces, revealing potential failure points before full-scale implementation.^[35] The primary advantages of physical models lie in their intuitive visualization, enabling stakeholders to interact hands-on with a tangible prototype for better spatial comprehension and iterative experimentation.^[36] However, limitations include challenges in scalability, as very large systems become prohibitively expensive and time-intensive to replicate accurately at reduced scales, often requiring compromises in material similarity or testing fidelity.^[29] A notable historical example is the Wright brothers' glider models in the early 1900s, which they constructed and tested at Kitty Hawk to refine wing designs and control mechanisms for powered flight, conducting over 1,000 glide trials to validate aerodynamic principles.^[37]

Mathematical and Computational Models

Mathematical models represent systems through symbolic equations that capture relationships among variables, enabling quantitative predictions and analysis. These models often employ differential equations to describe dynamic processes, such as Newton's second law of motion, formulated as F = ma, where F is force, m is mass, and a is acceleration, originally derived in the context of classical mechanics.^[38] Subtypes include continuous models using ordinary or partial differential equations for time- or space-dependent phenomena, and discrete models based on difference equations for sequential processes. Computational models extend these by implementing equations algorithmically on digital systems, such as finite element analysis (FEA), which discretizes complex structures into meshes to solve partial differential equations numerically for stress and deformation simulations; FEA originated in the 1940s with early structural engineering applications and was formalized in the 1960s.^[39] Key components of mathematical and computational models include variables (measurable quantities like position or temperature), parameters (fixed constants such as gravitational acceleration), and functions (relations defining interactions, e.g., force as a function of displacement). Models are classified as deterministic, yielding unique outputs from given inputs via fixed rules, or stochastic, incorporating randomness to account for uncertainty, as in Monte Carlo simulations that use repeated random sampling to estimate probabilistic outcomes in risk assessment or particle physics.^[40]^[41] The development process begins with formulation, where assumptions are defined and equations are derived to represent the system, followed by solution via analytical methods (exact closed-form expressions) or numerical techniques like Euler's method for ordinary differential equations (ODEs), which approximates solutions iteratively using y_{n+1} = y_n + h f(t_n, y_n), where h is the step size; this method was proposed by Leonhard Euler in the 18th century.^[42] Iteration refines the model by comparing outputs to real data, adjusting parameters, and reducing assumptions to improve accuracy.^[43] A representative example is the SIR model in epidemiology, which divides a population into susceptible (S), infected (I), and recovered (R) compartments, governed by the ODEs:

\begin{align} \frac{dS}{dt} &= -\frac{\beta S I}{N}, \\ \frac{dI}{dt} &= \frac{\beta S I}{N} - \gamma I, \\ \frac{dR}{dt} &= \gamma I, \end{align}

where \beta is the transmission rate, \gamma is the recovery rate, and N is total population; this compartmental framework was introduced by Kermack and McKendrick in 1927. Climate models, such as general circulation models (GCMs), rely on partial differential equations like the Navier-Stokes equations for fluid dynamics and energy balance equations to simulate atmospheric and oceanic interactions over global scales. The evolution of these models traces from analog computers in the 1940s, which solved differential equations mechanically for engineering problems like ballistic trajectories, to modern computational frameworks incorporating deep learning since the 2010s, where neural networks optimize parameters in large-scale simulations for pattern recognition in complex data.^[44]

Conceptual and Abstract Models

Conceptual and abstract models represent complex systems through non-physical, idea-based frameworks that emphasize qualitative relationships, processes, and structures without relying on numerical quantification. These models serve as mental or diagrammatic tools to simplify and communicate understanding of phenomena, capturing essential elements like interactions and hierarchies in an accessible form. Unlike physical replicas, they prioritize abstraction to facilitate reasoning about intangible aspects, such as decision-making flows or entity classifications, enabling users to grasp overarching patterns without delving into measurable details.^[45] Key types of conceptual models include verbal models, which use narrative descriptions to outline processes and relationships; graphical models, such as flowcharts, mind maps, and Unified Modeling Language (UML) diagrams employed in software design to visualize system architectures; and ontological models, which categorize entities, attributes, and relations within a domain to define foundational concepts. Verbal models rely on textual explanations for broad overviews, while graphical variants leverage diagrams for intuitive representation of dynamics, and ontological approaches formalize knowledge structures for consistency across applications. These types emerged from qualitative reasoning traditions, with tools like system dynamics diagrams—pioneered by Jay Forrester in the 1950s through stock-flow representations—providing early methods to depict accumulations and rates of change abstractly.^[6]^[46]^[47]^[48] Prominent examples illustrate their utility: Michael Porter's Five Forces model, introduced in 1979, conceptually frames competitive industry dynamics through threats from entrants, supplier and buyer power, substitutes, and rivalry to guide business strategy. In ecology, food web models depict species interactions as networked trophic relationships, abstractly illustrating energy flows and dependencies without quantitative simulations. Such models excel in fostering interdisciplinary communication by distilling complexity into relatable forms, promoting shared understanding among diverse stakeholders. However, they lack the precision of mathematical models, which can extend them quantitatively, potentially leading to oversimplifications that overlook nuanced variations.^[49]^[45]

