Case Study, Phase II: Focus on the Process of Adjustments

Rechnitzer, Tanja

doi:10.1007/978-3-031-04333-8_7

Tanja Rechnitzer²⁹

Part of the book series: Logic, Argumentation & Reasoning ((LARI,volume 27))

1018 Accesses

Abstract

This chapter describes the second phase of a case study in which reflective equilibrium is applied with the goal of justifying a precautionary principle. It focuses on how the RE criteria can be used to comparatively assess different candidate systems, i.e., theory choice, and to adjust commitments and system with respect to each other.

You have full access to this open access chapter, Download chapter PDF

The previous Chap. 6, explored how the RE steps can be used for the development of a candidate system, i.e., theory construction. The present chapter is the second phase of the case study. It focuses on how the RE criteria can be used to comparatively assess different candidate systems, i.e., theory choice, and to adjust commitments and system with respect to each other. In the next Chap. 8, a specific position will be fleshed out, and evaluated with respect to whether or not it is a justified position in RE.

7.1 Overview: Phase 2

In the second phase, I focus on implementing the two alternating steps of adjusting the system with respect to commitments (step A) and vice versa (step B) while being as detailed as possible. In order to have enough input in form of candidate systems to compare, I adopt proposals from the PP literature (see Chap. 4). Considering real alternatives, i.e., compelling rival principles, is important to provide a fuller justification in RE (cf. Knight 2017, 50). In the context of the case study, this is also important in order to test whether the RE criteria can be applied in a meaningful way to make a selection between different strong candidates. Additionally, I could have tried to obtain further candidates in ways similar to how the first candidate was developed in phase 1, or through exploring different ways for, e.g., increasing the theoretical virtues of the Principle 3-System. Developing a sophisticated candidate system takes time, so adopting candidates from the literature is not only a way to make sure that real alternatives are considered in order to avoid dealing with straw men; it is also a pragmatic decision in the context of this being a case study for RE—developing a broad range of original candidates would take up a lot of time and space.

Section 7.2 continues step A₃ from phase 1. The Principle 3 (P 3)-System that I developed in Chap. 6 is compared with the “Rawlsian Core Precautionary Principle (RCPP)” (Gardiner 2006) and what I call the “Utilitarian Uncertainty Principle (UUP)” (Bognar 2011). The RCPP is an influential and widely cited PP proposal, so it makes sense to assess how it fares in the RE process. It also has been criticized: for example, the UUP is an explicit counter-proposal to the RCPP, with Bognar (2011) claiming that it can account for all the cases in the application set of the RCPP, but that it is also a defensible decision-principle in other cases—making the RCPP superfluous. As the comparison of the two candidates shows, this is not true—at least not with respect to my commitments. My own preliminary attempt at a candidate system, the P 3-System, is also rejected as a result of assessing it in comparison with the other two candidates, due to its lack of theoretical virtues.

After adjusting the commitments with respect to the RCPP in step B₃ (Sect. 7.3), I introduce a modified version of the RCPP in step A₄ (Sect. 7.4), the “Maximin-PP for combinations of uncertainty and incommensurability”, drawing on the work of Aldred (2013). To have an additional sophisticated alternative, I adopt the proposal from Steel (2015) in the form of a “Tripartite Precautionary Approach (TPA)”.

The Maximin-PP is selected because its higher ranking with respect to theoretical virtues arguably outweighs the slightly better account value of the TPA. However, even after adjusting the commitments with respect to the Maximin-PP in step B₄ (Sect. 7.5), substantial value-commitments remain unaccounted for (which would also be true if the TPA were chosen). I argue that this is because the Maximin-PP is a normative principle for rational choice, and that something more will be required to meet my pragmatic-epistemic objective of formulating a defensible moral precautionary principle. The latter is then the focus of Chap. 8.

Results from phase 2 are recapitulated and discussed in Sect. 7.6, which also includes Fig. 7.9, a schematic overview of the steps of phase 2. Throughout phase 2, gray boxes are again used to summarize the main points of each step. As before, only relevant or exemplary aspects of the process are described in detail, and readers can refer to Appendix A at the end of the book for the full list of commitments, candidates for (parts of) the system, background information, and case descriptions.

