Abstract
Relations between chemistry and biology, and between biology and psychology, are often framed through reduction, realization, emergence, and explanatory autonomy. Yet both philosophy of science and scientific practice suggest that these relations are rarely governed by one final, globally sufficient semantics. Distinct scientific aims tend to stabilize distinct formal readings of the same domain. This article argues that the special sciences are therefore better understood as exhibiting disciplined semantic pluralism than as converging on one complete semantic foundation. On that basis, meta-rational pragmatics is introduced as a methodological layer for selecting, composing, and validating bounded interpretive skills. These skills are local semantic operators: explicit ways of reading theories, models, datasets, and explanations under task-specific constraints. Their scientific value lies not in completeness, but in validity, tractability, and controlled usefulness. The broader implication is that formalizing a science need not mean axiomatizing it absolutely. It can also mean rendering parts of its interpretive labor explicit enough to be represented, executed, reviewed, and partially verified in a computational setting [STRUCTURE-SEP] [PLURALISM-SEP].
1. Introduction
The classical literature on scientific theory distinguishes syntactic, semantic, and pragmatic conceptions of theory structure. That distinction remains important because it clarifies that scientific understanding may be reconstructed through formal sentences, mathematical models, or situated inferential and experimental practice [STRUCTURE-SEP]. For the present problem, however, the key point is that many mature sciences do not behave as if one unique semantic description settles all legitimate reasoning about a domain. Instead, they employ several disciplined representational regimes, each preserving different invariants and supporting different forms of judgment [PLURALISM-SEP].
This matters especially for inter-level relations. Chemistry constrains biology, and biology constrains psychology, but such dependence does not by itself determine one unique scientific meaning for biological or psychological concepts. Reduction in biology itself is standardly analyzed as involving ontological, epistemological, and methodological dimensions, not merely a single logical deduction from one vocabulary to another [REDUCTION-BIOLOGY-SEP]. The same pressure appears in philosophy of mind, where the literature on multiple realizability challenged simple one-to-one identification between higher-level kinds and lower-level realizers, even if the strongest classical versions of the thesis are now treated more cautiously [MULTIPLE-REALIZABILITY-SEP].
The proposal developed here is that meta-rational pragmatics becomes useful precisely under such conditions. It is not a denial of semantics. It is a way of handling cases in which scientific semantics is local, purpose-indexed, and only partially transportable across frameworks. More concretely, it treats these local semantic regimes as bounded interpretive skills that can be attached to AI agents and used to judge scientific artefacts under explicit criteria. The resulting aim is not metaphysical finality, but executable scientific discipline.
2. Why one final semantics is the wrong target
A strong reductionist hope would be that once the lower-level substrate is fixed, the higher-level science becomes semantically derivative in a complete sense. But that is not how the most influential contemporary accounts describe the situation. In biology, lower-level realization does not eliminate the methodological importance of higher-level concepts, explanations, or models [REDUCTION-BIOLOGY-SEP]. In the philosophy of mechanisms, explanation is often organized around entities and activities producing phenomena, and this framework is explicitly used to address relations between disciplines and levels without collapsing all scientific content into a single reductive vocabulary [MECHANISMS-SEP].
The resulting picture is neither strict monism nor vague relativism. It is closer to a constrained pluralism in which multiple formal readings of the same domain are legitimate because they answer different questions. Scientific pluralism, as surveyed in the Stanford Encyclopedia of Philosophy, is precisely a family of views that studies science in its synchronic and diachronic complexity rather than assuming that unification is always the correct regulative endpoint [PLURALISM-SEP]. That stance is especially compelling in the special sciences, where explanatory success often depends on selecting the right representational regime rather than on forcing all phenomena into one universal description.
This is the point where meta-rational pragmatics should be positioned more sharply. It should not be introduced merely as the claim that “there are many semantics.” That formulation is too loose and too easy to attack. A stronger formulation is that the special sciences often contain multiple local semantic regimes, each stabilized by a distinct inferential task, validation style, and computational profile. Meta-rational pragmatics then becomes the methodological layer that identifies, reifies, and coordinates those regimes as explicit interpretive skills.
3. Chemistry and biology: substrate dependence without semantic exhaustion
The chemistry–biology relation is often treated as the best case for reduction. Yet even there, several non-equivalent formalizations remain scientifically useful. Mechanistic explanation provides one important example. The philosophical literature on mechanisms treats explanation as involving organized entities and activities, and this approach has become central partly because it captures forms of biological explanation that are not well described as mere derivation from lower-level laws [MECHANISMS-SEP].
