Entity Realism

The Philosophical Problem

Angela Bogdanova
Published by Aisentica Development
Research Dossier · Philosophy of Science · American English
Prepared for the reserved canonical https://philosophybybogdanova.com/dossiers/entity-realism/
Editorial source edition: November 7, 2026 · Not released

A scientific theory may yield accurate predictions while misdescribing the mechanisms that produce them. The realist problem becomes acute when scientific discourse names objects and processes inaccessible to unaided sight: electrons, field configurations, molecular structures, and causal intermediaries. A commitment to their existence is a substantial ontological judgment; accurate calculation alone does not entail that the inventory of a particular theoretical model is correct. As Chakravartty's analysis of scientific realism distinguishes, realism involves epistemic, semantic, and metaphysical commitments that can be defended at different strengths.

Entity realism isolates a tractable question within that wider dispute. When scientists can reliably employ a proposed entity to produce changes elsewhere, what exactly have they established? Hacking's experimental argument from manipulation assigns special epistemic importance to doing things with entities rather than merely predicting observations through models. The philosophical difficulty is to specify the bridge from reliable intervention to warranted existential belief. The present study examines each step of that bridge and the ways laboratory practice can strengthen or defeat it.

Research Scope and Method

The inquiry is historical, analytical, and documentary. Its primary texts are Hacking's articles and book, Cartwright's books on laws and capacities, and van Fraassen's account of constructive empiricism. Later responses from Morrison, Resnik, Clarke, Chakravartty, and Nanay test the coherence of selective belief. Studies of experiments, physical instruments, and computation supply comparative context. Hacking's 1982 statement of the issue is interpreted through the bibliographic record and corroborating accessible extracts; where book chapters or articles are behind subscriptions, the analysis remains limited to visible publisher summaries, verified quotations found in open scholarly commentary, and securely identified propositions.

The argument separates four tasks: identifying what an author actually argued; reconstructing the reasoning that would make the claim valid; examining counterexamples or rival causal stories; and presenting an independent philosophical judgment. Constructed cases are identified as thought experiments, not as newly collected laboratory data. A source can prove the bibliographic existence of a publication without exposing its complete argument. Consequently the bibliography records provenance, and the evidence ledger separately records access scope and specific supported claims.

Two Kinds of Scientific Realism

In Representing and Intervening, Hacking (1983) separates realism about theories from realism about entities. The former concerns the truth or approximate truth of broad theoretical representations; the latter asks whether particular objects and processes invoked by science exist. One can accept the operation of an electron beam while suspending judgment about an exhaustive or final representation of electron nature. The separation gives experimentation independent philosophical standing.

This distinction involves a disciplined asymmetry of commitments. The experimenter can rely on a limited body of tested regularities even while revising higher-level models. An electron's experimentally relevant charge-dependent behavior can be stable through changes in theoretical depiction. Yet these regularities remain conceptual claims, and their application rests on assumptions about apparatus and causal linkage. Entity realism therefore must specify the limited descriptive commitments it accepts. The central question is how much theory is required to identify an intervention target without carrying the whole of the explanatory framework along with it.

Hacking's Experimental Turn

Hacking (1982) presented experiments as sources of phenomena rather than merely tests of antecedently stated hypotheses. His Experimentation and Scientific Realism distinguishes realist commitments to entities and to theories. Experiments often create conditions seldom available outside a laboratory, allowing scientists to separate causal contributions and use already stabilized effects in further investigation. A laboratory can thereby make the alleged entity part of an explanatory and technological chain.

The argument later receives its most recognizable formulation in Hacking's (1983) chapter on experimentation and realism. In this setting, intervention means using a putative unobservable as an instrument in producing and investigating another phenomenon. The strongest reading is comparative: repeatedly successful and targeted manipulation supplies evidence different in kind from a model's predictive accommodation. The inference becomes controversial when the target is itself described through theory-laden detection, or when a device's success permits several distinct causal interpretations.

From Detection to Intervention

Detection registers a response; intervention deliberately alters a causal setting. The difference matters because a single trace may be compatible with many hidden mechanisms, whereas controlled variations can discriminate hypotheses by producing contrasting outcomes. Hacking's (1981) microscope study helps illustrate why instrumentation changes the boundary between observing and inferring, even before the question of manipulation becomes central. Observations are mediated by optics, preparation methods, and trained practices.

