Specifying the Method of Reflective Equilibrium: A Methodological Framework

Abstract

This chapter discusses how we can obtain an applicable method of reflective equilibrium based on the theoretical foundations described in the previous chapter. It describes a methodological framework which identifies the tasks one has to complete in order to apply reflective equilibrium, the methodological decisions one has to make in order to be able to address these tasks, and the challenges one faces when making these methodological decisions.

Chapter 2 described the theoretical foundations of reflective equilibrium (RE) as an epistemological theory of justification. How do we get from this theoretical conception to a method? That is, how are we to obtain a method of RE that can be applied by actual researchers and practitioners?

This chapter develops a methodological framework which shows how the RE criteria can be specified, and how guidelines for its application can be defined—that is, how an RE method can be specified. As we saw in the introduction, there are different ideas about the aims and purposes of RE, as well as different ideas about the agent(s) involved in its application. For the purpose of this book, the objective is, firstly, to spell out a method that can be applied by one epistemic agent and does not require group deliberation or similar group processes. (However, this agent is of course not isolated, and needs to consider the inputs and epistemic achievements of others.) Secondly, reflective equilibrium is treated as a method that aims at justifying principles, theories, or similar; and not, for example, for finding a justified consensus, or as a decision framework. These restrictions do not mean that it is not possible to spell out RE methods for other purposes. However, they give the case study a more stringent focus while laying the groundwork for further work on different ways to spell out RE.

With respect to these two restrictions, this chapter develops an RE methodology: it describes guidelines for how RE can be specified and applied as a method, in a way that allows imperfect epistemic agents to apply it without sacrificing methodological rigor. It is worth pointing out that most of the methodological issues that need to be resolved for this purpose are not inherent to reflective equilibrium as a theory of justification (Walden 2013, cf.). For example, the RE epistemology tells us that commitments should have independent credibility, and that to be justified, the resulting commitments and the resulting system should be in agreement. But it stays silent on what, exactly, are sources of independent credibility, and how we can measure whether system and commitments are “in agreement”. Many questions that one needs to address to be able to obtain an applicable method of RE need to be answered by other research fields, and will be part of (epistemological) background theories. Like everything in RE, the answers to these methodological questions are not fixed once and for all. The goal of this chapter is thus not to spell out an RE method in the sense of defining all its elements and criteria as precisely as possible. Instead, it describes the tasks one has to complete in order to apply RE as a method, the methodological decisions one has to make in order to be able to address these tasks, and the challenges one faces when making these methodological decisions. Spelling out the method in detail for its specific application is part of Chap. 5, which describes the design and setup for the RE case study on precautionary principles.

To start, let us reconsider the criteria for an epistemic position to be in a state of reflective equilibrium:

• 1. The resulting commitments and the system are in agreement;

• 2. The resulting commitments and the system are supported by background theories;

• 3. The system does justice to the relevant theoretical virtues;

• 4. The resulting commitments respect the input commitments adequately;

• 5. The resulting commitments have independent credibility; and

• 6. The resulting position is at least as plausible as all available alternatives.

Notably, these criteria concern the resulting position. But how do we achieve such a position in RE? And can these criteria guide us in spelling out rules, or at least guidelines, for how to proceed when adjusting commitments and system?

As described in Chap. 2, the RE process starts from the initial commitments about a subject matter, and then systematically adjusts them with respect to a system (e.g., principles, a theory, or a model), which, in turn, is formulated and adjusted with respect to the commitments. The whole process is guided by theoretical virtues, like simplicity or scope, that put constraints on the system, while commitments have different weights that put constraints on which adjustments are admissible. Additionally, the position in the foreground should be supported by background theories.

Thus, if we want to apply reflective equilibrium, the first stage will consist in clarifying our initial epistemic position, which includes identifying commitments, relevant background theories, theoretical virtues, and candidates for the system (Sect. 3.1). Starting from this initial position, we will enter the stage of adjusting commitments and system alternately (Sect. 3.2), until the process comes to a preliminary end point. In the latter stage, we will then have to critically assess and evaluate the resulting position concerning whether and to which degree it meets the criteria for being in RE (Sect. 3.3).

3.1 The Starting Position of a Reflective Equilibrium Process

The main tasks of the first stage are (i) to clarify one’s goal, that is, to formulate the pragmatic-epistemic objective that one wants to achieve by applying reflective equilibrium, and (ii) to clarify one’s initial epistemic position, that is, to identify and select initial commitments, theoretical virtues, background theories and background information, and candidates for the system.

Being as clear as possible about one’s pragmatic-epistemic objective will help to (preliminarily) identify and specify all the other elements that are needed for the RE process. Some possible examples for such objectives are: to find a principled answer to the question of in which cases one is allowed to lie, to formulate and defend a theory of justice, to construct a weather model that can be used for predicting the local weather in mountain valleys, to justify a theory of inductive logic, or to justify a precautionary principle that can guide climate policy decisions. Having a clear goal will help to constrain the subject matter, which in turn helps to identify relevant commitments, relevant background elements, and existing candidates for the system. It also helps to determine which theoretical virtues are (likely to be) especially important for the target system in order to meet the objective. As with everything in RE, though, the formulation of the pragmatic-epistemic objective is not a one-way street: as we learn more about the subject matter, we might be able to formulate our objective more precisely.

