Contribution
It collects the realist responses to string- and QFT-dualities into a single diagram (Fig. 1, p. 7), with two consequences. The “discrimination” and “common core” strategies, previously listed as rivals ([35, 48]), turn out not to be disjoint. And each of the familiar strategies turns out to need an extra metaphysical premise to break underdetermination. It also names two alternatives to the common core that the philosophical literature had not separated out: the physicists’ “overarching theory” move (§8.1), and a new, admittedly strange “pluralism” (§8.2).
Key insight
Interpreting a duality is two separate choices. The first is which of the naively interpreted dual worlds (and the common-core world) you allow to be candidates for actuality. The second is whether the common core, a deeper theory or a union of all duals is the replacement ontology. Most confusions in the literature, the paper argues, come from running these together. For example, adopting the common core only adds another empirically adequate candidate world unless a principle such as the PII also removes the duals.
Assumptions
- Duality: a map between the solution spaces of two theories under which related models are empirically equivalent, i.e. they agree on “empirical substructures” in van Fraassen’s sense (p. 3). The map may be many-one (fn. 4).
- Naive interpretation: the ontological claims of a solution are read so that the world it represents is “isomorphic” to it (fn. 8, p. 5). This is limited to “internal interpretations” in Dewar’s and De Haro’s sense. The worry that world–structure isomorphism is a category error is set aside.
- “Actual world”: used non-standardly for a hypothetical world in which the data compatible with both duals are observed (fn. 9).
- Realism: antirealism (positivist and constructive empiricist) and structuralism, including OSR, are set aside (§4). The paper asks what the world is “really like” in extant ontological categories.
- Common core: assumed to be constructible in a well-defined way, citing De Haro & Butterfield’s “bare theory” ⟨S, Q, D⟩ (fn. 13, pp. 7–8).
Key results
- Taxonomy (§3, Fig. 1, pp. 7–9). Duals M1,…,M5 and common core Mc are given, each world coloured candidate or non-candidate:
- (1) all duals and Wc candidates, and (2) all duals but not Wc: underdetermination;
- (3) one dual and Wc, and (4) one dual only: discrimination;
- (5) Wc only, and (6) nothing: nihilism.
- The common-core approach is (1), (3) and (5), so discrimination and common core overlap in (3).
- (3) and (4) are shown to be distinct by a Newtonian absolute-velocity analogy, where Maudlin’s PSR argument favours rest (fn. 16, pp. 8–9).
- Underdetermination (§5). Moving from (2) to (1) by adding the common core “has made the situation worse” (p. 12). Super-empirical virtues and Occam’s razor are “first and foremost” practical and do not remove candidate worlds. The PII would take (1) to (5), but “why should we commit ourselves to such a metaphysical principle?” (p. 13).
- Discrimination (§6). Horowitz & Polchinski’s preference for the CFT side can be defended if read as privileging “those physical theories which we actually have to hand” rather than those we understand better, since the AdS side is only perturbatively constructed (pp. 13–14). The PII plus gauge redundancy in one dual (a many-one duality) supports preferring the other dual’s ontology (p. 15). This is an intra-theoretic use of the PII, contrasted with the inter-theoretic use in §5, and “both are in principle legitimate”.
- Nihilism (§7). It is compatible with realism if the duals are non-final, as Newtonian mechanics cannot account for the stability of matter, or if one rejects the programme wholesale on antecedent grounds (pp. 15–16).
- Overarching theory (§8.1). If the duals’ solutions simply are solutions of the deeper theory T̃, T̃ adds no distinct ontology and underdetermination remains. If T̃ does offer a distinct ontology, a further principle is still needed, and the PII need not deliver it because T̃ need not have less structure (pp. 17–18). An explanatory-power argument is doubted by analogy with shape dynamics vs GR (p. 19).
- Pluralism (§8.2). Every dual structure is instantiated as a numerically distinct part of one “fragmented” world (p. 19). There are two options for the overdetermination problem: (a) only one structure produces the data, which relocates the underdetermination; or (b) all are needed, which is doubted via the gauge analogy (pp. 20–21). Pluralism is judged “prima facie consistent” and in need of further study.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | The realist options for dualities are exhausted, as far as candidate status goes, by the six cases of Fig. 1 | moderate | combinatorial construction, informal; it is exhaustive over the scheme it sets up, but it presupposes naive internal interpretation |
| C2 | The discriminatory and common-core approaches are not disjoint, since option (3) is both | strong | follows from the taxonomy’s definitions; illustrated by the absolute-velocity analogy |
| C3 | Adopting the common core does not by itself break underdetermination; a metaphysical principle such as the PII is required | moderate | informal argument (§5) |
| C4 | Super-empirical virtues and Occam’s razor do not exclude candidate worlds | moderate | informal argument citing van Fraassen §4.1 on the virtue–truth gap |
| C5 | Discrimination is reasonable for AdS/CFT because only the CFT side is fully constructed | weak–moderate | informal argument, credited in part to a referee (fn. 21); the authors say it applies only to dualities where one side is incomplete |
| C6 | Embedding duals in an overarching theory does not resolve underdetermination without a further principle, and the PII may not supply it | moderate | informal argument (§8.1) |
| C7 | Pluralism is coherent but faces an overdetermination problem it has not answered | weak | informal argument by analogy with U(1) gauge copies; the “ontological problem” is named but not discussed |
| C8 | Several options in the taxonomy are “novel to the literature” (abstract) | moderate | the paper’s own reading of [35, 48]; its survey of that literature is short |
Concepts
- Duality — a map between solution spaces relating empirically equivalent models (weaker than De Haro & Butterfield’s).