Theoretical Foundations

Key Properties

In general model theory, models exhibit core properties that define their utility and structure, including fidelity, generality, and parsimony. Fidelity refers to the degree of realism or representational accuracy with which a model captures the essential features of its target system, ensuring that the model's outputs align closely with observed phenomena without unnecessary distortion. This property is central to effective modeling, as higher fidelity enhances the model's reliability for predictive or explanatory purposes, though it often involves tradeoffs with other attributes. Generality denotes the breadth of applicability of a model across diverse contexts or systems, allowing it to address a wide range of scenarios beyond its initial calibration. Parsimony, meanwhile, emphasizes simplicity in model construction, achieving explanatory power with the minimal number of assumptions or parameters necessary, thereby avoiding overcomplication that could obscure insights or reduce interpretability. These properties, as articulated in foundational works on scientific modeling, balance descriptive power against cognitive and computational demands.^[50]^[51]^[52] Scalability and modularity further characterize robust models, enabling them to adapt to increasing complexity or integration with other frameworks. Scalability allows a model to handle larger datasets, more variables, or extended time horizons without proportional loss in performance, often through techniques like hierarchical modeling that layer abstractions progressively. Modularity supports the decomposition of a model into independent components that can be developed, tested, or replaced separately, facilitating combination with complementary models for comprehensive analyses. For instance, in systems engineering, hierarchical modeling structures promote both scalability and modularity by organizing elements into nested levels, from fine-grained details to overarching behaviors. These attributes ensure models remain flexible and extensible in dynamic applications.^[53]^[54] Effective models also incorporate mechanisms for handling uncertainty, such as error bounds and sensitivity analysis, to quantify and mitigate the impact of input variability or structural approximations. Error bounds provide probabilistic or interval estimates around model predictions, delineating the range within which outcomes are likely to fall given known uncertainties. Sensitivity analysis evaluates how changes in model parameters or assumptions propagate to outputs, identifying critical dependencies and informing refinement priorities. These techniques are integral to maintaining model reliability, particularly in fields like environmental simulation where data incompleteness is common.^[55]^[56] Philosophically, these properties draw from insights in bounded rationality and abstraction levels. Herbert Simon's concept of satisficing, introduced in the context of decision-making under constraints, underscores that models need not achieve perfect fidelity or maximal generality but rather suffice for practical needs given limited information and resources.^[57] Complementing this, Herbert Stachowiak's general model theory posits models as operating across levels of abstraction, where reduction to key properties enables pragmatic representation without exhaustive detail.^[58] An illustrative metric for assessing such properties is the coefficient of determination, R^2, which measures goodness-of-fit in regression models by indicating the proportion of variance in the target variable explained by the model, typically ranging from 0 to 1; values closer to 1 signal high fidelity and parsimony in capturing relevant patterns.

Model Validation and Verification

Model verification and validation are essential processes to ensure that models are both correctly implemented and reliably representative of the systems they simulate. Verification focuses on confirming the internal consistency and correctness of the model, such as ensuring that computational code accurately solves the underlying mathematical equations without errors, often through methods like code debugging and numerical accuracy checks. Validation, in contrast, evaluates the model's alignment with real-world data by comparing outputs to empirical observations, thereby assessing its predictive accuracy for intended applications. Several techniques are employed to perform validation effectively. Cross-validation, particularly the k-fold method, partitions the dataset into k equally sized subsets (folds), trains the model on k-1 folds, and tests it on the remaining fold, repeating this process k times to provide a robust estimate of generalization performance while minimizing bias from a single data split. Sensitivity analysis tests model robustness by systematically varying input parameters or assumptions and observing the impact on outputs, helping identify critical dependencies and potential instabilities.^[59] For probabilistic models, Bayesian updating refines parameter estimates by incorporating new data to update prior distributions, enabling ongoing validation through posterior predictive checks that assess how well the model fits observed evidence.^[60] Standards and statistical tests provide structured frameworks for these processes. The ASME V&V 10-2006 guide outlines procedures for verification and validation in computational solid mechanics, emphasizing quantification of numerical errors and comparison of model predictions to experimental data to infer accuracy levels.^[61] Similarly, ASME V&V 20-2009 extends these principles to computational fluid dynamics and heat transfer, specifying methods to estimate modeling uncertainties.^[62] Statistical tests, such as the chi-squared goodness-of-fit test, measure discrepancies between observed and expected frequencies under the model, with low p-values indicating poor fit and the need for refinement.^[63] Challenges in validation include overfitting, where models capture noise rather than underlying patterns in training data, leading to poor generalization; this is commonly addressed through regularization techniques that add a penalty term to the loss function, such as L1 or L2 norms, to constrain model complexity and promote simpler, more robust solutions.^[64] In AI models, black-box issues arise from opaque internal mechanisms, making it difficult to trace decision pathways and verify causal relationships, which complicates validation and requires supplementary interpretability methods like feature attribution.^[65] A prominent case study involves the validation of global climate models, which have been rigorously tested against observational data compiled in IPCC assessments since the 1990 First Assessment Report. These models, such as those from the Coupled Model Intercomparison Project, demonstrate skill in reproducing historical temperature trends from 1990 onward, with projections aligning closely to observed global surface warming when accounting for external forcings like greenhouse gases, as evaluated in subsequent IPCC reports.^[66]^[67]