In the remainder of this section, I focus on the tasks, methodological decisions, and challenges that one needs to address when one wants to clarify and describe one’s initial position in order to start an RE process. Of course, these tasks can influence each other and do not have to be done in a strict order.

3.1.1 Identifying Initial Commitments

To improve our epistemic position and to work towards a state of reflective equilibrium, we need to assess the degree to which our commitments and our system are in agreement. When assessing this relation of agreement and deciding about possible adjustments, the independent credibility of the commitments has to be considered, that is, the credibility that the commitments have independently of their status in the RE process.

When we want to apply RE, one important task is thus to identify relevant commitments and their independent credibility. This poses a number of challenges. Without claiming to be exhaustive, I address now what I consider to be the five most important challenges.

Making One’s Own Commitments Explicit

Firstly, the epistemic agent needs to find ways to make her own commitments about the subject matter explicit, at least as far as possible. This is not trivial, as we typically are not aware of everything that we are committed to, and even less the reasons why we hold these commitments. Making one’s commitments explicit will to some extent be a creative process, but can be guided by systematic considerations: the agent can familiarize herself with the relevant literature, consider relevant cases in order to see how she would decide, use brainstorming techniques, discuss with others, and so on.

The process of making one’s commitments explicit will also typically be incomplete, and has to be continued throughout the process of adjustments. There is always the possibility that some new considerations emerge that were not considered before.¹

Considering Commitments of Others

Secondly, on her own, the epistemic agent may only come up with a limited selection of relevant considerations. To broaden the input, the agent should thus also consider the opinions of others, like other researchers, but also of practitioners and of affected parties—for example, in a biomedical context, the commitments of physicians, nurses, and patients (Van Thiel and Van Delden 2009). While the agent does not have to accept the commitments of others at face value, it is important to consider them in order not to overlook aspects of the subject matter that the agent herself might not have access to, or where the agent might be lacking the experience to form commitments. The agent thus has to identify relevant other parties whose commitments should be considered, and select appropriate methods to obtain them. What methods are appropriate depends on the project at hand, e.g., its objective, scope, and subject matter. Some examples are empirical studies, sending out surveys, studying relevant literature, interviews, and personal discussions.

Specifying Which Considerations Are Independently Credible

Thirdly, it needs to be specified what kinds of commitments have independent credibility. There is no general, project-independent answer to the question what gives a commitment independent credibility, and how much weight should be assigned to a commitment because of it. Instead, these are parameters that have to be addressed for each particular application of RE. To do so, the epistemic agent will have to refer to epistemological background theories, e.g., theories of confirmation, perception, intuition, and testimony (Baumberger and Brun 2017, Fn. 16). Which sources of independent credibility are relevant and how much weight they should give a commitment also depends on the project at hand, e.g., whether it is an empirical or a normative research project. In addition to perception or intuition, which were already named in Chap. 2 as possible sources of independent credibility, Van Thiel and Van Delden (2009, 236–37) name durability, transcendence, and experienced perception.

Durability :

means that we should have more confidence in commitments to judgments that are confirmed in a history of cases.²

Transcendence :

refers to the extent to which commitments are appreciated and affirmed by a community.

Experienced perception :

means that we should give more weight to commitments of people who have relevant experience, e.g., the commitments of medical practitioners in contexts of medical ethics.

Two things are important to stress here: firstly, as explained in Chap. 2, for a commitment to count as independently credible, we do not have to be able to give positive reasons for a commitment above and beyond that we are committed to it, i.e., are endorsing it and acting according with it in our actual practice. Often, we will have additional reasons to support it, but this does not have to be the case. Ultimately, whether or not such a commitment can be justified will depend on whether or not it can be shown to be part of a resulting position in reflective equilibrium. Secondly, no matter how the independent credibility of a commitment is identified, it is only a factor that gives additional weight to commitments—in no way does it make commitments safe from revision.

How to Individuate and Count Commitments?

Fourthly, the question of how commitments can be individuated and counted poses a challenge for listing commitments. Tersman (1993, 44) argues that it is highly dubious whether there is a useful way of counting and individuating the beliefs of a person—it is far from clear whether the commitments of a person can be represented by a determinate and finite set of propositions. Tersman concludes that this suggests that the question to what extent commitments and system are in agreement must be assessed intuitively.

Let us look at an example to see why it poses a challenge for the assessment of agreement, and for deciding how to adjust one’s position. Let us assume that I have a broad range of commitments of the sort “When coming home late from being out with friends, it is wrong to lie to your partner and say that you had to work”, “You must not lie to your mother and tell her that you are sick when in reality you just do not want to go to her birthday party”, “When you borrowed a book from a friend and lost it, you must not tell him that it was stolen”, etc. Based on these commitments, I adopt the principle “It is always wrong to lie”, which is in agreement with 25 commitments of this sort. However, I also have one single commitment expressed in the sentence “When you know that person A wants to murder B, then you should tell A that you do not know where B is, even when you are sure about the whereabouts of B.” This latter commitment conflicts with the principle that it is always wrong to lie. How should we resolve the conflict?