An experimental manipulation provides extra warrant when the intervention is selective, the predicted response is localized, and rival explanations are tested through changes in apparatus, timing, materials, or geometry. It does not acquire epistemic force simply from the fact that some causal sequence has been engineered. If a switch toggles an unknown electronic circuit, the successful operation warrants belief in a functioning causal arrangement but may not identify the internal component responsible. The ontological gain is proportional to the experiment's ability to discriminate among candidate causes.

Cartwright's Causal Realism

Cartwright's (1983) How the Laws of Physics Lie combines skepticism about the descriptive fidelity of broad idealized laws with confidence in well-established causal capacities. Her concern is that theoretical equations often describe idealized models, whereas particular causal claims can be confirmed by interventions in concrete conditions. The contrast directs attention to experimentally localized evidence without granting automatic truth to every general law used in the explanatory calculation.

Her account of causes and instrumental laws distinguishes causal inference from an inference to the globally best theoretical explanation. Cartwright's proposal stands near Hacking's in defending some existential commitments while resisting unrestricted theory realism. Their positions are nevertheless distinguishable: Hacking foregrounds the use of putative entities as experimental resources; Cartwright foregrounds causally specific knowledge and the reality of capacities. An adequate history of entity realism preserves the difference rather than assigning both thinkers one undifferentiated criterion.

Causal Knowledge and Ontological Commitment

The move from cause to existence contains an inference whose structure deserves explicit examination. If an observed effect has been produced by an intervention operating through a particular entity, that entity exists. The conditional itself is straightforward; the evidential work lies in establishing the antecedent: that it is this entity, rather than an instrument artifact, correlated process, or alternative mechanism, through which the intervention works. Cartwright's (1989) account of measurable capacities insists that causal powers can be investigated under appropriate conditions.

A credible argument begins with experimentally supported causal dependence, assesses which alternatives could yield the same observations, and identifies the minimally warranted referent. The level of commitment can vary. One test may justify that some charged carrier is causally active; a further family of tests may justify a stable type of carrier; a still richer theory assigns detailed microscopic properties. These are different epistemic achievements. Treating them as separate avoids both the premature acceptance of a complete theory and the equally premature dismissal of what material intervention has achieved.

What Counts as Manipulation?

Manipulation in this context involves reproducible, intentional changes to the circumstances in which a proposed entity produces or transmits an effect. It need not involve handling an object directly; scientists can vary fields, energies, apertures, chemical environments, or boundary conditions. A physically elaborate setup may manipulate a causal process through several mediating structures. Galison's (1997) history of microphysics instruments documents how experimental outcomes emerge from coordinated technical practices rather than a solitary direct sensory encounter.

The expression can become dangerously permissive if every successful technology is said to manipulate its favorite theoretical posit. A practical criterion requires three features: a clearly described intervention variable, a robust change in outcomes under controlled comparison, and evidence connecting that change to the asserted causal target. A fourth feature—successful resistance to serious alternative explanations—determines how specific the existential conclusion may become. The criterion is offered here as an analysis of the conditions already implicit in the dispute, without presenting it as a new named theory.

Existence, Identification, and Description

An existence claim is weaker than a comprehensive description, but it still requires reference. To say that a particular entity exists is to distinguish it from alternatives with sufficiently stable identity conditions. Chakravartty's (1998) proposal of semirealism draws attention to detection properties and auxiliary theoretical properties. The distinction clarifies how support can attach to causal features implicated in detection while other attributes remain uncertain.

Consider a signal repeatedly associated with a postulated particle. Unless the detection relation distinguishes that particle from other sources capable of producing the signal, the evidential warrant belongs to a broader causal category. Naming the preferred particle does not complete the identification. Conversely, stable interaction patterns across diverse instruments can justify more discriminating reference even when the complete microscopic picture is contested. The challenge for entity realism is not a demand for exhaustive knowledge before belief; it is the demand that the entity believed in be determined by the evidence more specifically than a placeholder for whatever caused the apparatus to respond.

Experimental Instruments and Mediation

Every experimental intervention is mediated by a material and interpretive apparatus. Experimentalists depend on detectors, calibration standards, mathematical corrections, engineering tolerances, and background assumptions. Franklin's (1986) study of neglected experiments emphasizes the historically rich practices through which results acquire credibility and bear on theoretical choice. The presence of interpretation is a normal condition of controlled measurement.