7.2 Step A_3.2: Comparing Principle 3-System, RCPP, and UUP

Two candidates from the literature, the RCPP and the UUP, and my own candidate from phase 1, the “Principle 3-System” are comparatively assessed with respect to their ability to account for commitments (Sect. 7.2.1, see Fig. 7.1 for a summary) and their theoretical virtues (Sect. 7.2.2). In the overall comparison (Sect. 7.2.3), I argue that although not without problems, the RCPP is the most defensible candidate, and I adopt it as the current system at the end of step A₃.

We are now resuming step A₃, which was started at the end of phase 1 in Chap. 6. In phase 1, I developed a candidate system, the Principle 3-System, which consists of Principle 3 and three additional parts.

Principle 3 (P 3) :: Where there are plausible threats of serious harm, precautionary measures that are proportional to the severity and plausibility of the threat should be taken.
P 3.1: Definition: Threat :: A threat is a possibility of harm that is uncertain.
P 3.2: Seriousness of Threats :: The seriousness of a threat is assessed according to (i) the potential for harm of the threat, and (ii) whether or not the possible harm is seen as reversible. [same content as IC 11]
P 3.3: ExplicPrec :: Explication of “Being a precautionary measure against an undesirable x”: An action a is precautionary with respect to something undesirable x if a fulfills the following necessary and jointly sufficient criteria:

1.
Intentionality: a is performed with the intention of preventing x.
2.
Uncertainty: the agent does not believe it to be certain or highly probable that x will occur if a is not performed.
3.
Reasonableness: the agent has externally good reasons (a) for believing that x might occur, (b) for believing that a will in fact at least contribute to the prevention of x, and (c) for not believing it to be certain or highly probable that x will occur if a is not performed.

I.e., Principle 3 is supposed to recommend those actions that fulfill the criteria of the explication, and are, on top of that, proportional to the severity and plausibility of the threat.

In the remainder of step A₃, this candidate system will be compared with the Rawlsian Core Precautionary Principle (RCPP) and the Utilitarian Uncertainty Principle (UUP). Gardiner’s RCPP is an influential PP interpretation (Gardiner 2006), whereas the UUP is derived from Bognar’s criticism and counter-proposal against the RCPP (2011). They have been selected based on the literature survey in Chap. 4.

The Rawlsian Core Precautionary Principle (RCPP) is a maximin-decision rule that is qualified through four jointly sufficient conditions. The first three conditions are cited (and slightly paraphrased) from Gardiner (2006, 74). The fourth condition is added on p. 51: “[The] RCPP needs some way of distinguishing a set of reasonable outcomes to contrast with those outcomes which are merely imaginable. This suggests that the three Rawlsian criteria mentioned so far must be supplemented with a further requirement: that the range of outcomes considered are in some appropriate sense “realistic,” so that, for example, only credible threats are considered.”

Rawlsian Core Precautionary Principle (RCPP)

If four conditions are fulfilled:

1.
No Probabilities: There is no, or no reliable, probability information about the possible outcomes available,
2.
Care Little for Potential Gains: decision-makers care relatively little for potential gains that might be made above the minimum that can be guaranteed by the maximin approach,
3.
Unacceptable Outcomes: the courses alternative to the one selected by maximin have unacceptable outcomes, and
4.
Reasonable Outcomes: the range of outcomes considered are in some appropriate sense “realistic” or reasonable,

then decision-makers should choose the course of action with the best worst case.

The Utilitarian Uncertainty Principle consists of a combination of the principle of maximizing expected utility and the principle of indifference: it tells us to treat all outcomes as equally probable if no probability information is available, then to calculate their expected utilities, and to select the option/course of action that has the highest expected utility:

Utilitarian Uncertainty Principle (UUP) :: If no or no reliable probability information is available, treat all outcomes as equally probable, and choose the option that has the highest expected utility.