Applications and Uses

In Scientific Research

In scientific research, models play a pivotal role in hypothesis testing by enabling the simulation of complex experiments that are often impractical or impossible to conduct directly, allowing researchers to predict outcomes and refine theories based on theoretical frameworks. For instance, in particle physics, quantum models governed by the Schrödinger equation simulate the behavior of subatomic particles, facilitating the testing of hypotheses about quantum states and interactions without requiring every physical trial. This approach has been instrumental in advancing understanding of fundamental forces and particle properties.^[68] Mathematical models in biology, such as the Hardy-Weinberg equilibrium, provide a foundational tool for testing hypotheses about genetic stability in populations under idealized conditions of no selection, mutation, migration, or drift. Formulated in 1908, the principle states that for a gene with two alleles at frequencies p and q, the genotype frequencies remain p^2 + 2pq + q^2 = 1 across generations, serving as a null model to detect evolutionary forces when deviations occur. In cosmology, the Lambda-CDM model, established following 1998 observations of type Ia supernovae indicating accelerated expansion, integrates cold dark matter and a cosmological constant to test hypotheses about the universe's composition and evolution, predicting cosmic microwave background fluctuations with high precision. The integration of models with large-scale data has further amplified their utility, as seen in genomics following the Human Genome Project's completion in 2003, which generated reference sequences enabling predictive models for gene function, variant effects, and disease associations through statistical and computational frameworks. These models process vast datasets to test hypotheses about genetic mechanisms, accelerating insights into complex traits. A landmark impact is evident in protein structure prediction, where AlphaFold, unveiled in 2020, achieved near-experimental accuracy for previously unsolved folds, revolutionizing hypothesis testing in structural biology by simulating folding pathways and enabling rapid exploration of protein interactions. Subsequent advancements, such as AlphaFold 3 released in 2024, have extended these capabilities to model interactions between proteins, DNA, RNA, and ligands with improved accuracy.^[69]^[70]^[71] Interdisciplinary hybrid models, combining physics and chemistry, have advanced material science by simulating atomic-scale behaviors to test hypotheses about novel properties, such as in quantum mechanical/molecular mechanical approaches that model reactions in heterogeneous environments with enhanced accuracy over purely empirical methods. These integrations not only expedite discoveries but also bridge theoretical predictions with empirical validation across natural sciences.^[72]

In Engineering and Design

In engineering and design, models play a pivotal role in iterative processes that facilitate rapid prototyping and refinement of built systems. Engineers employ computational models, such as computer-aided design (CAD) models, to simulate and test virtual prototypes throughout the design cycle, allowing for quick iterations without the immediate need for physical construction. This approach enables mechanical engineers to evaluate design variations, identify potential flaws early, and optimize performance before committing to manufacturing, thereby streamlining the transition from concept to production.^[73]^[74] Key examples of such models include finite element analysis (FEA) for stress assessment and proportional-integral-derivative (PID) controllers for system regulation. FEA divides complex structures into finite elements to approximate stress distributions under applied loads, solving the governing equations of elasticity through numerical methods like the stiffness matrix formulation [K]\{u\} = \{F\}, where [K] is the global stiffness matrix, \{u\} represents nodal displacements, and \{F\} denotes applied forces; stresses are then derived from strains via the constitutive relation \{\sigma\} = [D]\{\epsilon\}, with [D] as the material stiffness matrix and \{\epsilon\} as strains. This technique is widely used in structural design to predict failure points and ensure integrity under operational conditions.^[75]^[76] In control systems, PID models maintain stability by computing control outputs based on error signals, given by the equation

u(t) = K_p e(t) + K_i \int_0^t e(\tau) \, d\tau + K_d \frac{de(t)}{dt},

where u(t) is the control signal, e(t) is the error (difference between setpoint and measured value), and K_p, K_i, K_d are tunable gains for proportional, integral, and derivative terms, respectively; this model is foundational in automating processes like robotics and HVAC systems.^[77] Software tools like MATLAB and ANSYS support simulation-driven design by integrating modeling, analysis, and optimization capabilities. MATLAB facilitates dynamic system simulations and control design through its Simulink environment, while ANSYS provides multiphysics FEA for detailed structural and thermal evaluations, enabling engineers to couple models for holistic assessments. These tools accelerate design workflows by automating repetitive tasks and visualizing outcomes.^[78]^[79] The primary benefits of these models include substantial cost reductions and risk mitigation by minimizing reliance on physical prototypes and trials. In automotive engineering, finite element crash test models, developed since the 1970s with early codes like DYNA3D, simulate collision dynamics to predict occupant safety and structural deformation, avoiding the high expenses of destructive physical tests and enabling significant cost savings through iterative virtual validations. A notable case is Boeing's adoption of digital twin models for aircraft design starting in the 2010s, where virtual replicas of systems like the 787 Dreamliner's structures and avionics are used for real-time simulation, predictive maintenance, and lifecycle optimization, reducing prototyping time and enhancing reliability across production phases.^[39]^[80]^[81]^[82] In social and economic analysis, models play a crucial role in predicting complex human behaviors and informing policy decisions by simulating interactions within societies and markets. Agent-based models (ABMs), for instance, represent individuals or groups as autonomous agents whose decisions and interactions generate emergent patterns, enabling forecasts of social dynamics such as urban segregation or epidemic spread. A seminal example is Thomas Schelling's 1971 model of segregation, where agents with only mild preferences for similar neighbors spontaneously produce highly segregated communities, illustrating how micro-level choices can drive macro-level outcomes and guide urban planning policies.^[83] These models have been applied in policy contexts to evaluate interventions, such as simulating the effects of housing incentives on community integration or public health measures on behavior adoption.^[84] Economic models provide foundational tools for understanding market equilibria and strategic interactions. The Arrow-Debreu general equilibrium model, developed in 1954, formalizes how supply, demand, and prices balance across multiple goods and agents under assumptions of perfect competition and complete markets, serving as a benchmark for analyzing resource allocation and welfare in economies.^[85] Complementing this, game theory models like the Nash equilibrium, introduced by John Nash in 1950, describe stable outcomes where no player benefits from unilaterally changing strategy, widely used to predict behaviors in auctions, oligopolies, and international trade negotiations. These frameworks aid policymakers in designing regulations, such as antitrust laws or tariff policies, by quantifying potential market responses. In social applications, models adapted from epidemiology track the diffusion of behaviors, innovations, or norms through populations. Everett Rogers' 1962 diffusion of innovations model conceptualizes adoption as an S-shaped curve driven by interpersonal communication and adopter categories (innovators, early adopters, etc.), explaining phenomena like the spread of agricultural techniques or public health campaigns.^[86] This approach has informed social policies, such as vaccination drives or educational reforms, by predicting tipping points where majority adoption occurs based on network effects and perceived benefits. Despite their utility, modeling human systems faces significant challenges, particularly in capturing heterogeneity—variations in agent attributes like preferences or resources—and non-linearity, where small changes yield disproportionate outcomes due to feedback loops. Heterogeneity complicates calibration, as diverse agent behaviors can lead to unpredictable aggregations, requiring advanced computational methods to avoid oversimplification.^[87] Non-linearity, prevalent in economic cycles or social contagions, introduces bifurcations and chaos, making long-term predictions sensitive to initial conditions and challenging traditional linear assumptions in policy simulations.^[88] Modern advancements include network models that analyze social media influence, leveraging graph theory to map connections and assess propagation. Post-2000s research, such as the 2003 influence maximization framework, uses centrality measures—like degree (number of connections) or betweenness (control over information flow)—to identify key nodes that amplify message spread in platforms like Twitter or Facebook.^[89] These models, building on Freeman's 1978 centrality concepts, help predict viral trends and inform digital policies on misinformation or targeted advertising.