Instrumentation therefore generates a double responsibility for entity realism. The device extends causal access to previously inaccessible domains, while its internal workings require independent investigation. In a convincing experiment, the researcher can say why changing a specific setting should alter the target response, how common errors would appear, and how alternative detector models have been assessed. The larger the gap between physical intervention and displayed output, the more important that chain of justification becomes. Successful operation of a complex detector and correct ontological identification are connected achievements, but the latter does not follow automatically from the former.

Microscopy and the Reality of Observations

Microscopy offers a paradigmatic instance of instrument-mediated access. Hacking's (1981) question about seeing microscopically concerns how philosophers should understand image formation when unaided human vision is supplemented by elaborate optical and technical processes. Images can be made reliable through calibration, complementary preparations, and converging modalities; they remain products of interacting specimen, instrument, and interpretive practice.

The inferential issue is often misframed. The question is not whether an image is constructed—the instrument plainly constructs an image—but whether the constraints of that construction are sufficiently understood to support claims about what produced it. Changing illumination, specimen preparation, or imaging method may reveal whether a feature tracks the sample or the apparatus. Cross-method agreement increases confidence when errors are genuinely different. One should not count apparent agreement as independent confirmation when both instruments inherit the same untested processing assumption. This distinction is crucial to evaluating the experimental realism claimed for what microscopes disclose.

Electrons as Experimental Resources

Hacking's favored electron example is powerful because electrons are treated as resources in experiments that bring about other effects, rather than only as theoretical explanations of measured curves. His (1983) intervention chapter explicitly emphasizes manipulation of unobservables to investigate additional aspects of nature. An experimentally usable target has entered a stable causal repertoire.

The force of this example depends on what has been established across the whole experimental chain. Steering a beam with fields and observing characteristic interactions supports a family of causal claims. The evidence becomes stronger as beam properties, detector responses, and controlled interventions show consistent patterns under diverse conditions. Yet the exact microscopic description adopted in a particular theoretical interpretation may remain under revision. Entity realism is at its most plausible when the existence of a causal agent is reliably fixed by practice while disputed claims about its complete nature remain separately assessable.

Producing Effects and Inferring Causes

Producing an effect can be evidence for a cause without uniquely selecting that cause. A controlled apparatus may generate a repeatable trace through several possible mechanisms. A phenomenon that fits a particular causal model establishes compatibility first; comparative experiments establish discrimination. The distinction mirrors the wider warning in Cartwright's (1983) analysis of causal explanations, which separates detailed causal reasoning from general explanatory coverage.

Imagine, as a constructed comparison, a detector displaying a sharp pulse every time a field is switched. Suppose two plausible mechanisms predict the same pulse: movement of a hypothesized charged micro-object and rearrangement of charge in the detector itself. Repeated switching cannot decide between them while the two mechanisms remain observationally equivalent. Changing detector geometry or measuring a second downstream interaction can create a discriminating test. The experiment warrants a realist commitment to the hypothesized micro-object only to the extent that alternative causal explanations of the pulse have been materially constrained.

Cross-Instrumental Convergence

Cross-instrumental convergence has special philosophical significance because different apparatus designs can fail in different ways. A causal target that leaves appropriately related traces in independently calibrated devices is less easily reduced to one peculiar detector mechanism. Hacking's reflections on microscopic observation and Galison's (1997) account of images and counting illustrate the methodological variety that makes such convergence intelligible.

Convergence is nevertheless a claim to be tested, not a decorative description of agreement. If two devices share a calibration source, reconstruction software, or common model of the causal target, correlated error can imitate independence. Conversely, genuinely different methods may yield superficially dissimilar data that become coherent only through carefully justified transformations. Ontological confidence therefore grows from demonstrated compatibility across distinct error profiles rather than the number of devices employed. This reading preserves experimental realism's practical insight while refusing a simple vote-counting account of evidence.

Repetition, Calibration, and Error

Reproducibility is necessary for many forms of intervention-based inquiry, but repeated results from the same configuration can repeat the same error. Mayo's (1996) error-centered account of experiments emphasizes tests capable of revealing specific failures. Experiments gain force through opportunities for claims to be wrong in detectable ways. The distinction matters because a stable apparatus artifact can be reproducible with impressive precision.

