3.1. Truth-Enough

Abstractions, approximations, and idealizations are all thought to provide understanding while being false. A successor concept, such as Elgin’s “truth-enough” or Potochnik’s “representation as-if,” is frequently proposed in truth’s stead. Propositions are true enough just in case they are close enough to the truth to achieve one’s epistemic goals. For this to undercut veritism, true-enough propositions could be neither true answers to relevant questions nor a means to acquiring such answers. However, true-enough propositions appear capable of serving either role. Thus, they pose no challenge to veritism.Footnote ⁶

Begin with abstractions, in which certain truths are omitted. Precisely what gets abstracted appears to be a function of which questions are relevant. For instance, Schelling’s (Reference Schelling1971) model of segregation omits organizational practices and specialized communication systems as causes of segregation in order to answer the highly circumscribed question, “How could segregation result from the interaction of individual choices?” Moreover, since Schelling is only looking for a possible (rather than an actual) way for segregation to occur via individual choices, his model provides a true answer to this question.Footnote ⁷ Thus, models with abstractions can still provide true answers to relevant questions. Consequently, abstractions do not threaten veritism.

Next, consider approximations, in which something is close to the truth but not perfectly accurate. Here, we should distinguish between (i) the true claim that approximately p and (ii) the false but approximately true claim that p. Since the former is true and not merely true enough, veritism allows it to serve as an answer to a question. Moreover, questions often dictate suitable standards of approximation. For example, in their empirical evaluation of Schelling’s hypothesis, Bobo and Zubrinsky (Reference Bobo and Zubrinsky1996) appear to be answering questions such as, “Controlling for age, occupation, education, and sex, how strongly do whites’ in-group preferences correlate with their preferences to live in segregated neighborhoods with $p\le .05$?”

Indeed, this provides the ingredients for a veritistic treatment of false but approximately true claims. In short, scientists are actually communicating (i) true claims about approximations although they are uttering (ii) false but approximately true claims. Since the former pose no threat to veritism, neither do the latter. For instance, in Bobo and Zubrinsky’s question, the aforementioned demographic variables are typical controls in sociological research; the .05 p-value is the default scientific convention. Hence, as bits of common knowledge, these standards of approximation are frequently left implicit. In such contexts, utterances of the form “p” typically express that p falls within the accepted standards of approximation. Hence, although what is said is (strictly speaking) false, what is communicated is a true claim about an approximation. Thus, even the “tough” approximations conform with veritism.

While both abstractions and approximations offer true answers to relevant questions, idealizations cannot, for they are not even approximately true. For instance, the ideal gas law is

${\mathrm{pV}}_m=\mathrm{RT}\mathrm{.}$

Here, p, V_m, and T denote a gas’ pressure, molar volume, and temperature, respectively, and R is the ideal gas constant. Some statistical-mechanical models of this law entail, contrary to fact, that particles in ideal gases do not interact. This idealization enhances scientists’ understanding of the ideal gas law.Footnote ⁸

Inquisitive veritism treats idealizations as false answers to irrelevant questions. For instance, in the aforementioned derivations of the ideal gas law, scientists are not seeking to answer, “Do particles in ideal gases interact?” If they were, then they would be sorely misled.Footnote ⁹ Nevertheless, idealizations are epistemically valuable because they are a means to arriving at true answers to other, relevant questions. For instance, the aforementioned derivation seeks to answer, for example, “Why does the ideal gas law hold (at low pressure and high temperature)?” As several authors (e.g., Strevens Reference Strevens2008) have noted, whether particles interact is no part of the answer to this latter question. Rather, the answer resides in the partition function—which is true.Footnote ¹⁰ The idealization earns its keep by making the partition function’s relevant consequences—concerning the statistical mechanics underlying the ideal gas law—readily apparent. Thus, according to inquisitive veritism, the idealization is of instrumental epistemic value. Importantly, this account of idealizations only requires answers to be true or false and only requires questions to be relevant or irrelevant. Truth-enough is otiose.