Limitations and Ethical Considerations

Inherent Limitations

Models inherently involve simplifications that trade fidelity for tractability, leading to approximations that omit intricate real-world details. For instance, the ideal gas law, expressed as PV = nRT, assumes point-like particles with no intermolecular forces or volume, ignoring attractions and repulsions that affect real gases at high pressures or low temperatures, resulting in deviations from predicted behavior.^[90] Such trade-offs enable computational feasibility but introduce systematic errors by reducing complex systems to manageable equations.^[91] Key sources of error in models include model mismatch from incorrect assumptions about underlying processes, parameter uncertainty due to imprecise estimates of inputs, and emergence of unpredicted behaviors in complex systems where interactions produce novel outcomes not captured by initial formulations. Model mismatch arises when the chosen structure fails to reflect reality, such as overlooking nonlinear dynamics in simplified linear approximations.^[92] Parameter uncertainty stems from variability in data or estimation methods, amplifying prediction errors, while emergence manifests in systems like ecosystems or networks where collective effects defy component-level predictions.^[93] Among these, certain error types highlight fundamental unmodelability, such as black swan events—rare, high-impact occurrences beyond typical distributions that models cannot anticipate due to reliance on historical data. Nassim Nicholas Taleb's 2007 analysis describes these as outliers with extreme consequences, often rationalized post-hoc but inherently unpredictable in probabilistic frameworks.^[94] Scaling issues further compound limitations when extrapolating from micro-level details to macro-scale phenomena, as relative effects identified in small-scale data may not translate to absolute levels at larger scopes, leading to the "missing intercept" problem in aggregation.^[95] A prominent example of these limitations occurred during the 2008 financial crisis, where models like the Gaussian copula underestimated default correlations in mortgage-backed securities, assuming independence in tails that masked clustering risks and contributed to widespread failures in risk assessment.^[96] To mitigate such uncertainties, ensemble modeling combines outputs from multiple models to average errors and quantify variability, reducing epistemic uncertainty through weighted averaging and providing more robust predictions than single-model approaches.^[97]

Ethical and Philosophical Issues

Ethical concerns surrounding models, particularly in artificial intelligence and predictive analytics, often center on biases that perpetuate social inequalities. For instance, facial recognition systems developed in the 2010s exhibited significant racial biases, with error rates as high as 34.7% for darker-skinned females, compared to 0% for lighter-skinned males, leading to disproportionate misidentifications of people of color in applications like law enforcement.^[98] Such biases arise from training datasets that underrepresent certain demographics, amplifying historical inequities in technology deployment.^[99] Misuse of models in public policy has also raised ethical alarms, as flawed predictions can influence decisions affecting millions. During the 2020 COVID-19 pandemic, influential epidemiological models, such as the Imperial College London projections estimating up to 2.2 million U.S. deaths without intervention, prompted strict lockdown policies but faced criticism for methodological flaws like overreliance on unverified parameters, leading to debates over their disproportionate socioeconomic impacts.^[100] These cases highlight the moral responsibility of modelers to ensure transparency and robustness to avoid unintended harms from policy missteps.^[101] Philosophical debates on models revolve around their ontological status, pitting scientific realism against instrumentalism. Realists argue that successful models approximate objective truths about unobservable realities, while instrumentalists, as articulated by Bas van Fraassen in his 1980 framework of constructive empiricism, view models merely as empirical tools for prediction and explanation without committing to their literal truth.^[102] This tension questions whether models represent the world or serve pragmatic functions, influencing how scientists and policymakers interpret their reliability. A seminal critique appears in Patrick Suppes' 1960 analysis, which contrasts mathematical models as abstract structures with empirical models as approximations of reality, emphasizing their limited fidelity and the philosophical pitfalls of conflating the two.^[103] Issues of trust in models stem from overreliance, which can expose "model fragility"—sensitivity to assumptions that undermines policy applications. In climate policy debates, excessive dependence on integrated assessment models has been criticized for producing uncertain projections due to subjective inputs like discount rates, fostering fragility in long-term decision-making and public skepticism.^[101] This overreliance risks amplifying uncertainties, as seen in divergent model outcomes for emission scenarios that shape international agreements.^[104] To address these challenges, regulatory frameworks are emerging to govern high-risk models. The EU AI Act, which entered into force on 1 August 2024 with provisions applying from February 2025 onward, classifies certain AI models as high-risk if they pose threats to health, safety, or rights—such as biometric categorization systems—and mandates risk assessments, data governance, and transparency to mitigate biases and misuse.^[105]^[106] These measures aim to balance innovation with ethical accountability, requiring providers of high-risk systems to ensure human oversight and conformity evaluations.^[107]