Calibration ties measurements to independent standards and establishes the operating domain of instruments. Blinding, control conditions, sensitivity analyses, and changes in relevant parameters can test whether a causal attribution survives foreseeable confounders. In entity realism the key question is whether the method would likely have produced a recognizably different result if the alleged entity were absent or had the hypothesized interaction profile wrong. An answer grounded in concrete error mechanisms advances existential warrant more than an unqualified statement that the apparatus works well.

The Scope of Local Causal Claims

Entity realism is strongest within the range of interventions in which causal claims have been carefully tested. Cartwright's (1989) measurement of capacities gives a reason to privilege well-specified conditions over the unrestricted extrapolation of law-like formulas. A capacity may be manifested differently as backgrounds, couplings, or interfering processes change. Demonstrating one regularity establishes only so much about its operation in different contexts.

This locality has philosophical consequences. A laboratory can warrant the existence of a causal factor in a controlled environment while leaving unsettled its role in a distant or inaccessible setting. Equally, observing a stable capacity does not identify every essential property of the entity that exercises it. The right inferential practice records the conditions under which the manipulation is robust, the substitutions it tolerates, and the points at which expected effects fail. Such documentation is a strength rather than a concession: it supplies the domain over which the entity claim has been earned.

Morrison's Theory-Dependence Objection

The viability of entity realism depends on the distinction between knowledge adequate for successful intervention and commitment to extensive theoretical descriptions. Margaret Morrison (1990) challenges the prospect that experiment can establish the existence of theoretical entities while remaining as independent of theory as Hacking's sharpest formulations suggest. Her Theory, Intervention and Realism concerns the conceptual conditions of experimental success: an intervention must be planned, recognized, and interpreted through some information about the target and the instruments that affect it. What looks like purely practical warrant often includes explanatory and theoretical assumptions.

The objection has a stronger form and a weaker form. In the weaker form, manipulation requires beliefs about certain stable interaction patterns, and experimental realism need only concede limited dependence on them. In the stronger form, the same high-level theoretical structure that fixes the entity's identity is indispensable to operating and interpreting the experiment, leaving no stable basis for realist belief in the entity independently of belief in the theory. These forms should be distinguished because theoretical dependence by itself does not establish that the whole theory must be true.

Morrison's challenge is best understood as a demand to audit what an experiment genuinely relies upon. If its success requires only locally tested relationships that survive transformations in the broader theory, selective belief remains intelligible. If the target is specified entirely by an uncorroborated theoretical ontology and the experimental result can be redescribed without loss, the apparent independence is illusory. Whether a particular experiment occupies the first or second position is a case-sensitive matter. The positive contribution of the criticism is to replace a rhetorical distinction between representing and intervening with a detailed accounting of theoretical dependencies.

Resnik's Challenge to Experimental Realism

David B. Resnik (1994) analyzes Hacking's claim that intervention can justify realism without a commitment to approximately true scientific theories. His critical article on experimental realism identifies the central problem: knowing that one's experimental intervention targets an entity appears to require representational and inferential resources. Intervention does not remove the question of how scientists know which causes their devices employ.

The objection can be clarified by separating practical mastery from ontological discriminability. An operator may know how to produce and suppress a signal while possessing several equally successful descriptions of the mechanism. Practical control is real in such circumstances, but belief in one particular theoretical constituent requires further evidence. The device's behavior may constrain the range of candidates without resolving which member of that range should be accepted.

Resnik also places Hacking within a realism debate formerly centered on representation, explanation, and prediction. The shift to practice is important because experiments can create new evidence rather than merely reproduce data already predicted. Yet the claim that experimental activity escapes representation would overshoot that insight. Control requires variables and target descriptions. The realist can respond by treating representation as revisable and selective: successful practice may establish that at least a causally efficacious something exists at a location or within an interaction class, while leaving its exhaustive character unsettled. This response meets the objection only if the level of existential commitment is honestly adjusted to the evidence actually obtained.

Clarke and Low-Level Generalizations

Steve Clarke (2001) argues that entity realism possesses a defensible middle position between comprehensive scientific realism and empiricist antirealism. The abstract of his Defensible Territory for Entity Realism explicitly critiques Cartwright's appeal to a distinction between inference to the best explanation and inference to the most probable cause. The underlying problem is whether a realist can know enough about putative entities to manipulate them while withholding commitment to broad theories that describe those entities.