Note first that, in reflective equilibrium, this problem is independent of the method of counterexamples. We have a supported claim ‘It is always wrong to lie’ and an independently plausible case that is in tension with it. That just sets a problem. The mere fact that there is a counterexample to the principle doesn’t determine whether we should call for a revision of the principle, or whether we should adjust or even reject the conflicting commitment.³

So how can we determine how to resolve the conflict? As I already mentioned, the individuation of commitments seems relevant for such problems. On the one hand, we could argue that the conflicting commitment is outweighed by the sheer number of other commitments that agree with the principle and thereby lend support to the principle over the conflicting commitment. On the other hand, we could also try to come up with even more commitments that are similar to the conflicting commitment, thereby increasing the number of commitments with which the system conflicts. Should we then say that the system is outweighed by the huge number of conflicting commitments?

One might think that a solution could be to somehow count commitments based on the relevant features they refer to. In the lying case, we could try to distinguish between classes of commitments where lying does negatively affect others, and cases where it would protect them. However, this raises the question of whether it is even possible to classify commitments before we have a system—after all, it is the purpose of the system to identify relevant features of commitments and to systematize them. Claiming that there is a relevant difference between these situations of lying is already a proposal for systematizing the commitments.

Possible ways to deal with this challenge could be to work with a fixed set of sentences that represent commitments, and to accept that the justification reached will be relative to this fixed set (as is the case, e.g., in the formal model of Beisbart et al. 2021). Another way is to start with a relevant selection of explicit input commitments, but to be aware that they are only a part of the whole set of commitments, that there are further implicit input commitments which can emerge during the process of adjustments, and to systematically search for them when adjusting system and commitments.

When assessing agreement between commitments and different candidate systems, it will be important to compare candidate systems with respect to the same set of commitments, and to search systematically for potentially conflicting or supporting commitments that need to be made explicit. This does not solve the problem of how to individuate and count commitments, but at least helps to ensure that not one candidate is rejected on basis of counterexamples that are not considered for another candidate. It also means that a resulting position can be defended as being in a state of RE only with respect to the commitments that were explicitly considered, and that it is always possible that there will be further emerging commitments that might or might not unhinge it.

Making a Selection of Initial Input Commitments

This means that, fifthly, a selection of those initial input commitments that should explicitly enter the process of adjustments needs to be made. One objection against reflective equilibrium is that too much—actually all that is relevant—depends on the selection of the input: from implausible or repugnant commitments can result implausible or repugnant principles (Kelly and McGrath 2010).

As an answer to this objection, some RE proponents propose to filter the commitments in order to only admit credible ones. This is in line with the demand of Rawls (1999, 42) that only those judgments in which we have high confidence, and which are made under “conditions favorable to the exercise of the sense of justice, and therefore in circumstances where the more common excuses and explanations for making a mistake do not obtain.” Not just any judgment should be included, but only our considered judgments. For example, Beauchamp and Childress (2013, 405–09) argue that we should start with those moral beliefs that are part of the “common morality” and made by “moral judges” that have relevant epistemic virtues, such as being impartial, or having sympathy and compassion for the welfare of others.

To arrive at a justified output, the argument goes, it is important to start from credible inputs (cf. Van Thiel and Van Delden 2009). And indeed, one of the criteria of the RE conception described in Chap. 2 is that commitments should have independent credibility. Does this mean that we should “filter” our commitments in order to only include those that have a high independent credibility?

The RE conception from Chap. 2 does not presuppose a strong filtering condition for input commitments: commitments should have independent credibility, but to be included in the process of adjustments, this credibility can be very minimal—including the limiting case where nothing speaks against this commitment. Commitments are considered during the process, and considering and “vetting” or “filtering” them in piecemeal fashion prior to entering the process of reflective equilibrium, is, as Knight (2017, 49) argues, counterproductive: “we have no way of knowing whether these isolated speculations will be consistent with the most plausible overall position.”

Van Thiel and Van Delden (2009, 235) also argue against a strong filtering condition in the sense of only considering highly credible commitments, because this might (a) lead to excluding intuitions from minority groups of agents, and (b) narrow down the available input too much—e.g., selecting commitments based on whether they belong to the “common morality”, which is a set of norms all morally serious persons share, may result in a very small set of input commitments that does not provide a sufficient basis for moral theorizing. Furthermore, (c) “limiting the set of moral intuitions in this way complicates the task of integrating the relevant moral experience of others. For example, from agents who may have moral intuitions that are not shared by all morally serious persons, because their intuitions stem from moral experience that is uncommon” (Van Thiel and Van Delden 2009, 235). Only considering input commitments that we take to be highly credible might also (d) lead to conservatism, since we will be wary of altering them.

All this speaks in favor of also accepting commitments with a very low independent credibility, and to explore how the various commitments and their relative weights work together. Since (candidate) systems should account for all current commitments, making the set of commitments as broad as possible will enrich the process by adding more (potentially) relevant factors that might or might not be a suitable basis for systematization. Making sure that their weights are considered adequately means that commitments with a very low credibility will not easily lead to the elimination of more credible commitments (after all, that independent credibility of commitments has to be respected is a criterion of RE). At the same time, the low independent credibility of a commitment can be overridden by its being incorporated into a credible position—this might even be true for commitments that are initially “incredible” (cf. Elgin 1996, 119).