- Common core (Mc, Wc) — the mathematical structure shared by the dual solutions, and the world it naively represents.
- Discrimination — privileging exactly one dual’s ontology.
- Nihilism — no dual solution legitimately describes the actual world.
- Overarching theory (T̃) — a deeper theory into whose solution space the duals embed, e.g. M-theory, or the unification of matrix and wave mechanics.
- Pluralism — all dual structures co-instantiated as numerically distinct parts of one fragmented world.
- Legitimate candidate — a world not discounted on super-empirical or philosophical grounds (fn. 9).
Connections
It builds on Matsubara (2013) and Read (2016), which first listed discrimination and common core as the main realist options. It takes its notion of common core from De Haro & Butterfield’s Schema and “bare theory” (LIT-199). Read closely, that Schema defines duality by isomorphism of the bare theory: a stronger notion than this paper’s empirical equivalence, which the authors acknowledge in fn. 4. De Haro & Butterfield’s later account of “internal interpretation”, and their conclusion that duality-related models are physically equivalent only when interpretations commute, amount to a principled common-core position with an argued route to case (5). This paper leaves that route open (§5, “why should we commit ourselves…?”).
On structural realism, it sets OSR aside rather than weighing it. It cites Ladyman’s SEP entry (LIT-045) and French among its structuralism references, but does not argue whether OSR dissolves the problem. Its note that the common-core approach is “sometimes identified with structural realism” but is distinct from both versions of structuralism (fn. 19) bears on the OSR/ISR cluster (LIT-151, LIT-215). Its overdetermination objection to pluralism mirrors the standard treatment of gauge redundancy, the same move Belot’s chapter makes in treating reduction as removing “physically otiose” variables (LIT-122).
The metaphysical thesis it presupposes is realism about a unique actual world describable in “extant ontological categories”. Its question, which of several structurally equivalent worlds is actual, is a question of metaphysics conditioned by physics, and it deploys the PII, the PSR, “multiple grounding” (Dasgupta, fn. 29) and fragmentation. The metaphysics tag is justified.
Bearing on the record
It supports no THEORY document directly and would be a good citation for any record entry that needs a map of interpretive options for dualities or symmetry-related models. The dossier’s link to LIT-215 (Karpenko’s use of dualities for informational structural realism) should be read with the caveat that this paper sets structuralism aside and so offers ISR no support either way. It carries nothing for ML practice, and nothing here belongs in the Anthology of the SOTA.
Note of 2026-10-16: THEORY-687 now cites LIT-113 as a source.
Limitations
- It surveys options and does not argue for one. The authors’ own sympathies (discrimination for AdS/CFT; both uses of the PII “legitimate”) are stated briefly.
- The empirical-equivalence definition of duality lets in cases, such as theory-level underdetermination, that the stronger isomorphism definitions exclude. Conclusions about “dualities” may not transfer to the Schema’s dualities.
- Pluralism’s “ontological problem” is named but not addressed.
- Naive internal interpretation is assumed throughout. External interpretations, where the duality map does not relate referents, are not considered.
- It is a 2018 survey. It predates De Haro & Butterfield’s monograph and later work on the common core.
Open questions
- Can pluralism answer the overdetermination objection, e.g. via multiple grounding, so that co-instantiated duals account non-redundantly for the data? The authors pose this and a comparison with Pitts on gravitational energy (fn. 33).
- Is there independent support for the PII strong enough to take case (1) to case (5)? That is the paper’s central unmet premise.
- Does the AdS/CFT discrimination argument survive a non-perturbative construction of the bulk side?
Corrections to the seeded skim
- The dossier’s page numbers are the PDF’s, which run one ahead of the manuscript’s printed numbers. For example, §3 Taxonomy is printed pp. 7–9 (PDF 8–10), and §8 is printed pp. 16–21.
- The definition of duality is deliberately weaker than De Haro & Butterfield’s. It requires only a map between solution spaces under which related models are empirically equivalent. The authors choose this over the stronger requirement that all physically meaningful quantities be preserved, and they allow the map to be many-one (fn. 4, p. 3). The dossier’s “formal map between solution spaces of two empirically equivalent theories” is right, but the contrast with the Schema (LIT-199) matters for comparing the two.
- §6 is not only “discrimination defended via PII plus gauge redundancy”. Its main new argument is a reading of Horowitz & Polchinski on AdS/CFT (pp. 13–14). Discrimination is justified not because the CFT is better understood, but because it is the only side whose full mathematical structure is “to hand”, since the bulk string theory is known only perturbatively. The PII-plus-gauge-redundancy argument (p. 15) is a second, intra- vs inter-theoretic use of the PII.
- §7 (nihilism, pp. 15–16) is missing from the dossier’s bullets. It gives two routes compatible with realism: dual theories may be non-final, or one may reject a whole research programme, such as string theory, on antecedent grounds.
- The dossier asks how §4 treats OSR’s claim to dissolve the problem. It does not engage it. Structural realism and OSR are “set aside” (§4, pp. 10–11; fn. 18), because the authors want to say what the world is like in “extant ontological categories”. The common-core approach is explicitly distinguished from both versions of structuralism (fn. 19, p. 12).
- Pluralism faces two named problems, overdetermination and an “ontological problem” (what a fragmented world is). Only the first is discussed (pp. 19–21). Its analogy is gauge-related vector potentials, where co-instantiating every gauge copy is “metaphysically otiose” (p. 21).
- metaphysics tag: justified. The paper’s question is which worlds exist or are actual given dual theories, and it invokes the PII, PSR, grounding and fragmentation.
- primary topic: philosophy-of-science (keep as listed). The problem is set as underdetermination for the scientific realist, and metaphysics is the right second tag.