References

[1]
Models | Internet Encyclopedia of Philosophy
A model is considered to be a representation of some object, behavior, or system that one wants to understand.Models in Science · Physical Models · Mathematical Models · State Spaces
[2]
Models in Science - Stanford Encyclopedia of Philosophy
Feb 27, 2006 · Many scientific models are representational models: they represent a selected part or aspect of the world, which is the model's target system.
[3]
What is a Model? - SERC (Carleton)
Oct 12, 2009 · A pedagogical content page on "What is a Model?" that defines scientific models as human constructs approximating real-world systems, ...
[4]
Developing and Using Models - University of Hawaii at Manoa
Models are used to represent systems, or parts of systems, in order to study and communicate ideas about those systems.
[5]
Scientific Models | CK-12 Foundation
There are three types of models: physical, conceptual, and mathematical. Many models are created on computers due to their ability to handle and manipulate ...
[6]
Types of Models - SEBoK
May 24, 2025 · 1.1 Formal versus Informal Models · 1.2 Physical Models versus Abstract Models · 1.3 Descriptive Models · 1.4 Analytical Models · 1.5 Hybrid ...Formal versus Informal Models · Physical Models versus... · System Models
[7]
Models in science
Scientific modelling In science, a model is a representation of an idea, an object or even a process or a system that ...
[8]
Practices of Science: Using Models - University of Hawaii at Manoa
Models can take many forms, including diagrams, three-dimensional models, and computer models.
[9]
2.1: Models in Science and Engineering - Math LibreTexts
Apr 30, 2024 · 1. Observe nature. · 2. Develop a hypothesis that could explain your observations. · 3. From your hypothesis, make some predictions that are ...Missing: types | Show results with:types
[10]
SEP2: Developing and Using Models - The Wonder of Science
Models, starting concrete, progress to abstract, include diagrams, physical replicas, and simulations. They represent systems, aid explanations, and generate ...
[11]
model - Understanding Science
In science, the term model can mean several different things (e.g., an idea about how something works or a physical model of a system that can be used for ...
[12]
Scientists Say: Model
Mar 4, 2024 · Models are representations of real-life systems or processes that we use to ask questions, make predictions and test our knowledge.
[13]
Scientific Models | Texas Gateway
Scientific models are representations of objects, systems or events and are used as tools for understanding the natural world.<|separator|>
[14]
[PDF] Basic model types, key definitions, and a general framework ... - CNS
Abstract. A model is a systematic description of an object or phenomenon that shares important characteristics with its real-world counterpart and supports ...
[15]
[PDF] Scientific Models - Monash University
A scientific model is a physical, mathematical, or conceptual representation of ideas, events, or processes, and a powerful way to represent simplifications.
[16]
Making Models Work - DukeSpace
Scientific models are used to investigate reality. Here “model” refers to a representation which is created by an agent for a particular inferential purpose.
[17]
[PDF] Notes on Thermodynamics, Fluid Mechanics, and Gas Dynamics
Dec 15, 2021 · Models are often used in fluid mechanics to predict the kinematics and dynamics of full-scale (often referred to as prototype) flows.
[18]
[PDF] An introduction to metatheories, theories, and models
Metatheory is about the investigation of theory itself. Theory is a system of assumptions. Model is a tentative structure used as a testing device.
[19]
[PDF] Theories, Models, and Effects—Oh My! Differentiating Similar ...
Theories are broad explanations, models are for prediction, and effects are descriptive. Theories are broad, models are moderate, and effects are narrow in ...
[20]
[PDF] Models and Modeling: An Introduction | Ambitious Science Teaching
Modeling is how scientists represent ideas and change them based on new evidence. A scientific model represents a system or phenomenon, and can be drawings, ...Missing: definition | Show results with:definition
[21]
[PDF] ARCHITECTURE AND MATHEMATICS IN ANCIENT EGYPT
Highly illustrated with plans, diagrams and figures, this book is essential reading for all scholars of ancient Egypt and the architecture of ancient cultures.
[22]
Ptolemy's Model of the Solar System - Richard Fitzpatrick
Ptolemy constructed an ingenious geometric model of the moon's orbit which was capable of predicting the lunar ecliptic longitude to reasonable accuracy.
[23]
(PDF) Reconstructing Galileo's Inclined Plane Experiments for ...
Galileo performed his free fall experiments with the inclined plane in 1603 and published them in his Discourses on Two New Sciences (1638).
[24]
https://digitalcommons.dartmouth.edu/cgi/viewcontent.cgi?article=1032&context=facoa
[25]
The world's first general purpose computer turns 75 | Penn Today
Feb 11, 2021 · Presper Eckert, ENIAC was the fastest computational device of its time, able to do 5,000 additions per second, but because it had no ...Missing: post- | Show results with:post-
[26]
a study of the function of theory, probability and law in science.
Braithwaite, R. B. (1953). Scientific explanation; a study of the function of theory, probability and law in science. Cambridge University Press.
[27]
World modeling - UNESCO Digital Library