But if commitments are not selected based on their degree of independent credibility, how can we identify those commitments that we should explicitly consider in the process of adjustments? Since it is not possible for human epistemic agents to consider all and every commitment that they hold, a selection has to be made, even if one rejects a strict filtering condition (Van der Burg and Van Willigenburg 1998, 4). Firstly, it is important that this selection is made in a way that other researchers are able to understand why certain commitments were selected, even if they perhaps would have selected others (de Vries and van Leeuwen 2010, 498). Secondly, this highlights the role of “emerging” input commitments for practical applications: the epistemic agent should not only start with a specific selection of commitments, but should strive to broaden the set of relevant commitments throughout the whole process (Van Thiel and Van Delden 2009, 235). This can happen through systematically searching for counterexamples (or further supporting examples), through engaging with the literature, through considering thought experiments, through presenting and discussing one’s commitments with others, etc.

It makes sense to start with a selection of what we take to be commitments to “core cases”, “central problems”, and “paradigm examples” of the subject matter. We also have general commitments about what does or does not belong to the subject matter, and should name some examples that we take to be especially relevant. They can of course also be adjusted, and, depending on the pragmatic-epistemic objective, sometimes it might be better to deviate more from the initial subject matter to further our epistemic goals, and sometimes less. It just needs to be “traceable back”, so that it is plausible that we solved the problem we originally set out with. This might include recognizing that the problem is better described in another way than we originally thought—as long as we ensure that we did not abandon the problem altogether without such reasoning, i.e., did not simply leave it because it was too hard, or uninteresting, and went doing something else instead (which can be also legitimate, but in such a case we should not claim to still be talking about the initial problem).

Tasks to Identify Initial Commitments

• Identify relevant literature and other information sources on the subject matter and familiarize yourself with it;

• Employ strategies to elicit your own commitments concerning the subject matter;

• Identify relevant parties whose commitments should be considered and select appropriate methods to obtain them;

• Identify sources of independent credibility (based on epistemological theories and on what is relevant for the subject matter at hand);

• Make a selection of explicit initial input commitments that you deem representative, and explain your reasons for making this selection in a way that is comprehensible to other researchers.

3.1.2 Selection of Theoretical Virtues

To be part of a position that is in reflective equilibrium, the target system needs to have theoretical virtues—we want a systematic account of the relevant commitments, and this means that the target system—be it a (set of) principles, a model, or a theory—needs to have certain virtues like being simple, fruitful, or having unifying power. Which theoretical virtues are relevant, and how they should be weighted, depends on the overall pragmatic-epistemic objective that is pursued. For example, if you want to come up with a general moral theory—e.g., a theory of justice—the virtue of “broad scope” might be more important than if you are addressing a specific issue of applied ethics in a specific context, e.g., ethical questions concerning the patenting of genetic resources and ownership of digitized sequence information. Making a preliminary selection of theoretical virtues will thus help to further clarify what kind of project is pursued. Being as clear as possible about how one understands these virtues will help when comparing candidate systems in order to argue in favor or against them.

The virtues and their interpretation are not fixed once and for all, but can be seen as second-order commitments (Elgin 2014) that are also open to revision if, e.g., it turns out that prioritizing the virtue of broad scope is no longer contributing to our pragmatic-epistemic objective.

Nonetheless, to describe the initial setup of the RE process, it is necessary to make a selection, even if it is preliminary.

Making a Preliminary Selection of Theoretical Virtues

• Which theoretical virtues are likely to contribute to the pragmatic-epistemic objective? That is, which theoretical virtues are likely to be relevant for the target system given the pragmatic-epistemic objective?

• Which of those virtues are likely to be more important for the target system, and why?

3.1.3 Description of the Background

To be justified according to reflective equilibrium, the epistemic position of an epistemic agent should not only be internally coherent in the sense that commitments and system are in agreement. To be fully justified, this position needs to be in reflective equilibrium with respect to the best available background theories and to all available background information that is relevant for the subject matter (Baumberger and Brun 2017, 174–75).

For an epistemic agent who wants to apply RE, this poses the challenge of getting an overview of the background which is comprehensive enough without requiring processing so much information that the method becomes unworkable. As with the selection of initial input commitments, it might make sense to work with a relevant selection, while being aware of, and familiar with, a broader picture. Background theories and background information is also one aspect of applying RE where its social aspects become especially salient: even if a single epistemic agent is applying RE, they will have to rely on the epistemic achievements of others in order to identify justified background theories that they can use.

Which theories will be relevant to, e.g., constructing arguments in favor of or against commitments or candidate systems cannot be fully determined before starting the process. But thinking about it, considering relevant literature, and consulting others, will again help to get a clearer idea of one’s subject matter: what are the theories that I presuppose and that I do not aim to justify as part of the position in the foreground? Which theories have sufficient justification to be used as external support for my position? For example, in the case of thinking about ownership of digitized sequence information, relevant background theories might concern ownership in other contexts.

That these background theories are, for the purpose of the RE process, presupposed, does not mean that they are safe from revision: just that they are currently not in the focus of what is being justified. It is possible that the process makes clear that one of these background theories has problematic implications and should be revised, but as soon as we do this, we are changing our epistemic project.

Identifying Relevant Elements in the Background

• What are relevant and plausible background theories?

• What information, e.g., factual information about the subject matter, is relevant?

• Are there any assumptions that need to be made?

The answers to these questions will also depend on the selection of initial commitments and on possible candidates for the system, e.g., what background theories are needed in order to correctly interpret them.