Thus, the main reasons for rejecting veritism in favor of an axiology predicated on truth-enough are found wanting. We can do just as well by recognizing that scientists do not deem every question about a given phenomenon relevant. Epistemically valuable abstractions, approximations, and idealizations depart from the truth only insofar as they answer irrelevant questions. Hence, they are compatible with veritism, and do not support U1.

3.2. Empirical Adequacy

Sometimes, scientists forgo accuracy in order to increase understanding. However, requiring some parts of understanding to be accurate appears to be a good way of distinguishing genuine scientific understanding from the pseudo-understanding afforded by crackpot theories. For this reason, some claim that empirical adequacy, not truth, fittingly fastens understanding to the facts (De Regt Reference De Regt2017; Rowbottom Reference Rowbottom2019).

Let empirical statements only describe observable entities, properties, and processes. Let empirical questions have empirical statements as true answers. Then, insofar as empirical questions are relevant, veritism accommodates this brand of empiricism. Veritism also mimics its proponents’ flexible notion of empirical adequacy. For instance, De Regt (Reference De Regt2017, 38) writes that “there may be variation in how [empirical adequacy is] ranked and applied in specific cases.” Similarly, veritism suggests that different inquiries will deem different empirical questions relevant and thereby allow for different kinds and degrees of empirical adequacy.Footnote ¹¹

Like empiricism, veritism allows some nonempirical questions to be irrelevant and thereby unsuitable as epistemic goals. However, veritism also allows for some relevant nonempirical questions. As a description of scientific practice, this appears desirable, as many scientists seem genuinely interested, if not professionally obligated, to ascertain whether unobservables (e.g., leptons) exist. It is unclear how empiricists about understanding accommodate this facet of scientific practice. Thus, all told, construing empirical adequacy as encapsulating the idea that some scientific questions are properly answered by empirical statements appears to accommodate all of the virtues of appealing to empirical adequacy and arguably avoids some potential pitfalls of this approach.

3.3. Explanation

Many require tight connections between understanding and explanation (Strevens Reference Strevens2013; De Regt Reference De Regt2017; Khalifa Reference Khalifa2017; Potochnik Reference Potochnik2017). It is generally thought that if A explains B, then both A and B are true. Moreover, explanations are answers to why-questions. All of this accords with veritism.

However, de Regt directly challenges the requirement that correct explanations require true explanantia.Footnote ¹² For him, intelligible theories can yield falsehoods F that nevertheless explain some phenomenon B. Thus, de Regt’s view suggests that F is false but nevertheless correctly answers the question, “Why B?” This appears incompatible with veritism.

I offer two replies to de Regt’s challenge. First, his semantics for explanation is somewhat implausible (Khalifa Reference Khalifa2017, 162–64). If de Regt is correct, then statements of the form “A explains B but not-A” should be acceptable. However, statements such as “the patient’s allergy explains her symptoms, but the patient has no allergies” border on incoherence. Consequently, it is unclear that falsehoods correctly answer why-questions. Indeed, such sentences are more fruitfully regarded as elliptical for true counterfactuals of the form: “if A were true, A would explain B.” Thus, whenever questions about what would explain a phenomenon are relevant, veritism readily captures these statements’ epistemic value.Footnote ¹³

Second, even if explanations tolerate false explanantia, such explanations do not appear epistemically valuable unto themselves. Otherwise, crackpot explanations are difficult to rule out. Hence, epistemically valuable but false explanations must be a means to some other cognitive good. Indeed, de Regt requires explanations to promote empirical adequacy and systematicity. As section 3.2 shows, veritism ably accommodates empirical adequacy; section 3.6 does the same with systematicity. Thus, even if falsehoods explain, then veritism can account for their value. In this way, veritists can still resist de Regt’s understanding objection by denying U1, even while granting his controversial semantics.Footnote ¹⁴

3.4. Inference/Reasoning

It seems possible to provide true answers about a topic without understanding it. Scientists’ opening of various “black box” models illustrates the demand that understanding involve the theoretical and methodological reasons for why something is the correct answer.