Clarke's emphasis on low-level generalizations helps identify what an experimentalist may justifiably take as established: a field variation changes a beam trajectory within a known range; a characteristic detector response follows a calibrated interaction; a particular adjustment suppresses a specific effect. These commitments are more articulate than an entirely theory-neutral technique but narrower than acceptance of an encompassing theory.

The term 'low-level' cannot itself settle the matter. Some apparently simple relationships presuppose elaborate corrections and calibration models, whereas a powerful abstract theoretical framework may be empirically entrenched in only a limited domain. The defensible distinction is functional and evidential: identify which assumptions are indispensable for a successful intervention and which additional theoretical claims are not implicated by its success. There is no universal linguistic altitude at which a generalization becomes safely realist. The distinction survives insofar as the evidence supports the particular claims in use without requiring the broader theory to be endorsed wholesale.

Detection Properties and Semirealism

Anjan Chakravartty (1998) proposes a selective realism concerned with properties implicated in detection. His original presentation of semirealism distinguishes properties causally connected with the regular behavior of detectors from auxiliary theoretical properties attributed to the same putative entities. A realist may defend commitment to the former while suspending judgment about the latter. This approach specifies in property-level terms what the manipulation argument suggests at the level of entities.

The distinction is attractive because it directly addresses theory change. If different theories preserve experimental regularities while assigning incompatible unseen properties, the experimentalist need not treat every inherited attribute as confirmed. Yet an additional question appears: how securely can the detection property be assigned to an enduring object if the identity of the object is itself part of the disputed account? A property-bearing entity is more than a list of regularities; calling two interactions features of the same entity requires an individuation argument.

Semirealism offers resources for answering that question through stable causal structure and detection relations. Its strength depends on whether those relations establish a particular referent or only an empirically equivalent pattern. The historical debate thereby reveals that realism can be selectively calibrated across several dimensions: existence, causal power, identity, and descriptive characterization. An argument about one dimension does not settle all the others. This careful calibration is part of philosophical explanation rather than a newly introduced theory.

Nanay's Singularist Alternative

Bence Nanay (2019, first available online in 2016) directly revisits the independence of entity realism from theory realism. His open-access Entity Realism and Singularist Semirealism argues that the theoretical commitments required for manipulation create pressure on an entity realist who hopes to remain entirely neutral about theories. Nanay connects the debate to his earlier Singularist Semirealism, which considers the status of singular and general scientific claims separately.

One important feature of the challenge is its attention to what precisely a scientific statement asserts. A particular observed causal interaction may be accepted without assuming that the general theoretical classification under which scientists describe it is universally correct. But the singular proposition, too, contains predicates whose applicability may depend on background knowledge. The prospect of avoiding all theory therefore recedes.

Nanay's proposal underscores the importance of the granularity of acceptance. The realist may be warranted in accepting some singularly identified processes and experimentally stabilized property instances while withholding stronger generalizations. Whether this alternative ultimately preserves Hacking's intended ontological confidence depends on the extent to which such singular commitments are genuinely informative about entities and not merely recordings of operational regularities. The present analysis retains the issue as a contest between articulated positions, not as a settled logical refutation of every form of experimental realism.

Constructive Empiricism and Acceptance

Bas van Fraassen's (1980) The Scientific Image offers an influential alternative to scientific realism. Constructive empiricism holds that science aims at empirical adequacy and that accepting a scientific theory involves belief in its empirical adequacy rather than belief in the literal truth of all its claims about unobservables. This position is frequently misunderstood as a refusal to use theoretical language or to take sophisticated technologies seriously. In fact, acceptance includes practical and methodological commitments to the theory's use.

Against this position, the entity realist asks whether intentional deployment of unobservable entities as tools changes what a rational experimenter ought to believe. An antirealist can concede that the devices work and that experimental regularities support further predictions while refusing to infer an unobservable ontology. A realist can reply that control over unseen causal powers is more naturally understood as interaction with existing targets. That reply must show how existential commitment adds epistemic content beyond the empirically adequate description of the apparatus.