3.1.4 Selection of Candidate Systems

To get the process of adjustments going, we will need candidates for the systematization of the commitments in form of principles or a theory. Through familiarizing oneself with the subject matter and debates concerning it, one will most likely be able to draw up a list of the main contending candidates for the system. Further candidates should be added, if one can think of any that seem plausible (Knight 2017, 57). If there are no available candidates, one will have to formulate them in order to get the process going. This is largely a creative process which cannot be guided by explicit rules, but it will be helpful to think about the theoretical virtues that one is aiming for, and to examine the initial commitments to see whether some of them might be suitable to be reformulated as parts of the system. The selected candidates should be real alternatives, not just straw men, while at the same time the selection has to be kept at a manageable number. As part of this, it is particularly important to also consider candidates that one does not agree with, but that are, e.g., widely endorsed.

Selecting Candidates for the System

• What are existing plausible candidates from, e.g., the literature?

• Are there commitments in the set of initial input commitments that could be suitable as a (part of) the system?

• Can you come up with further plausible candidates for the system?

We did now address challenges for describing the initial position, and for identifying the elements that enter the process of adjustments. But, if we want to make progress from this initial position, we need to know how to bring virtues, weights, etc. to bear on the process of adjustments. We look at this in the next section.

3.2 The Process of Adjustments

An epistemic position that is in a state of reflective equilibrium is reached through a process of mutually adjusting commitments and systematic elements (see Chap. 2). This process of adjustments is often described in terms of moving “back and forth” between commitments and system—neither side takes priority nor is safe from being adjusted (cf. Goodman 1983, 64; Rawls 1999, 18; Cath 2016, 214).

The goal of this stage is thus to adjust commitments and system with respect to each other in order to maximize the RE criteria. The question is whether we can spell this out in the form of rules, or at least guidelines, that can be applied methodically. Can we structure the process of adjustments in a way that allows users of RE to conduct it systematically, and that will help them to make progress?

3.2.1 Steps of the Equilibration Process

While in principle nothing is safe from revision in RE, we cannot justify everything at once. When spelling out the methodical steps of the RE process, I thus work with the background/foreground distinction in the sense that the justificatory project takes place in the foreground, and adjustments should primarily made to the commitments and the system. The goal is neither to justify the elements in the background, nor the theoretical virtues, nor the sources of independent credibility (unless, of course, this is our pragmatic-epistemic objective, but in which case it constitutes the project in the foreground). The idea is that there is something that is adjusted—there is a specific objective with respect to which we want to make progress—and other things that are preliminarily taken to be fixed. (With a special emphasis on preliminarily fixed, for the purpose of a given justificatory project—they can always come into focus in other projects.)

Thus, I propose to adjust commitments and system in two alternating steps, in which either the commitments are adjusted with respect to the current system, or the system is adjusted with respect to the current set of commitments. As we want to make progress not only with respect to agreement between commitments and system, but also with respect to the other RE criteria, they should come to bear on the decisions about which adjustments should be made. I thus propose two general kinds of steps.

Two Kinds of Steps of the RE Process

Adjusting Commitments :

Keeping the system constant, find the set of commitments that maximizes the combination of (i) agreement with the current system, and (ii) independent credibility, and (iii) respect for input commitments, and (iv) support from background theories.

Adjusting the System :

Keeping the current set of commitments constant, find a system that maximizes the combination of (i) agreement with the current system, and (ii) theoretical virtues, and (iii) support from background theories.

For reasons of cognitive manageability, it is also possible to hold parts of the position constant and only adjust a subset of the commitments with a part of the system. This also leaves room for, e.g., explicating concepts as part of an RE process (Brun 2020). In fact, it is to be expected that in practical applications, steps of the RE process will often consist of explicating concepts or systematizing subsets of the current commitments: in most cases we do not have (m)any pre-developed, fully fleshed-out candidate systems that we can compare. What Brun (2020, 934) says about explication also has consequences for RE:

[In] many projects of explication the target theory is not readily available, and explicating concepts must therefore go hand in hand with developing a target theory[.]

Like explicators, RE agents typically cannot draw on a pre-developed system, and this is because developing such a system is one of the objectives of the RE process. But the building blocks of such a system, e.g., explicated concepts, are typically not readily available. An RE process might proceed in steps of partial systematizations which in turn impose further constraints on the continuation of the process. Brun (2020, 938–39) illustrates this with an example from Goodman (1951):

[We] can imagine starting with an explication of point in terms of intersecting lines; this will have consequences for a subsequent explication of to the left of, the extension of which will now need to include certain pairs of pairs of intersecting lines; and this explication will in turn restrict our choice of explicata for to the right of to converses of the relation to the left of; and so on for further explications.

Thus, the two alternating steps of the RE process can be carried out both with the current system/set of commitments as a whole, or with parts/subsets of them. The important point is that if a “sub-process”, e.g., an explication, is conducted, this always has to happen with respect to the overall pragmatic-epistemic objective for the position as a whole.

A challenge that becomes immediately clear is that the different criteria that should be maximized will often pull in different directions in both kinds of steps. What if one way to adjust commitments leads to a set that has a higher degree of agreement with the current system than another set, but the latter has a higher degree of independent credibility? How should such trade-offs be decided? And how in the first place can we assess whether or not a set of commitments has a higher degree of agreement than another?

Thus, the challenges for this stage are to specify the criteria of the two steps in a way that allows us to assess the extent to which they are met, and to decide how situations should be resolved in which trade-offs between criteria need to be made.

I first focus on the specific challenges for defining and assessing the RE criteria as part of specifying the method, before focusing in particular on the question of how trade-offs can be handled (in Sect. 3.2.3).