However, this poses no problem for inquisitive veritism. Recall that inquiries are networks of interrelated questions and answers. In many scientific inquiries, this networking results from answering one question Q with A and then asking “follow-up” questions about A, for example, “Why A?” or “Why believe that A is a correct answer to Q?”Footnote ¹⁵ Reasons are then answers to these follow-up questions. For instance, Bobo and Zubrinsky claim that in-group preferences negligibly affect preferences to live in racially homogeneous neighborhoods. Why believe this? Because it is the result of a methodologically sound linear regression analysis, with people’s preferences to live in communities of varying levels and kinds of racial diversity as the dependent variables, and “feeling thermometer” results, in which respondents rank the “warmth” of their feelings toward various racial groups (including their own), as one of several independent variables. Furthermore, understanding frequently improves in proportion to the number of follow-up questions correctly answered. Thus, inasmuch as understanding involves the recruitment of reasons, it appears to conform with veritism.

Admittedly, scientists sometimes use falsehoods to infer true answers; the aforementioned idealization in deriving the ideal gas law is an example. According to veritism, scientists should not cite these falsehoods as answers to follow-up questions. However, even here, veritism is unthreatened, for, as already discussed, such falsehoods still function as a means to true conclusions that answer relevant questions even if they cannot serve as true answers unto themselves.Footnote ¹⁶ Hence, as concerns reasoning, U1 is unsubstantiated.

3.5. Nonpropositional Representations

Much of scientific understanding involves nonpropositional representations, for example, models and diagrams. For this reason, some have turned to denotation and other nonpropositional forms of representational success as a way to accommodate understanding-providing falsehoods. Roughly, the thought is that such falsehoods still are sufficiently tethered to the world by denotation, but the added slack afforded by eschewing truth allows them to provide greater understanding.

This stands in tension with veritism, as (nonelliptical) answers to questions are propositional. However, this tension is illusory. Even when construed nonpropositionally, scientific representations are widely thought to support “surrogative inference”; very roughly, competent and informed users can take statements about the representational source as premises and draw deductively valid or inductively warranted conclusions about the target (Suárez Reference Suárez2004). Surrogative inferences are sound when their conclusions are true. Let a scientific representation A of B be successful in context C just in case the sound surrogative inferences that appeal to A in their premises have conclusions that correctly answer every relevant question about B in C.Footnote ¹⁷ Such a formulation requires neither A nor B to be propositions. In this way, nonpropositional representation can be seen as derivative of the surrogative inferences that are consonant with veritism.

3.6. Systematicity

One might think that even if one can provide true answers to relevant questions, one’s understanding suffers if those answers do not cohere in the proper way. Such coherence is partly a function of the explanatory and inferential relations discussed above, but sometimes the claim appears to be that even if two bodies of statements were inferentially and explanatorily articulated to the same degree, they could differ in terms of global properties or theoretical virtues, such as simplicity and scope (De Regt Reference De Regt2017; Elgin Reference Elgin2017). Call a healthy balance of these virtues “systematicity,” and assume that it is an epistemically valuable feature of understanding.

Unfortunately, few theories of understanding unpack systematicity in any detail. I offer one proposal. Should this miss the mark, it behooves veritism’s critics to propose alternatives. To begin, systematicity is not itself an answer to any question. But certain propositions—paradigmatically, invariant generalizations or “law-like statements”—drive this systematicity. If these law-like statements are true, then their suitability for answering questions such as, “What do these seemingly disparate phenomena have in common?” seems to be precisely why the web of commitments comprising an inquirer’s understanding exhibits a certain kind of systematicity. In such cases, veritism clearly bestows epistemic value upon these law-like statements.

Indeed, section 3.1 shows that these law-like statements need only be approximately true in order to accord with veritism in this way. But could law-like statements that fall short of this threshold still secure epistemically valuable kinds of systematicity? Those resistant to veritism sometimes suggest that Newtonian mechanics fits the bill. While that is debatable, let’s grant, if only for the sake of argument, that Newtonian mechanics does not provide an approximately true answer to the question, “What do tidal patterns, planetary orbits, swinging pendula, falling apples, and so on, have in common?” Then, we need some reason to think that Newtonian mechanics is not systematic in the same way as, for example, conspiracy theories. After all, ex hypothesi, it is claiming that various phenomena are related in ways that they are not. Even here, veritism offers a tidy solution: unlike conspiracy theories, classical mechanics’ law-like statements would still be an effective means for making predictions and thus for answering empirical questions (see sec. 3.2). In other words, veritism would still imbue these statements with instrumental epistemic value. Thus, whether a law-like generalization—the engine of systematicity—is true or not, veritism does it justice. Yet again, U1 can be rejected.

