{
  "id": "first-principles-thinking",
  "title": "First-principles thinking",
  "lang": "en",
  "source_lang": "en",
  "summary": "First-principles thinking is an approach to inquiry and problem-solving that examines basic assumptions and reconstructs explanations or solutions from explicitly justified starting points.",
  "category": "philosophy",
  "subcategory": "metaphysics-epistemology",
  "type": "concept",
  "aliases": [
    "Reasoning from first principles",
    "first-principles reasoning"
  ],
  "generated_at": "2026-11-06",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/first-principles-thinking/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/first-principles-thinking/",
      "json": "https://aiwiki.page/zh/first-principles-thinking.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "First-principles thinking is an approach to inquiry and problem-solving that identifies the basic assumptions, facts, or constraints relevant to a question and reasons from them rather than relying primarily on precedent, convention, or analogy. Its central task is not simply to divide a problem into smaller parts, but to distinguish what is genuinely necessary or well supported from what has merely been inherited. The expression connects an older philosophical concern with the foundations of knowledge to contemporary methods of analysis in science, engineering, and practical decision-making. It does not designate a single standardized method, nor does it guarantee that its starting points are certain."
    },
    {
      "heading": "What counts as a first principle?",
      "text": "A first principle is a starting point within a system of explanation or reasoning. It is “first” in the order of justification, not necessarily in time or historical discovery. A proposition may serve as a first principle because it is treated as self-evident, adopted as an axiom, established independently, or accepted as a sufficiently reliable constraint for the problem at hand.\n\nSeveral kinds of starting point must be distinguished:\n\n- Logical principles govern acceptable inference, such as the law of noncontradiction in classical logic.\n- Mathematical axioms specify the assumptions under which theorems are proved.\n- Empirical principles express well-supported regularities or physical constraints, but remain answerable to observation.\n- Definitions and modeling assumptions establish how a problem is represented and which features are included.\n- Practical objectives and constraints specify what a proposed solution must achieve and which limits it must respect.\n\nThese categories do not have the same epistemic status. An objective such as minimizing cost is not a fact about nature, and a simplifying assumption is not a demonstrated law. Careful first-principles reasoning makes such differences explicit rather than treating all starting points as equally certain.\n\nThe relevant level of fundamentality also depends on the question. A structural calculation may begin with established properties of materials without deriving them from quantum mechanics. Those properties function as starting points for that calculation, although they are not fundamental in an unrestricted sense."
    },
    {
      "heading": "Philosophical background",
      "text": "Aristotle and demonstrative knowledge\n\nAristotle developed an influential account of first principles in his treatment of demonstrative knowledge. In the Posterior Analytics, a demonstration explains a conclusion through premises that are prior to it in the relevant explanatory order. If every premise required another demonstration, justification would either continue indefinitely or become circular. Demonstrative knowledge therefore requires starting points that are not themselves established by demonstration of the same kind.\n\nAristotle did not present the acquisition of first principles as merely inventing assumptions. His account connects perception, memory, experience, and the recognition of universals with the apprehension of principles. His treatment thus distinguishes the order in which people discover something from the order in which a scientific explanation is organized.\n\nThis historical account should not be equated without qualification with the modern business or engineering expression. Aristotle was concerned with the structure of knowledge and explanation, not with a general recipe for innovation.\n\nDescartes and the search for secure foundations\n\nRené Descartes pursued a different foundational project by questioning beliefs that could be doubted and seeking a secure basis for knowledge. His Discourse on the Method also proposed dividing difficulties into parts and proceeding in an orderly way from simpler matters to more complex ones.\n\nThese ideas resemble aspects of contemporary first-principles thinking: examining assumptions, decomposing difficulties, and reconstructing an argument. Nevertheless, methodical doubt, decomposition, and reasoning from first principles are distinct activities. Breaking a problem into pieces does not establish that the resulting pieces are true or fundamental.\n\nFoundationalism\n\nIn epistemology, foundationalism holds, broadly, that some justified beliefs provide support for others without depending on them in the same way. First-principles thinking has a structural resemblance to this view, but a practical user of the method need not endorse a complete foundationalist theory.\n\nAn engineer can provisionally accept measurements and physical laws while remaining open to revising them. Conversely, coherentism emphasizes mutual support among beliefs rather than a one-directional hierarchy resting on basic beliefs. This contrast raises a central question for first-principles approaches: whether a proposed foundation is independently warranted or gains its credibility from the broader system in which it operates."