7.2.1 P 3-System, RCPP, and UUP: Account for Commitments

See Fig. 7.1 for a comparison of the Principle 3-System, the Rawlsian Core Precautionary Principle, and the Utilitarian Uncertainty Principle with respect to account for the current commitments C₂. Changes in commitments from C₁ to C₂, i.e., the emerging commitments EC 14–EC 20, are marked by different table borders.

The P 3-System reaches the highest account value out of the three, 164. The RCPP reaches 72, whereas the UUP only has an account value of 35 because it in fact conflicts with some strong commitments, for example:

IC 5:: Don’t risk great harm in pursuit of modest benefit. [high]
EC 13:: Precautionary measures should be effective in preventing or substantially ameliorating either a threat or the harm of a threat. [high] [emerged at Step B₁]

The toy example Disproportionate Outcomes 2 further helps to illustrate how the UUP conflicts with my commitments:

Case 8: Disproportionate Outcomes 2

We have to decide between two alternative courses of action, “Safe 2” and “Risky 2” (see Fig. 7.2).

IC 20:: In case 8, Disproportionate Outcomes 2, the option “Safe 2” should be chosen. [high]
IC 21:: In case 8, Disproportionate Outcomes 2, the option “Safe 2” should be chosen because the worst case of option “Risky 2” is disproportionately worse than what we could gain from it as compared with “Safe 2”. [medium]

The commitments concerning precautionary measures, EC 5–EC 14, are accounted for by the P 3-System via the explication of “being a precautionary measure against an undesirable x”, and they are consistent with the RCPP and the UUP. However, the commitments on precautionary measures are only relevant for the P 3-System because it needs the explication of “being a precautionary measure against an undesirable x” in order to be able to yield somewhat determinate verdicts. The other two candidates yield determinate verdicts without this “detour” of first assessing whether a measure is precautionary or not. Still, there are two important differences between the two: while measures recommended by the RCPP will meet the criteria for being a precautionary measure, this is not the case for every measure the UUP will recommend, as, e.g., its failure to account for EC 13 shows.^{Footnote 1}

This is a point in favor of the RCPP, because as I argued in Sect. 6.8.2, not every precautionary measure is defensible. A justified PP should only recommend such measures that are warranted—but neither should it recommend measures that cannot even be classified as precautionary in the first place. Thus, the RCPP no longer needs the explication of precautionary measures, which was developed in the first phase of the case study (Chap. 6). We can thus move the explication to the background, as it is no longer needed as a part of the system.

7.2.2 P 3-System, RCPP, and UUP: Theoretical Virtues

The three candidates are comparatively assessed and ranked with respect to their theoretical virtues. In the descriptions, I focus on the most salient features that distinguish the three candidates, setting aside those that they share. For example, all three candidates have to presuppose that we can distinguish plausible or reasonable outcomes that should be considered for the decision process from those that are not. This decreases their determinacy. But since all of them face the same problem, it does not make a difference for their ranking.

Determinacy

The virtue of determinacy demands that “The target system should, together with relevant factual information, yield determinate verdicts, i.e., both its conditions of application and its verdicts should be precise and clear enough” (see Sect. 5.5 for more on this and the other virtues).

In specifying the kind of possible outcomes that “trigger” precautions, the RCPP is slightly more determinate than the P 3-System, since there is at least a “trivial”, straightforward interpretation of “unacceptable” available, whereas “serious” leaves more room for discretion. But aside from this, “unacceptable” is still not very determinate without further specification, and it is also far from clear whether the trivial interpretation is the correct/best one.

The “care little for potential gains” criterion of the RCPP decreases its determinacy, since this criterion can be interpreted in several ways: does it mean that the minimum (best worst case) is already really good, so that additional gains don’t mean much? Or that there is only a small difference between the best worst case and the best case?

On the other hand, the determinacy of P 3-System is impaired through reference to “irreversibility”: the relevant sense of “irreversible” is far from clear, and it could be spelled out in a number of ways (Randall 2011, 57–60; 70–72).

Lastly, RCPP and UUP refer to probabilities without clarifying the relevant sense of probability, e.g., should we apply them when we have no objective probabilities available? This might often be the case. Or when we have no subjective probabilities available? That would be much more rare (Hansson 2008; Roser 2017, cf.). On the other hand, the P 3-System refers to threats which are possibilities of harm that are uncertain, which is similarly unclear.