3.2.2 Defining and Assessing the RE Criteria

To specify the two alternating steps proposed above, we need to find ways to measure and compare the following:

• Agreement between (sets of) commitments and candidate systems;

• Support from Background Theories;

• Theoretical virtues of candidate systems;

• Respect of current commitments for input commitments.

On this basis, we can then address the issue of how to handle and decide trade-offs between agreement, theoretical virtues, and weights of commitments.

In the following, I describe what challenges need to be addressed in order to specify these measures. How I specify them for the purpose of my case study on precautionary principles is part of Chap. 5.

3.2.2.1 Measuring Agreement

The central relation that has to be brought into equilibrium is the one of agreement between commitments and a system. Agreement can be spelled out in different ways, but is typically understood to require more than mere consistency. One way is to spell it out in the form of account, i.e., that it has to be possible to obtain the commitments via inference from the system (Beisbart et al. 2021). These inferences can be specified as deductive or also as non-deductive inferences, and they will typically require that the system is applied to relevant background information. For a full agreement between system and commitments, we might require that the system can account for all of the commitments, and that we are committed to everything that follows from the system.

When measuring the account between the current commitments and the system, we face at least two distinct challenges:

Firstly, not all input commitments are likely to be explicit in the current commitments. This means that when comparing different candidate systems with respect to their ability to account for commitments, their success will be contingent on which commitments are explicitly considered.

Secondly, we will typically not have an overview of everything that does follow from a candidate system, that is, we will not have an overview about its logical closure.

Both challenges mean that it is often difficult, if not impossible, to get an accurate measurement of agreement (unless we work with a restricted, fixed set of inputs, like the dialectical structures used in the formal model of Beisbart et al. 2021). Thus, it is possible that a candidate system S _a is accepted over another S _b on the basis that it can better account for current commitments, while it later might turn out that there were actually further—implicit—input commitments that S _b can account for while S _a actually conflicts with most of them. Additionally, even if we decide to assess account only with respect to the commitments that are currently made explicit, it will still not be possible to compare those commitments with everything that follows from the system if we cannot overview all of it.

Possible ways to deal with these challenges are to keep these limitations in mind when applying the two steps, and to use heuristics to compensate to some degree for these shortcomings: thus, we should always make explicit with respect to which set of commitments agreement is measured, and we must systematically search for commitments that are not made explicit yet but might cause problems, as well as for inferences from the system that might lead to conflicts.

Defining a Measure for Agreement

• Decide how explicitly and exactly agreement should be measured (this also depends on what is feasible within the project at hand);

• Specify different degrees of agreement (e.g., account, consistent non-account, conflict), and decide how they should be weighed.

3.2.2.2 Measuring Support from the Background

To be in reflective equilibrium, the position should not only be internally coherent—agreement between system and commitments—but should also be reasonable given all of the best available background theories and background information. Given that one is aware of the background, it seems likely that support from background theories could be measured analogously to agreement. However, this will quickly become very complex and unworkable for a single epistemic agent, who would have to be familiar with all relevant background theories and their implications. In most applications, it will thus be more sensible to use references to the background selectively, e.g., as tie-breakers in cases of trade-offs, and to assess whether already well-developed positions can reasonably be seen to be in a state of reflective equilibrium (see Sect. 3.3). Assessing the degree to which a position is supported by background theories will typically have a strong social component, requiring that the RE applicant relies on the epistemic achievements of others who, e.g., justified said theories.

Defining a Measure for Support from the Background

• Decide how explicitly background elements should be included, and at what point(s) support from them should play a role.

3.2.2.3 Measuring Theoretical Virtues

The system does not only have to be adjusted with respect to the (current) commitments, but also with respect to theoretical virtues: in order to provide a systematic account of the commitments, i.e., an account that helps us to meet the pragmatic-epistemic objective that we pursue in conducting the RE process, the resulting system should have theoretical virtues like scope or simplicity. The (preliminary) selection of a set of virtues was already addressed in Sect. 3.1. But if we want to be able to compare different candidate systems with respect to their theoretical virtues as part of the process of adjustments, we also need a way to measure and comparatively assess the virtues.

This poses a challenge, as it seems implausible that all relevant theoretical virtues will always be operationalizable on a ratio scale or even an interval scale (Stegenga 2015, 269). Okasha (2011, 102) illustrates this with the example of the virtue of “fruitfulness”:

Conceivably, one could order a set of theories by how fruitful they are, but it is hard to believe that differences in fruitfulness can be compared; a statement such as ‘the difference in fruitfulness between T1 and T2 exceeds the difference between T2 and T3’ hardly seems meaningful. If this is right, then the real-valued ‘utility’ function that represents the fruitfulness preference order is merely ordinal—any increasing transformation can be applied to it without loss of information.

Different measurement scales may be appropriate for different criteria, and may also depend on the scope of the project that we are pursuing (cf. Okasha 2011, 103). Thus, for other virtues, it might be possible to measure them on ratio or interval scales. But as long as even one virtue remains that is measured on an ordinal scale, no complete ordering will be possible that meets reasonable conditions of theory choice (Okasha 2011, 93), because it is not possible to establish unambiguous trade-offs (Kemp and Grace 2010, 401).