    },
    {
      "heading": "Reasoning process",
      "text": "There is no universally prescribed sequence, but a first-principles analysis commonly contains several distinguishable operations.\n\nDefining the question\n\nA problem must be specified before its assumptions can be examined. “Improve a transport system,” for example, could mean reducing travel time, lowering operating costs, increasing accessibility, or reducing energy consumption. Different objectives produce different analyses.\n\nProblem definition therefore includes separating descriptive questions—what happens and why—from normative questions about what should be achieved. Technical facts alone do not settle the choice of goals.\n\nIdentifying and classifying assumptions\n\nThe analysis then distinguishes observations, established principles, conventions, estimates, and preferences. Claims such as “this material has a measured strength,” “this component is normally purchased from a supplier,” and “customers require this particular design” require different kinds of support.\n\nThe purpose is not to reject inherited knowledge automatically. It is to identify which claims are indispensable and which are open to change. Established expertise can supply reliable starting points; customary practice may also encode constraints that are not immediately visible.\n\nConstructing an explanation or solution\n\nOnce the premises are explicit, reasoning connects them to possible conclusions. In a deductive argument, the structure can be represented as\n\n\\[P_1, P_2, \\ldots, P_n \\vdash C,\\]\n\nwhere the premises \\(P_1,\\ldots,P_n\\) entail the conclusion \\(C\\) under specified rules of inference.\n\nDeductive reasoning can establish what follows from the premises, but logical validity does not establish that those premises are true. A formally valid argument can still be unreliable as a description of the world.\n\nPractical inquiry also uses inductive reasoning to infer regularities from observations and abductive reasoning to propose explanations. First-principles thinking is therefore not necessarily a purely deductive procedure.\n\nTesting and revision\n\nWhen conclusions concern the empirical world, they require comparison with evidence. Predictions may be assessed through observation, an experiment, simulation, or a prototype. Disagreement can reveal incorrect premises, omitted mechanisms, numerical errors, or a mismatch between the model and its intended application.\n\nThis makes empirical first-principles work an iterative process rather than a one-way derivation from unquestionable truths. Its relationship to the scientific method lies in making assumptions explicit and exposing consequences to testing."
    },
    {
      "heading": "Applications",
      "text": "Mathematics and formal systems\n\nIn mathematics, reasoning from axioms and definitions provides a precise model of derivation from stated premises. A proof establishes a result relative to those premises and the accepted inference rules.\n\nWorking “from first principles” can also have a narrower pedagogical meaning: deriving a result directly from a definition rather than invoking an already established theorem. For example, calculating a derivative from its defining limit illustrates this usage. It does not require rebuilding all of mathematics from its foundations.\n\nPhysics and engineering\n\nIn physics and engineering, a first-principles model derives behavior from underlying laws rather than relying solely on a fitted relationship or an existing design. Such models still require boundary conditions, material properties, approximations, and computational choices.\n\nFor example, the idealized minimum energy needed to raise a mass \\(m\\) through a height \\(h\\), near Earth's surface where gravitational acceleration \\(g\\) is approximately constant, is\n\n\\[E_{\\min}=mgh.\\]\n\nThis provides a baseline for evaluating a lifting system. It does not determine the actual electricity consumption, which also depends on efficiency, friction, control behavior, and other losses. Nor does it determine the system’s purchase price, maintenance cost, or safety performance.\n\nThe example illustrates a distinction between a physical lower bound and a realizable design. A bound can constrain expectations without supplying a complete solution.\n\nDesign and organizational decisions\n\nIn design and management, the approach is used to question whether existing arrangements are necessary. An analysis may distinguish a required function from the familiar product or process that currently performs it.\n\nA cost analysis, for example, can separate material inputs, fabrication, labor, transport, financing, and overhead instead of treating an existing market price as an irreducible fact. However, the cost of raw inputs is not automatically an achievable production cost. Yield losses, quality assurance, tooling, coordination, and scale may account for much of the difference.\n\nThe method is most informative when it makes these dependencies visible, rather than simply assuming that established practices are wasteful."