As for the determinacy of verdicts, the RCPP and the UUP both yield determinate verdicts by singling out exactly one course of action that should be taken. The P 3-System, on the other hand, is not so clear-cut, since while it does include an explication of “precautionary measure”, it does not further explicate when such a measure is proportional to the seriousness and plausibility of a threat. The use of this concept relies thus on our existing (pre-theoretical) understanding of proportionality, which leaves substantial room for discretion.

Based on this assessment, I rank the three candidates as following with respect to their determinacy:

$$\displaystyle \begin{aligned}UUP > RCPP > P3\end{aligned}$$

Practicability

A candidate system has the virtue of practicability when it is applicable in the sense that it specifies relevant information about actions and other items of evaluation that human beings can typically obtain and use to arrive at moral verdicts. E.g., it should process inputs that are typically available to us, and yield verdicts that are realizable by us. For more explanation, see p. 16.

The problem with the practicability of the P 3-System is that its determinacy is so low that it is not really possible to assess how accessible the relevant information typically is for us—because it is unclear what the relevant information would be.

For the RCPP, we have to be able to rank outcome values ordinally, and to be able to identify the worst case of an option. While the first speaks for its practicability—ordinal rankings are much easier to obtain than rankings on interval or ratio scales—the latter is not very practicable as long as we are missing a clear criterion for “reasonable” outcomes.

The UUP seems more practicable than both RCPP and P 3-System. Its main problem as regards practicability is that we need be able to assign cardinal utilities to the outcomes in order to calculate expected utilities, i.e., we need to be able to rank outcome utilities on interval scales. This might not always be possible.

I rank the practicability of the three candidates as following:

$$\displaystyle \begin{aligned} UUP > RCPP > P3\end{aligned}$$

Broad Scope

A candidate system has a broad scope when it has a broad range of applicability, i.e., it should tell us in as many cases as possible whether or not (which specific) precautionary measures are required. For more on scope, see p. 16.

The range of applicability of the RCPP consists of all situations in which outcomes are in some relevant sense reasonable, decision-makers care little about potential gains that can be made above the minimum that can be guaranteed by following maximin, no (reliable) probability information is available, and alternatives to the option recommended by maximin all include unacceptable outcomes. This is the smallest range of applicability of the three candidates.

Then we have the UUP, which covers all situations where values of reasonable outcomes can be ranked on cardinal scales.

The P 3-System can also handle situations when outcomes are incommensurable, as opposed to the UUP, where such situations are not in its range of applicability. I thus rank the three candidates according to their scope as following:

$$\displaystyle \begin{aligned} P3 > UUP >RCPP\end{aligned}$$

Simplicity

I understand the theoretical virtue of simplicity as demanding that the conceptual apparatus of the target system should be economical in the sense that the concepts it includes that cannot be reduced to each other are kept to a minimum. See also Sect. 5.5.

The P 3-System includes thirteen different concepts: threat; plausible (threat/outcome); serious (threat/harm); proportionality (of precautionary measures to plausibility and severity of a threat); potential for harm (of a threat); (ir)reversibility (of harm); prevention; intention; belief; certain; highly probable; externally good reasons (for believing something); contributing to the prevention of an event.

The theoretical apparatus of the RCPP consists of seven concepts: probability; outcome; course of action/option; maximin; “to care little for potential gains above a specific guaranteed minimum”; to be unacceptable; reasonable outcomes. The technical apparatus of the UUP consists of five concepts: probability; outcomes; option/course of action; utility; expected utility.

By including thirteen different concepts, the P 3-System does not seem to be particularly simple, even if considered on its own. The RCPP is considerably simpler in only including seven concepts, while the UUP is even simpler with only five. Consequently, the ranking of the three candidates according to their simplicity is:

$$\displaystyle \begin{aligned}UUP > RCPP > P3\end{aligned}$$

Overall Ranking: Theoretical Virtues of P 3, RCPP, and UUP

Here is a short overview of how the three candidates fare overall with respect to the theoretical virtues.