The concept of the gene offers a second example. The SEP entry on the gene explicitly discusses diversification in the meaning and reference of the term “gene,” showing that the concept does not function as a single stable semantic unit across all scientific contexts [GENE-SEP]. This is already enough to weaken any naive expectation that biology inherits one straightforward semantics from chemistry. Scientific practice instead shifts among conceptions depending on explanatory and experimental needs.
A third example is Chemical Organisation Theory. Dittrich and Speroni di Fenizio define a chemical organisation as a closed and self-maintaining set of components, thereby formalizing persistence and organizational viability in reaction systems [DITTRICH-2007]. This is not simply “more chemistry.” It is a distinct formal reading that foregrounds closure and self-maintenance rather than, for example, local reaction enumeration. The important methodological lesson is that two mathematically explicit descriptions of the same domain may preserve different invariants and thereby support different uses.
From an MRP perspective, these become distinct interpretive skills. A mechanistic skill asks how organized parts and activities produce a phenomenon. An organizational skill asks whether a reaction network forms a closed and self-maintaining regime. An ontology-oriented skill asks whether entities and relations can be rendered interoperable and computationally tractable. None of these needs to represent the whole domain completely. Their legitimacy comes from bounded validity.
4. Biology and psychology: multiple frameworks, bounded validity
The case of biology and psychology makes the point even more strongly. Cognitive science has long resisted the idea that one framework alone should dominate all explanation. Marr’s framework became influential precisely because it distinguished different explanatory questions rather than reducing them all to one level [MARR-2010]. Even when particular readings of Marr are debated, the enduring lesson is that explanatory adequacy is question-relative.
This conclusion is reinforced by the literature on explanatory pluralism in cognitive science. Dale, Dietrich, and Chemero argue that framework debate in cognitive science is unresolvable in the strong sense that no single framework is likely to capture the full complexity of cognitive phenomena [DALE-2009]. That does not mean that anything goes. It means that different frameworks may be appropriate to different explanatory purposes.
Here again MRP can be made more concrete than in the earlier draft. A computational skill may judge a model in terms of task decomposition and representational function. A mechanistic skill may judge the same case in terms of organized bodily or neural realization. A dynamical skill may judge it in terms of trajectories, coupling, and stability. A clinical skill may judge whether distinctions are robust enough for diagnosis or intervention. These skills need not be semantically identical. Their practical value lies in the fact that they offer valid judgments over restricted classes of questions.
This is where the diversity of judgment becomes methodologically productive. We do not need completeness if the skill produces valid and inspectable results for the target task. In other words, semantic incompleteness is often acceptable, and sometimes desirable, when it supports better tractability and clearer validation boundaries.
5. Interpretive skills as executable local semantics
The central move of meta-rational pragmatics is therefore to treat local semantic regimes as explicit interpretive skills. A skill is not merely a prompt pattern or a loose heuristic. It is a bounded inferential role attached to an artefact. It specifies what distinctions matter, what evidence counts, what kind of output is expected, and what counts as partial success or failure.
This reframing matters because it makes formalization operational. Instead of searching first for one final semantics of a science, one can construct a family of bounded semantic operators and ask whether they are explicit enough to support review, comparison, consistency checks, structured criticism, or partial proof. In such a framework, an AI agent does not “understand biology” or “understand psychology” in one absolute sense. It possesses a set of scientific reading skills, each tuned to a narrower class of judgments.
That is the deeper sense in which one may speak of meta-rational semantics. It is not a single model-theoretic semantics replacing all others. It is a controlled ecology of executable interpretive competences. The methodological gain is that scientific reasoning becomes more representable: not because ultimate meaning has been captured once and for all, but because some of the interpretive labor has been decomposed into reusable, inspectable units.
6. Efficiency is not secondary
The earlier draft was right to mention efficiency, but the point should be stated more strongly. Efficiency is not merely an implementation detail added after philosophical analysis. It is one of the reasons pluralism is rational.
Different interpretive skills impose different representational restrictions. Some are narrower, more compressed, and easier to run. Others are richer, slower, and harder to validate. A skill may therefore be preferable not because it is more complete, but because it is sufficiently valid and computationally advantageous for a class of problems. In practice, that can matter more than representational maximality.
Biomedical ontology work is a clear case. The OBO Foundry was explicitly developed to support interoperability and logical well-formedness across biomedical ontologies [OBO-2007]. Such ontologies do not attempt to encode all of biology. Their strength comes precisely from selective formal restriction. Likewise, the Gene Ontology is valuable because it offers a controlled, machine-readable representation of certain biologically relevant relations, not because it is the final semantics of living systems [GENE-ONTOLOGY]. This is a good model for MRP: bounded formalisms can be scientifically powerful because their restrictions make them operational.