4.1. Salience

Potochnik (Reference Potochnik2017, 95) holds that “idealizations contribute to understanding by representing as-if to the end of [i] depicting a causal pattern, thereby [ii] highlighting certain aspects of that phenomenon (to the exclusion of others) and revealing connections with other, possibly disparate phenomena that embody the same pattern or, in some cases that are exceptions to that pattern.” This is plausibly interpreted as assigning only instrumental epistemic value to idealizations. Consider i: the depiction of causal patterns. Strictly speaking, depictions need not be accurate, so this might suggest that Potochnik has provided a genuine alternative to veritism. However, inaccurate causal depictions do not seem to have any epistemic value unto themselves. Presumably, they serve some further end. Indeed, the passage above suggests that Potochnik requires understanding-conferring causal depictions to highlight aspects of phenomena and reveal connections between phenomena.

Hence, if this challenges veritism, then it is because “highlighting and revealing” is an epistemically valuable feature of understanding. Furthermore, such a proposal is plausible: those who understand frequently notice things that others do not. So, let’s consider whether highlighting possesses epistemic value that eludes veritism; “revealing” admits of similar treatment.

First, a small semantic point: the verb highlight is naturally glossed as entailing a kind of accurate representation. If x highlights y, then y exists; if x highlights that p, then p is true. Furthermore, suppose that one’s representation “pseudohighlighted” an aspect of a phenomenon, for example, it vividly represented the phenomenon as having a property that it did not actually have. In isolation, such representations’ epistemic value is mysterious, to say the least.

However, true claims need not be highlighted. So perhaps this remainder is epistemically valuable in a manner that eludes veritism. Unfortunately, Potochnik says little about how to fund the difference between highlighted and nonhighlighted aspects of a phenomenon.Footnote ¹⁹ Following Koralus (Reference Koralus2014), a highlighted truth might simply be one that answers a relevant question. After all, those truths that serve our most pressing inquiries seem especially attention grabbing. However, this proposal dovetails seamlessly with inquisitive veritism.

So, for argument’s sake, let’s propose an alternative on Potochnik’s behalf: that salience captures the difference between highlighted and nonhighlighted truths. The problem is that any conceivable epistemic good or its exact opposite can be made salient. Crucially, in the latter case, our epistemic situation frequently worsens as a result of this salience. (Consider, e.g., the falsehoods and misunderstandings made salient in certain fringe social media bubbles.) This suggests that salience has no inherent epistemic value: everything hinges on what’s being made salient. Consequently, if salience is what puts a truth on the highlight reel, then, contra U2, highlighted truths appear to have no greater epistemic value than nonhighlighted truths. Hence, highlighting has no epistemic value over and above what veritism counsels.

4.2. Skills

Understanding frequently is thought to involve more than the possession of information. It also involves some kind of ability or skill (Newman Reference Newman2012; Wilkenfeld Reference Wilkenfeld2013; De Regt Reference De Regt2017). On the one hand, if such abilities are no more than the capacity to answer relevant questions correctly or to make inferences that ultimately lead to such answers, then we have already seen that veritism readily accounts for their epistemic value. If, on the other hand, the abilities involve more than this, then we are owed an argument as to why they are epistemically valuable. After all, both questions and the reasoning that delivers their answers are quite varied, so it is not obvious that anything of epistemic value eludes their purview. To my knowledge, no such argument has been made. Moreover, abilities accrue much nonepistemic value. For example, being able to do things gives us a sense of accomplishment, makes us employable, increases our prospects of survival, and so on. Perhaps these nonepistemic benefits account for any value that abilities have in excess of the epistemic value captured by veritism. This, however, is cold comfort for purveyors of understanding objections.