... science, systems dynamics and econometrics into what today comprises global modeling. ... 1970s is frequently associated with the beginning of world modeling ...
[28]
An Introductory Review of Deep Learning for Prediction Models With ...
We present in this paper an introductory review of deep learning approaches including Deep Feedforward Neural Networks (D-FFNN), Convolutional Neural Networks ...
[29]
Physical Model Definition, Examples & Limitation - Lesson | Study.com
Helping scientist visualize or predict things. Advantages of physical models include accuracy, safety, visualization, education, and trying things out that ...Advantages of Physical Models · What Is a Limitation of a...
[30]
What Are the Four Types of Models in Engineering? - BytePlus
Physical models are tangible, scaled representations of real-world systems, structures, or processes. These three-dimensional representations provide engineers ...Physical Models In... · Mathematical Models In... · Descriptive Models In...<|separator|>
[31]
Physical Scale Model: Visualization, Uses, Challenges, Prices, and ...
Jan 10, 2025 · Common materials for PSMs include foam board, balsa wood, cardboard, plastic sheets, clay, and 3D-printed parts. Foam board, famous for its ...
[32]
https://formlabs.com/blog/3d-printing-architectural-models/
[33]
Introduction to Scaling Laws - Av8n.com
So if the width is scaled up by a factor of K and the height is scaled up by a factor of K, the area necessarily gets scaled up by a factor of K2. 1.2 Volume, ...
[34]
Wind Tunnel Test - an overview | ScienceDirect Topics
Wind tunnel tests refer to experiments conducted in a wind tunnel, which is designed to simulate controlled airflow conditions around a physical model, ...Missing: applications integrity
[35]
[PDF] Wind-Tunnel Investigations of the Aerodynamics of Bridge Stay ...
Experiments on stationary sectional models of scaled replica of bridge stay cables were carried out in a 2- by 3-m wind tunnel. The cross-sectional shapes of ...
[36]
3.3 Physical Modelling | Design Technology - Ruth-Trumpold
Designers use physical models to visualize information about the context that the model represents. It is very common for physical models of large objects ...
[37]
Researching the Wright Way | National Air and Space Museum
The 1900 glider was the Wrights' first piloted aircraft, flight tested at Kitty Hawk that fall. It incorporated the wire-braced biplane structure and wing- ...
[38]
Newton's Laws of Motion | Glenn Research Center - NASA
Jun 27, 2024 · Sir Isaac Newton's laws of motion explain the relationship between a physical object and the forces acting upon it.Newton's First Law: Inertia · Newton's Second Law: Force · The Acceleration Of An...Missing: foundational | Show results with:foundational
[39]
Eighty Years of the Finite Element Method: Birth, Evolution, and Future
Jun 13, 2022 · This document presents comprehensive historical accounts on the developments of finite element methods (FEM) since 1941, with a specific ...
[40]
[PDF] CHAPTER-1 MATHEMATICAL MODELLING
A mathematical model can be represented as a functional relationship of the form. Dependent variable = f( independent variables, parameters, forcing functions).
[41]
[PDF] A Comparison of Deterministic vs Stochastic Simulation Models for ...
Deterministic models have a known set of inputs which will result in an unique set of outputs. A stochastic simulation model has one or more random variables ...
[42]
Differential Equations - Euler's Method - Pauls Online Math Notes
Nov 16, 2022 · We'll use Euler's Method to approximate solutions to a couple of first order differential equations. The differential equations that we'll ...
[43]
[PDF] An Introduction to Mathematical Modelling
(e.g. population size - or any other variable!) over time. In the case of stochastic models a more statistically well founded method of parameter estimation.
[44]
Deep learning: Historical overview from inception to actualization ...
This study aims to provide a historical narrative of deep learning, tracing its origins from the cybernetic era to its current state-of-the-art status.
[45]
[PDF] CONCEPTUAL MODELING: DEFINITION, PURPOSE AND BENEFITS
The conceptual model is a concise and precise consolidation of all goal-relevant structural and behavioral features of the SUI presented in a predefined format.
[46]
Conceptual Modeling - an overview | ScienceDirect Topics
Written languages for conceptual modeling may be graphical (diagrams) and/or textual, with the terms “abstract syntax” and “concrete syntax” distinguishing ...
[47]
Introduction to Ontology Concepts and Modeling - Boxes and Arrows
Nov 2, 2021 · An ontology is a formal system for modeling concepts and their relationships. Unlike relational database systems, which are essentially interconnected tables.
[48]
[PDF] Jay Wright Forrester and the Field of System Dynamics
hour sketching the stock/flow diagram of what became one of the most significant and widely debated models in the history of system dynamics (see Fig. 14) ...
[49]
Food Web: Concept and Applications | Learn Science at Scitable
Food web is an important conceptual tool for illustrating the feeding relationships among species within a community, revealing species interactions and ...
[50]
[PDF] The Structure of Tradeoffs in Model Building
Despite their best efforts, scientists may be unable to construct models that simultaneously exemplify every theoretical virtue. One.Missing: PDF | Show results with:PDF
[51]
The Structure of Tradeoffs in Model Building - ResearchGate
Aug 7, 2025 · ... generality can be modiﬁed due to the interaction of ﬁdelity ... model properties. I then use three case studies to show that, rather ...