    },
    {
      "heading": "Relationship to other approaches",
      "text": "Reasoning by analogy transfers insights from a familiar case to a new one. First-principles thinking instead emphasizes the assumptions and mechanisms within the target problem. The approaches are complementary: analogy can generate a candidate solution, while a first-principles model can help assess whether the relevant similarities are substantive.\n\nThe Socratic method uses questioning to examine beliefs, definitions, and apparent inconsistencies. It can expose unsupported assumptions, but questioning alone does not necessarily produce an alternative explanation or design.\n\nOccam’s razor concerns the avoidance of unnecessary assumptions or complexity when comparing explanations. First-principles thinking concerns the identification and use of starting points. An explanation with fewer assumptions is not automatically better grounded.\n\nA further distinction concerns a priori and a posteriori knowledge. Reasoning from first principles is not necessarily reasoning independently of experience. A scientific model may begin with principles supported by extensive empirical investigation."
    },
    {
      "heading": "Limitations and criticisms",
      "text": "Uncertain or hidden foundations\n\nThe principal vulnerability is the selection of starting points. A purportedly basic fact may be a convention, a poor measurement, or an oversimplification. Labeling a claim a “first principle” does not provide its epistemic justification.\n\nClaims that a conclusion follows inevitably can also conceal choices about definitions, system boundaries, or acceptable trade-offs.\n\nIncomplete decomposition\n\nDividing a system into components may omit interactions among them. The behavior of an organization, ecosystem, or complex machine cannot always be inferred by considering each part in isolation. Feedback, coordination, and conditions at the system boundary can be decisive.\n\nA simplified model may remain useful, but its usefulness does not establish completeness.\n\nDependence on expertise\n\nRejecting precedent indiscriminately can discard accumulated knowledge. Some practical constraints are embedded in tacit knowledge: skills and judgments that practitioners exercise without fully articulating them. An apparently unnecessary procedure may address a rare failure mode or a coordination problem that a simplified analysis overlooks.\n\nFirst-principles reasoning therefore does not eliminate the need for expertise, historical evidence, or learning from existing solutions.\n\nNormative and computational limits\n\nPhysical facts do not by themselves determine what ought to be valued. The is–ought problem remains relevant when technical analysis is used to support ethical or political conclusions. A derivation can show consequences of an objective without establishing that the objective should be accepted.\n\nEven when the premises and goals are clear, deriving a useful answer may be computationally infeasible. Approximation, empirical estimation, and analogy can be more effective than attempting a complete derivation. The methodological issue is whether the assumptions are appropriate, their consequences are understood, and the resulting claims remain open to correction."
    }
  ],
  "links": [
    "knowledge",
    "law-of-noncontradiction",
    "quantum-mechanics",
    "aristotle",
    "rene-descartes",
    "discourse-on-the-method",
    "epistemology",
    "foundationalism",
    "coherentism",
    "deductive-reasoning",
    "logical-validity",
    "inductive-reasoning",
    "abductive-reasoning",
    "evidence",
    "experiment",
    "scientific-method",
    "mathematics",
    "physics",
    "socratic-method",
    "occams-razor",
    "a-priori-and-a-posteriori",
    "epistemic-justification",
    "tacit-knowledge",
    "is-ought-problem"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}