The disagreement is strongest where both camps accept exactly the same experimental data and disagree about their ontological interpretation. A decision cannot be obtained by repeating the experimental success already admitted by both. It requires an argument about whether reliable causal manipulation supplies evidence for particular underlying entities, how it excludes alternatives, and how much unobservability should matter in evaluating the result. The debate concerns justified belief, not permission to continue successful scientific work.

Underdetermination and Historical Change

Historical replacement of scientific explanations demonstrates that practical success can coexist with ontological mistakes. Scientists have employed theoretical categories that later lost their standing while retaining experimentally useful procedures. The challenge is neither to erase historical achievement nor to assume that every successful experimental intervention names the right underlying entity. Chakravartty's survey of realist commitments explains how scientific realism confronts underdetermination, theory change, and the selective defense of retained structure.

Entity realism makes a promising adjustment by favoring causal practices that can outlast the theories originally used to organize them. For example, stable experimental regularities may survive changes in the descriptions of matter and fields. This continuity provides an empirical constraint on replacement theories. It does not guarantee that the historical name attached to the regularity referred to an independently persisting kind.

The strongest historical reading is consequently comparative. Where a newly successful theory revises the causal story but preserves the same independently detectable experimental target, some ontological continuity may be warranted. Where the new explanation eliminates that target and explains all prior traces as manifestations of a different mechanism, past success supports a weaker conclusion. The record of scientific change weighs against indiscriminate realism, while it does not by itself invalidate carefully discriminated existential commitments.

Material Culture and Collective Experiment

Scientific entities become experimentally accessible through organized systems of instruments, skilled personnel, reference materials, and institutional procedures. Peter Galison's (1997) Image and Logic reconstructs the history of experimental particle physics through changing detection systems and coordinated practices. The philosophical implication is that experimental control belongs to an interlocking material practice, not to a single isolated manipulation.

Such collectivity can amplify reliability through independent work, standardized procedures, and diverse expertise. It can also conceal shared error when techniques, instruments, and background models are inherited without renewed scrutiny. Interlaboratory agreement is strongest when it includes meaningful methodological variation and transparent access to failure modes. What should be counted is the degree of independent exposure to error, not merely the number of institutions repeating an expected result.

From an ontological standpoint, a community's ability to build devices around a target establishes an important body of causal knowledge. The transition from technical reliance to existential commitment still requires an audit of the dependencies that make the devices work. Successful technology may constrain ontology very tightly when many alternative accounts have been tested; it can remain ontologically open where incompatible explanations produce indistinguishable results. Experimental labor is an epistemic achievement in either case, but the warranted entity description is not always equally specific.

Simulation, Computation, and Physical Control

Computer simulation complicates the vocabulary of intervention because manipulating a variable in code is not identical to physically manipulating the system the model depicts. Paul Humphreys's (2004) Extending Ourselves examines the epistemology of computational science and the extension of scientific capacities through computational methods. Computational exploration can identify consequences of assumptions, reveal sensitivity, and suggest experimental tests, even when no new physical interaction with the alleged entity occurs.

An entity realist must distinguish at least three situations. First, a researcher manipulates variables inside a simulation and observes outputs implied by the simulation's formal structure. Second, a computational system adjusts the settings of an actual experiment, after which physical sensors record changes. Third, a computational model interprets the resulting physical data. Only the second case unambiguously adds physical intervention; the first and third can contribute indispensable but different kinds of epistemic work.

Consider a hypothetical simulation in which changing the modeled charge of a particle alters a detector's predicted output. The calculation demonstrates a dependency within the model. It does not independently demonstrate that the model's particle exists. If an automated experimental platform then performs the corresponding field variations and the measured physical response discriminates competing mechanisms, the evidential situation changes. The philosophical issue is the provenance and independence of the resulting constraint on reality.

Artificially Mediated Inquiry

Research increasingly employs artificial systems to recommend experiments, adjust instruments, classify traces, or construct candidate physical models. The existence of these methods broadens the philosophical relevance of intervention without changing the requirement that an ontological claim be supported by physically discriminating evidence. Hacking's (1983) emphasis on experimental practice remains applicable wherever an intervention is physically enacted and its causal outcome becomes evidence.