[52]
[PDF] on model building - DSpace@MIT
Science is concerned with describing the universe with fidelity and parsimony. ... and generality of their discovered knowledge. Model-building for problem ...
[53]
Principles of modularity, regularity, and hierarchy for scalable systems
Aug 6, 2025 · This paper offers a number of observations about properties, dependencies and tradeoffs among these principles and proposes a formal model where ...
[54]
Principles of modularity, regularity, and hierarchy for scalable systems
Jun 30, 2007 · This paper offers a number of observations about properties, dependencies and tradeoffs among these principles and proposes a formal model where
[55]
[PDF] Sensitivity and Uncertainty Analyses - US EPA
Sensitivity Analysis – The computation of the effect of changes in input values or assumptions (including boundaries and model functional form) on the outputs.
[56]
[PDF] 9. Model Sensitivity and Uncertainty Analysis - ResearchGate
Uncertainty in model output can also result from errors in the model structure compared to the real system, and approximations made by numerical methods ...
[57]
Sensitivity analysis of complex models - ScienceDirect.com
It determines how sensitively the output of a sensitivity analysis depends on the values assigned to the model parameters.
[58]
A systematic approach to model validation based on Bayesian ...
The rejection procedures are based on Bayesian updates, where the prior density is related to the current candidate model and the posterior density is obtained ...
[59]
[PDF] Standard for Verification and Validation in Computational Fluid ...
ASME V&V 20-2009. Standard for. Verification and Validation in Computational Fluid. Dynamics and Heat Transfer. AN AMERICAN NATIONAL STANDARD. Page 2. iii.
[60]
1.3.5.15. Chi-Square Goodness-of-Fit Test
The chi-square test (Snedecor and Cochran, 1989) is used to test if a sample of data came from a population with a specific distribution. An attractive feature ...
[61]
What Is Regularization? | IBM
By increasing bias and decreasing variance, regularization resolves model overfitting. Overfitting occurs when error on training data decreases while error on ...
[62]
What Is Black Box AI and How Does It Work? - IBM
Users cannot easily validate a model's outputs if they don't know what's happening under the hood. Furthermore, the opacity of a black box model can hide ...What is black box artificial... · Why do black box AI systems...
[63]
[PDF] Climate Models and Their Evaluation
from the IPCC Special Report on Emission Scenarios (SRES). The limitations of palaeoclimate tests are that uncertainties in both forcing and actual climate ...
[64]
Climate predictions and observations | Nature Geoscience
Additional context on the evolving understanding of climate change can be found by looking at the IPCC conditional predictions made in 1990, 1995 and 2007.<|control11|><|separator|>
[65]
The Human Genome Project: big science transforms biology and ...
Sep 13, 2013 · The Human Genome Project has transformed biology through its integrated big science approach to deciphering a reference human genome sequence.
[66]
Highly accurate protein structure prediction with AlphaFold - Nature
Jul 15, 2021 · In particular, AlphaFold is able to handle missing the physical context and produce accurate models in challenging cases such as intertwined ...Missing: impact | Show results with:impact
[67]
Hybrid QM/classical models: Methodological advances and new ...
Oct 27, 2021 · Hybrid methods that combine quantum mechanical descriptions with classical models are very popular in molecular modeling.
[68]
What is Iterative Design | Autodesk
Iterative design is the cyclical process of rapid prototyping, testing, and refining. In this approach, designers understand that the perfect product is only ...
[69]
Benefits of 3D CAD Modeling for Mechanical Engineering Design
Aug 31, 2022 · 3D CAD lets mechanical engineers create and change designs quickly, run simulations, and make adjustments without building physical models.
[70]
What is Finite Element Analysis (FEA)? - Ansys
Finite element analysis (FEA) is the process of predicting an object's behavior based on calculations made with the finite element method (FEM).
[71]
[PDF] Finite Element Analysis (FEA) or Finite Element Method (FEM)
The Finite Element Analysis (FEA) is a numerical method for solving problems of engineering and mathematical physics. Useful for problems with complicated.
[72]
https://pubs.aip.org/aip/cpr/article/2/4/041303/281332/Hybrid-QM-classical-models-Methodological-advances
[73]
Model-Based Systems Engineering (MBSE) - MATLAB & Simulink
Engineers use model-based systems engineering (MBSE) to manage system complexity, improve communication, and produce optimized systems.
[74]
Ansys | Engineering Simulation Software
Ansys engineering simulation and 3D design software delivers product modeling solutions with unmatched scalability and a comprehensive multiphysics ...
[75]
Car Crash Simulation | Make Software, Change the World!
In 1976, John Hallquist at Lawrence Livermore National Laboratory (LLNL) created DYNA3D, which used the finite element method to measure the impact of nuclear ...
[76]
Crash Simulation Software: An In-Depth Guide for Engineers
Aug 12, 2024 · Crash simulations ensure that designs adhere to these regulations, minimizing the risk of costly redesigns and delays. Cost Efficiency: Physical ...
[77]
[PDF] 787 Dreamliner Difference - Boeing