When defining measures for the theoretical virtues, it thus makes sense to define them as precisely as possible, but to keep the pragmatic-epistemic objective in mind. Often, it might be better to use an ordinal scale instead of trying to force everything onto an interval or a ratio scale. However, virtues should at least be comparable on an ordinal scale, as theoretical virtues that do not allow for any comparison of candidate systems are not feasible for the RE process (and neither for any other purpose that aims at making a selection).

Defining Measures for the Selected Theoretical Virtues

• For each virtue, define a measurement scale that is at least ordinal.

3.2.2.4 Measuring Respect for Input Commitments and Independent Credibility

In Chap. 2, I argued that two senses of “respecting input commitments” can be distinguished: (1) as meaning that we have to respect their independent credibility, a criterion that also extends to adjusted or newly inferred commitments insofar as they have independent credibility, and (2) in the sense that, referring to the pragmatic-epistemic objective and the background, we have to be able to argue that the resulting commitments do not constitute a radical change of subject when compared with the input commitments.

For measuring, aggregating, and comparing the independent credibility of commitments, similar challenges might arise as with measuring the theoretical virtues: it might not always be possible to measure different sources of independent credibility—e.g., durability, experienced perception, or strength of intuition—in a way that allows us to obtain a complete ordering of (sets of) commitments. As with measuring virtues, it will make sense to be as precise as possible while ensuring that the measurement is still meaningful.

Concerning the question of whether the subject was changed, one way to interpret this criterion might be as demanding that as few as possible of the input commitments are adjusted. This could give us a negative measurement of “respect”, where penalties could be assigned for differences between the input commitments and the current, adjusted set of commitments. The weights of commitments (i.e., their independent credibility) could be integrated into this measurement by, e.g., increasing the penalty for eliminating a commitment with a higher weight (cf. Beisbart et al. 2021).

However, we could also argue that while the input commitments constrain the subject matter in some ways, they do not define it completely. For example, I might be committed to the proposition “We should not research and develop climate engineering technologies to alleviate adverse effects of anthropogenic climate change”. If I adjust this commitment in the sense of reversing so that I am now committed to the proposition that we should research and develop these technologies, did I really change the subject? This suggests that assessing whether the subject matter was changed in a (too) radical way is more a pragmatic criterion. Assessing it at every step might not be particularly useful—if we assume that the commitments constrain the subject matter, and respect their independent credibility when adjusting them, then this might be enough of a guide for the process. Whether or not the subject matter was changed too radically, and whether or not we were successful in achieving our pragmatic-epistemic objective, might better be assessed retroactively, once we have reached a (preliminary) end point of the equilibration process.

After addressing the challenges we face when defining measurements for the individual criteria, let us next focus on the question of how to handle the trade-offs that can result.

3.2.3 Handling Trade-offs and Path-Dependency

When adjusting the position in the foreground—commitments and system—we want to make adjustments in such a way that we make progress with respect to maximizing the criteria. The goal is to find a resulting position that is in a state of reflective equilibrium. This includes the requirement that it is at least as plausible as relevant alternatives, i.e., we need to be able to comparatively assess positions. I defined two steps of the RE process, in which either commitments or the system are adjusted with respect to the RE criteria:

Adjusting Commitments :

Keeping the system constant, find the set of commitments that maximizes the combination of (i) agreement with the current system, (ii) independent credibility, (iii) respect for input commitments, and (iv) support from background theories.

Adjusting the System :

Keeping the current set of commitments constant, find a system that maximizes the combination of (i) agreement with the current system, (ii) theoretical virtues, and (iii) support from background theories.

The discussion of how the individual criteria can be measured has revealed (a) that, often, criteria might only be measurable on ordinal scales, in which case forming a weighted sum (i.e., an unequivocal measure for which set of commitments, or which candidate system, maximizes the criteria) is not possible, and (b) that there can be trade-offs between the criteria which cannot be resolved unambiguously, i.e., where it is not clear which adjustments lead to an overall better position.

We might still think that there is one fundamental constraint for selecting a candidate position (or, in the specific steps of the RE process, a candidate system or a candidate set of commitments): whenever there is a candidate that is pareto optimal, i.e., that is at least as good as all other alternatives with respect to all criteria, and better in at least one, then this candidate should be chosen. When a position P₁ ranks as well as another Position P₂ with respect to, e.g., theoretical virtues and respecting input commitments, and ranks better with respect to account, then choosing P₂ anyway simply does not seem defensible. However, during the process, we need to be aware of the danger of local maxima: maybe sometimes we need to accept temporary epistemic setbacks in order to make overall progress.

In any case, it is unlikely that there will always be such a dominant option. What if P₁ and P₂ were equally good with respect to account, but the system of P₁ is more theoretically virtuous, whereas the set of current commitments of P₂ respects input commitments to a higher degree? Or when comparing candidate systems: what if a candidate S₁ is more fruitful than another S₂, but the latter has a broader scope (assuming that these virtues can only be measured on ordinal scales)?

Additionally, the weighing or even the selection of specific criteria might change during the process. For example, in the earlier stage of an RE process, it might make sense to trade off agreement for an increase in scope, because this will force us to critically examine more commitments and thereby works against conservatism. One reason why a system cannot account for a commitment might be that the latter is not in its range of applicability. We might then, in the early stages of an RE process, prefer a candidate system that has a broader scope, but conflicts with more commitments (i.e., fares less well with respect to the criterion of agreement), over one that has a narrower scope, but conflicts with fewer commitments (i.e., that fares better with respect to agreement). If, however, during several steps of the RE process, we simply cannot get rid of conflicts with central commitments, then we might want to reconsider, and trade off scope in favor of agreement.