When an automated controller manipulates a laboratory process, human and artificial contributions may be distributed across design, hypothesis selection, execution, data reduction, and review. The resulting warranted claim depends on calibration, causal identification, independence from circular training assumptions, and error testing. An autonomous system could repeatedly optimize a measured outcome while exploiting an unnoticed artifact. Equally, a well-validated controller can improve the range and precision of feasible interventions.

The broader Artificial Era invites a careful distinction between the bearer or organizer of inquiry and the ontological target of inquiry. Whether a research process is performed by a human operator, an automated platform, or a coordinated artificial system has consequences for provenance, accountability, and the interpretation of results. The physical existence of a proposed entity is still evaluated through evidence about its interaction with the world. This discussion is an application of the philosophy of experiment to emerging research practices; no empirical priority, specific achieved scientific discovery, or new ontological framework is asserted.

Three Comparative Cases

The first case is a constructed electron-beam investigation. A researcher varies field strength and geometry, observes a systematic change in downstream interactions, and reproduces the result with substantially different sensing methods. Hacking's (1982) argument from experimental use explains why the practical employment of an alleged unobservable becomes a realist consideration. The conclusion is strongest if the evidence specifically tracks a class of charged carriers, while independent measurements substantiate that this class is the same across experiments. The hypothetical case illustrates the argument; it is not a report of a new experiment.

The second case concerns a calibrated imaging artifact. Suppose several visually convincing features disappear when an image-processing step is changed. Hacking's (1981) microscope discussion shows the relevance of understanding instrument-mediated observation. Here successful control over the instrument does not favor realism about the apparent specimen features. On the contrary, a discriminating intervention identifies a source of error and retracts an unwarranted entity claim. The success of experiment is compatible with negative ontological conclusions.

The third case concerns an automated simulation-guided experiment. The computational model predicts that either of two different microscopic mechanisms produces the observed signal. A robotic platform conducts a physical intervention designed to make the candidates diverge; the data then fit one mechanism in a well-calibrated domain. Humphreys's (2004) study of computational science clarifies the methodological role of the model, while the physical intervention supplies a distinct causal test. This case illustrates how a computationally designed experiment can improve ontological discrimination without turning simulated outputs into direct evidence for physical existence.

Taken together, the cases show that intervention is epistemically valuable through its power to distinguish possibilities. It can confirm some existence claims, disconfirm others, and leave the remaining alternatives unresolved.

What Experiment Can Establish

A well-designed experiment can establish reproducible causal dependencies under specified conditions, identify the effects of varying certain parameters, and constrain explanations of the resulting phenomena. In strong cases it can also justify belief that a particular unseen causal factor exists. The success of experimental inference, however, is sensitive to the specificity of the target, the range of alternatives tested, the calibration of the apparatus, and the independence of the available evidence. Mayo's (1996) error-testing approach emphasizes this dependence of warrant on serious opportunities to uncover error.

Ontological claims can be arrayed from modest to demanding. The minimal conclusion is that some stable causal interaction occurs. A stronger conclusion is that the intervention involves an entity with a specified set of causally detectable characteristics. Stronger still is the claim that the entity has a particular microstructure and persists as a member of a given theoretical kind. Each increase in specificity demands additional discrimination. The labels assigned by an accepted model do not themselves add that evidence.

For this reason, entity realism is most persuasive as a disciplined selective practice. It connects justified belief with what investigators can physically do, but its ontology should track the actual depth of experimental resolution. If alternatives remain compatible with all interventions performed, the unresolved possibility should be represented explicitly rather than silently removed from the conclusion. This result makes experimental realism robust under fallibilism.

The Limits of the Entity Realist Inference

The inference from manipulation to existence encounters four recurrent limits. First, causal mediation permits a device to respond reliably while scientists misidentify the internal mechanism. Second, theories contribute indispensable descriptions that may be vulnerable to future revision. Third, the same experimental regularity can instantiate different causal structures under distinct background conditions. Fourth, repeated technological success may reflect shared instruments or assumptions rather than independent confirmation. Resnik's 1994 critique and Morrison's discussion of intervention give these worries their proper historical place.

These limits do not imply that interventions are epistemically idle. On the contrary, experiments are especially capable of revealing weaknesses in causal attribution. A clever control may turn what looked like confirmation into an artifact diagnosis; a new sensor can break a tie between competing mechanisms; a stable response under surprising conditions can force a radical reconstruction of the underlying theory.