Sep 17, 2024 · example is the use of digital twins in product development and testing. Digital twins allow software developers, engineers and production.
[78]
Dynamic models of segregation - Taylor & Francis Online
Aug 26, 2010 · This is an abstract study of the interactive dynamics of discriminatory individual choices. One model is a simulation in which individual members of two ...
[79]
Agent-based modelling as a method for prediction in complex social ...
Feb 22, 2023 · This special issue presents the range of positions on ABM and prediction, tackling methodological, epistemological and pragmatic issues.
[80]
Existence of an Equilibrium for a Competitive Economy - jstor
A. Wald has presented a model of production and a model of exchange and proofs of the existence of an equilibrium for each of them. Here proofs of the.
[81]
Diffusion of Innovations - Everett M. Rogers - Google Books
Title, Diffusion of Innovations ; Author, Everett M. Rogers ; Publisher, Free Press of Glencoe, 1962 ; ISBN, 0598411046, 9780598411044 ; Length, 367 pages.
[82]
Structural Effects of Agent Heterogeneity in Agent-Based Models
Jun 30, 2022 · We show that initializations with unstructured heterogeneity can interfere with a structural understanding of emergent processes, especially ...
[83]
Nonlinearity and Chaos in Economic Models - jstor
A substantial amount of recent research has sought to elucidate the role of nonlinearity and chaos in macroeconomic models. Some of the work has been.
[84]
[PDF] Maximizing the Spread of Influence through a Social Network
ABSTRACT. Models for the processes by which ideas and influence propagate through a social network have been studied in a number of do-.
[85]
10.8 The Behavior of Real Gases
Molecular volumes and intermolecular attractions cause the properties of real gases to deviate from those predicted by the ideal gas law. Key Equation. van der ...
[86]
The limitations of mathematical modeling - Strategy+business
Jan 3, 2023 · A new book demonstrates the danger of the perfect model.
[87]
Error and uncertainty in modeling and simulation - ScienceDirect.com
This article develops a general framework for identifying error and uncertainty in computational simulations that deal with the numerical solution of a set ...
[88]
Merits and Limitations of Mathematical Modeling and Computational ...
Aug 11, 2021 · The recent mathematical models about COVID-19 and their prominent features, applications, limitations, and future perspective are discussed and reviewed.
[89]
[PDF] Black swans, or the limits of statistical modelling - risk-engineering.org
... Taleb (2007). 9 / 58. Page 11. Rare events are not all black swans. ▷ The ... ▷ By definition, black swan events are not predictable. ▷ You can try to ...
[90]
The 'missing intercept' problem with going from micro to macro
Sep 9, 2025 · The 'missing intercept' problem occurs when scaling micro estimates to macro, as cross-sectional data only identifies relative effects, not ...
[91]
[PDF] The Gaussian Copula and the Financial Crisis: A Recipe for Disaster ...
Multivariate normal distributions of credit risk, such as produced by the Gaussian copula, fail to capture default clustering; in times of crisis if one obligor ...
[92]
Ensemble deep learning models for prediction and uncertainty ...
A key strength of ensemble models is their ability to reduce model prediction uncertainty (Murray 2018). The uncertainty in the model parameters is known as ...
[93]
Gender Shades: Intersectional Accuracy Disparities in Commercial ...
In this work, we present an approach to evaluate bias present in automated facial analysis algorithms and datasets with respect to phenotypic subgroups. Using ...
[94]
[PDF] Understanding bias in facial recognition technologies
Bias in facial recognition stems from historical racism, white privilege, imbalanced datasets, and unstable categories of race, ethnicity, and gender.
[95]
Influential Covid-19 model shouldn't guide U.S. policies, critics say
Apr 17, 2020 · Epidemiologists are criticizing an influential coronavirus model as flawed and warning against relying on it as the basis for government ...
[96]
[PDF] The Use and Misuse of Models for Climate Policy - MIT
Apr 8, 2015 · Abstract: In recent articles, I have argued that integrated assessment models (IAMs) have flaws that make them close to useless as tools for ...
[97]
Scientific Realism - Stanford Encyclopedia of Philosophy
Apr 27, 2011 · Van Fraassen (1980) reinvented empiricism in the scientific context, evading many of the challenges faced by logical empiricism by adopting a ...What is Scientific Realism? · Antirealism: Foils for Scientific...
[98]
[PDF] a comparison of the meaning and uses of models - Stanford University
A COMPARISON OF THE MEANING AND USES OF MODELS. IN MATHEMATICS AND THE EMPIRICAL SCIENCES by. PATRICK SUPPES. TECHNICAL REPOR!' NO. 33. August 25, 1960.Missing: critique | Show results with:critique
[99]
Flawed Climate Models - Hoover Institution
Apr 5, 2017 · The problem is that these models have serious limitations that drastically limit their value in making predictions and in guiding policy.
[100]
Article 6: Classification Rules for High-Risk AI Systems - EU AI Act
AI systems of the types listed in Annex III are always considered high-risk, unless they don't pose a significant risk to people's health, safety, or rights.
[101]
EU AI Act: first regulation on artificial intelligence | Topics
Feb 19, 2025 · High risk. AI systems that negatively affect safety or fundamental rights will be considered high risk and will be divided into two categories: ...