Similarly, we might be more reluctant at the beginning to adjust commitments before having explored several candidate systems. Even if the current system, e.g., was the best available candidate in the step before, we cannot be sure whether there is not a much better candidate to be had, which we just did not come up with yet. Thus, adjusting with respect to the “weak” candidate would lead us astray. This does not have to be a negative thing—exploring different ways in which the position can be adjusted, and making each such pathway as strong as possible constitutes an important epistemic achievement, and also contributes to the justification of the position that ultimately results, because we can then truly show that it is at least as good as all alternatives. But, e.g., cognitive limitations and time-constraints will typically make it unfeasible to explore every possible pathway. At least when coming up with a first systematization, we will want to focus our energy on pathways that look somewhat promising (even if there is never a guarantee that we will not find still another, better candidate). This means that in the early stages of the process, it might make sense to put up with the conflicts in the position, and only to start adjusting commitments once a candidate system has in some sense “proved its worth”, e.g., by being the strongest candidate in several rounds—or if actually all available candidates would conflict with the commitment. And of course, whenever we adjust a commitment that has independent credibility, we have to be able to give reasons why this adjustment is warranted—as long as we are not able to do this, we can at best tentatively explore whether adjusting this commitment in this way would, overall, lead to a position that is more defensible than a position in which the commitment was not adjusted in this way.

Thus, while in general we want to maximize the criteria, there does not seem to be a one-size-fits-all solution for trade-offs, and they rather have to be decided on a case-by-case basis with respect to the overall objective. The process of adjustments is neither infallible in the sense of guaranteeing that it leads to a state of reflective equilibrium, nor is it mechanical (Bonevac 2004, 386).

Specifying the Two Steps of the RE Process

• Define the RE criteria as precisely as possible while keeping them informative enough given the project at hand (i.e., the subject matter and pragmatic-epistemic objective in question, as well as the available resources);

• Give a preliminary weighting of the different criteria, noting whether any of them are more important with respect to the pragmatic-epistemic objective;

• Concretize the two alternating steps of the process by inserting the criteria so-defined.

3.3 Preliminary Conclusion of the Process and Evaluation of the Resulting Position

The process of adjustments comes to an endpoint when the position stabilizes, that is, when neither of the two steps improves the position anymore. But this does not yet guarantee that a full reflective equilibrium was reached. We thus have to assess to what degree the criteria are met in order to appraise the resulting position. This will include asking the following questions:

• Are the resulting commitments and the system in agreement?

• Can the position be supported by background theories?

• Does the system do justice to theoretical virtues?

• When comparing input commitments and resulting commitments, is it plausible that we did not abandon the subject?

• Do (at least some of) the resulting commitments have independent credibility?

• Is the resulting position at least as plausible as relevant alternatives?

The answers to these questions will most likely not be a clear yes or no, but rather a matter of degree: a position can be more or less supported by background theories, a system can have a higher or a lower degree of theoretical virtuousness, etc. To answer the question of whether a reflective equilibrium was reached—whether, on balance, the different criteria are met to a sufficient degree—thus also depends, once again, on the pragmatic-epistemic objective; and on whether or not there are better alternatives available. The only thing that should not be traded off in favor of other criteria is consistency—admitting jointly inconsistent claims would undermine the epistemic enterprise (Elgin 1996, 103).

If the answers to these questions are not satisfactory, we could, e.g., retrace some of the steps of the RE process and explore different pathways of adjustments, e.g., choosing to adjust the system in case of a conflict where previously the commitments had been adjusted.

Indeed, there is no guarantee that the method will lead to a justified account, as has for example been brought forward against RE by Kelly and McGrath (2010). But this is why we have the criteria to assess whether such a state was reached, and if so, to what degree. And if such a state was reached, then we have also reached a “fallible, provisional, but reasonable epistemological stance” (Elgin 2014, 255). At any point, our equilibrium can become unbalanced again, so that we will need to continue the process of equilibration. But this does not make the justification via RE useless, or unreasonable.

3.4 Recapitulation: A Methodology of Reflective Equilibrium

This chapter developed a methodological framework for specifying RE as a method which is intended to be applied by one epistemic agent. This methodology was developed with respect to the goal of justifying systematic accounts of a subject matter, for example, in the form of theories or principles. The chapter highlighted methodological and pragmatic decisions that one has to make in order to concretize a method of RE for specific justificatory projects. A lot of these issues are not inherent to reflective equilibrium, but have to be resolved by other research fields that can tell us, e.g., what counts as good evidence, what a strong non-deductive inference is, when testimony is reliable, and so on. Still, these decisions need to be made in order to be able to apply RE as a method. How they are made always depends on the specific project at hand.

The specific project in this book concerns the justification of a precautionary principle as a case study on the applicability of RE. Thus, in the next Chap. 4, I conduct a literature survey on precautionary principles in order to familiarize myself—and my readers—with the subject matter. This survey will also serve as a source for identifying commitments, background elements, and candidates for the system.

Chapter 5 then describes the setup for the case study. It identifies the elements of the initial position, and makes the necessary methodological decisions in order to concretize the RE method for its application to this specific project, that is, it addresses the tasks that have been identified in the present chapter. The case study itself is conducted in Chaps. 6–8.