The important philosophical correction is that the expression 'we can manipulate it' combines several propositions that should be separately defended. Scientists have manipulated an apparatus; they have produced a particular outcome; they have identified a cause of the outcome; and they have established that the cause is the very entity specified by an accepted ontology. These propositions often support one another but have different evidential conditions. A claim of entity realism becomes convincing when their connection has been established, not when they are treated as interchangeable.

A Critical Synthesis

The dispute is commonly staged as a choice between accepting theoretical objects and withholding belief because theories change. The literature supports a more articulate analysis. Hacking (1983) demonstrates the epistemic significance of experimentally using putative entities as instruments. Cartwright (1983; 1989) clarifies the importance of causally localized knowledge. Morrison (1990) and Resnik (1994) expose dependencies on representation. Clarke (2001) and Chakravartty (1998) examine more selective commitments, while Nanay (2019) challenges the stability of neutrality toward theoretical content. The contemporary realism survey maps these positions within the wider dispute.

The original interpretive conclusion of this monograph is that successful manipulation has its greatest ontological force when intervention, causal identification, and referential discrimination are supported together. Experimental success first provides warrant for a causal regularity. Carefully varied tests can establish that a stable source is responsible. Convergence across independent methods can support identification of a particular type of source. Further experiments may justify additional properties. No step requires acceptance of every proposition in the theory that initiated the investigation.

This synthesis explains both why experiment matters and why criticism of naïve entity realism is warranted. Interventions can fix the existence of particular entities more securely than global theory success alone, while their reach remains local, historically revisable, and constrained by identification evidence. An entity realist can therefore defend selective belief through rigorous experimental practice without claiming absolute certainty or theory-free cognition. This position is presented as an analytic reconstruction and judgment within the established philosophical debate.

Source and Methodological Limitations

The primary-source research for this dossier combines open reference scholarship, accessible scholarly articles, original publisher records, bibliographic indexing, and publisher abstracts for subscription-controlled works. Nanay's (2019) full open-access critique provides directly examinable sustained argumentation. Cambridge, Oxford, Springer, JSTOR, and journal pages confirm the existence and bibliographic identities of the other selected works. In some cases only official abstracts or excerpts are available in the research session. Those sources support appropriately limited attribution and do not authorize claims to exhaustive full-text coverage.

Cartwright's Nature's Capacities and Their Measurement was first published in 1989, while the Oxford Academic record carries a 1994 online-edition date. The manuscript consistently cites the historical first-publication year and records the edition distinction in the bibliography. Resnik's original article appeared in 1994 despite a contemporary Cambridge platform date; Nanay's article was online in 2016 and appeared in a 2019 journal issue. These distinctions matter for accurate historical attribution.

No new experiments, laboratory data, quantitative meta-analysis, or systematic bibliometric review were conducted. The three comparative cases are explicitly constructed philosophical examples, not independent empirical findings. The conclusions concern the logical scope and evidential conditions of existing experiments and arguments. Further historically specific laboratory analysis could change judgments about individual cases without requiring revision of the central distinction between causal control and successful identification.

Conclusion

Entity realism directs philosophy of science toward a distinctive source of knowledge: the ability to use a proposed entity as a causally effective resource in experiments. Its historical importance lies in the shift from defending the truth of broad theories through predictive success to examining what experimentalists can establish through intervention. Hacking's experimental realism and Cartwright's causal realism remain powerful because both recognize that scientific inquiry often earns specific beliefs in the midst of theoretical disagreement.

The examination of objections establishes the limits of that achievement. A reproducible effect does not automatically identify its hidden cause. A successful manipulation can depend on theoretical assumptions, detector models, computational processing, and shared technical infrastructures. Consequently experimental success supports ontological claims at different degrees of specificity. The existence of a causal factor, its stable identification as a particular entity, and the acceptance of an extensive account of its nature are distinct questions.

The substantive conclusion is that intervention can justify realist belief in particular theoretical entities when experiments independently identify the causally effective target, discriminate it from consequential alternatives, and withstand serious tests of error. Where identification remains underdetermined, the warranted conclusion is narrower than the preferred theoretical description. This assessment secures the legitimate insight of entity realism while preserving the epistemic discipline on which its strength depends. The decisive value of intervention is its capacity to impose new and discriminating constraints on claims about the world.