The same argument extends to theory review. A restricted proof procedure, a local consistency check, a bounded ontology alignment, or a framework-specific criticism may establish only partial results. But partial theoretical demonstration is often pragmatically valuable. It can expose contradictions, certify compatibility under a given regime, or reject a class of invalid inferences. In an MRP setting, such bounded demonstrations are not second-best imitations of absolute proof. They are practical instruments of scientific control.
7. What MRP is trying to formalize
The most defensible way to state the ambition of meta-rational pragmatics is therefore this: it aims to provide a framework for the controlled normalization of scientific interpretation. Its purpose is not to discover one ultimate foundation of biology or psychology. It is to make heterogeneous scientific judgments more explicit, more comparable, and more executable.
That aim includes at least four concrete functions. First, MRP supports review of scientific results by allowing the same claim to be interrogated by different interpretive skills. Second, it supports selective partial proof or consistency assessment, even when full formalization is unattainable. Third, it supports comparison between incompatible or only partially compatible foundations. Fourth, it supports computational efficiency by routing different classes of questions to skills whose representational burdens are appropriate to them.
This, in turn, suggests a revised conception of formalization. To formalize a science is not always to axiomatize it completely. Often it is to render enough of its interpretive practice explicit that parts of it can be represented, executed, checked, and reused. On that understanding, MRP is not an anti-semantic proposal. It is a proposal for executable semantic pluralism.
8. Conclusion
The special sciences do possess semantics, but often not one final semantics that is complete, unique, and computationally practical across all purposes. Chemistry constrains biology, and biology constrains psychology, yet these relations do not erase the need for multiple formal viewpoints. Mechanistic, organizational, ontological, computational, and other readings may all be scientifically legitimate while remaining only partially compatible [MECHANISMS-SEP] [PLURALISM-SEP].
Meta-rational pragmatics becomes valuable exactly here. It treats the plurality of scientific meaning as a space of bounded interpretive skills. These skills are selected not because they are semantically total, but because they are valid enough, explicit enough, and efficient enough to support real scientific work. Their restrictions are not merely limitations. They are often what make execution, review, and partial theoretical control possible. In that sense, the most defensible formulation is this: meta-rational pragmatics is a methodological framework for turning semantic pluralism in the special sciences into executable, inspectable, and task-sensitive scientific reasoning.
References
- [DALE-2009] Dale, R., Dietrich, E., Chemero, A. Explanatory Pluralism in Cognitive Science. Cognitive Science, 2009. Link: https://onlinelibrary.wiley.com/doi/10.1111/j.1551-6709.2009.01042.x
- [DITTRICH-2007] Dittrich, P., Speroni di Fenizio, P. Chemical Organisation Theory. Bulletin of Mathematical Biology, 2007. Link: https://link.springer.com/article/10.1007/s11538-006-9130-8
- [GENE-ONTOLOGY] The Gene Ontology Consortium. Gene Ontology Resource. Link: https://geneontology.org/
(Official resource; included here as the primary project site.) - [GENE-SEP] Griffiths, P. E., Stotz, K. Gene. Stanford Encyclopedia of Philosophy, 2022. Link: https://plato.stanford.edu/entries/gene/
- [MARR-2010] Shagrir, O. Marr on Computational-Level Theories. Philosophy of Science, 2010. Link: https://www.cambridge.org/core/journals/philosophy-of-science/article/marr-on-computationallevel-theories/21C3C13FB31C51228AAD3D086E273990
(Used conceptually; the core point about Marr’s levels is standard in the literature.) - [MECHANISMS-SEP] Craver, C. Mechanisms in Science. Stanford Encyclopedia of Philosophy, 2015. Link: https://plato.stanford.edu/entries/science-mechanisms/
- [MULTIPLE-REALIZABILITY-SEP] Bickle, J. Multiple Realizability. Stanford Encyclopedia of Philosophy. Link: https://plato.stanford.edu/entries/multiple-realizability/
(Referenced for the general philosophical debate; current SEP entry URL.) - [OBO-2007] Smith, B., Ashburner, M., Rosse, C., et al. The OBO Foundry: coordinated evolution of ontologies to support biomedical data integration. Nature Biotechnology, 2007. Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC2814061/
- [PLURALISM-SEP] Ludwig, D., et al. Scientific Pluralism. Stanford Encyclopedia of Philosophy, 2021. Link: https://plato.stanford.edu/entries/scientific-pluralism/
- [REDUCTION-BIOLOGY-SEP] Brigandt, I. Reductionism in Biology. Stanford Encyclopedia of Philosophy, 2008. Link: https://plato.stanford.edu/entries/reduction-biology/
- [STRUCTURE-SEP] Winther, R. G. The Structure of Scientific Theories. Stanford Encyclopedia of Philosophy, 2015. Link: https://plato.stanford.edu/entries/structure-scientific-theories/