Determinacy:	UUP > RCPP > P3
Practicability:	UUP > RCPP > P3
Scope:	P3 > UUP > RCPP
Simplicity:	UUP > RCPP > P3

The UUP is always better than the RCPP, so we get overall UUP > RCPP. The RCPP is always better than the P 3-System aside from scope, where the P 3-System even outranks the UUP. How to handle this trade-off is discussed in the next subsection, where the candidates are compared overall, i.e., with respect to both account and theoretical virtues.

7.2.3 Overall Comparison of P 3, RCPP, and UUP

The P 3-System has the broadest scope of the three candidates, and also the highest account value (164). On the other hand, it ranks last with respect to the virtues of determinacy, practicability, and simplicity. We also have to consider that its high rank with respect to account (a) rests on the assumption that all outcomes are plausible in the relevant sense, and (b) is also partly due to the fact that it can simply account for some of the commitments because they are also part of the P 3-System (e.g., EC 2–4). I argue that compared with the available alternatives, this trade-off between account and theoretical virtues is not defensible: we should reject the P 3-System as a candidate at this point.^{Footnote 2}

With the UUP and the RCPP, we are also facing a trade-off between account and theoretical virtues: while the UUP ranks higher with respect to overall theoretical virtues than the RCPP, the latter has a higher account value (RCPP: 72, UUP: 35, see Fig. 7.1, p. 3). The low account value of the UUP is partly due to its conflict with some central commitments that have a high weight, e.g. IC 5, EC 13, and IC 20, and others with a medium weight, e.g., IC 21. And as the assessment of theoretical virtues has shown, even though the UUP ranks highest with respect to determinacy and scope, it is not completely unproblematic.

Thus, I argue that the theoretical virtues of the UUP cannot compensate for its low account value, and that for now, the RCPP is the most convincing candidate, despite its problems. I thus adopt the RCPP as the current system.

7.3 Step B₃: Adjusting Commitments to the RCPP

Commitments are adjusted with respect to the current system, the RCPP. In Sect. 7.3.1, I discuss whether the set of current commitments can be adjusted to increase agreement with the current system. The adjustments of excluding from the subject matter (i) cases with probabilities, or (ii) with more than one acceptable worst case/without any acceptable worst case, are rejected as not defensible. After searching for further relevant commitments in Sect. 7.3.2, the current set of commitments at the end of step B₃ is described in Sect. 7.3.3 and summarized in Fig. 7.6.

At the end of the last step, the Rawlsian Core Precautionary Principle (RCPP) was chosen as the current system. However, the assessment and comparison of the three candidate systems in step A₃ has shown that even though the RCPP is currently the most defensible candidate, it still has significant room for improvement. When adjusting the current commitments with respect to the RCPP, we have to keep this in mind and should not give too much weight to the current system.

7.3.1 Trying to Increase Account

First, I discuss whether commitments that are in conflict with the RCPP should be adjusted (a). Then, I discuss whether cases where either (b) probabilities are available or (c) cases that have more than one acceptable worst case, respectively no acceptable worst case, should be excluded from the subject matter to increase account, since the RCPP does not yield verdicts in such cases. Lastly, some commitments are clarified (d).

(a) Adjusting Conflicting Commitments

The RCPP conflicts with four commitments that are all general commitments concerning precaution and precautionary decision-making, not case-specific judgments.

First, we have the commitments to the Wingspread PP:

IC 2:: When an activity raises threats of harm to human health or the environment, precautionary measures should be taken even if some cause-and-effect relationships are not fully established scientifically. (Wingspread Formulation of the Precautionary Principle) [low]

There can be cases where IC 2 demands that measures should be taken whereas RCPP recommends against them: for example, if there is a course of action that has a small harm to the environment in its set of possible outcomes, but another alternative course of action has an unacceptably huge economical loss in its set of outcomes, then RCPP will recommend the first alternative, given that the other criteria of the RCPP are fulfilled. That is, RCPP will recommend the action that entails a threat to the environment when otherwise we would face unacceptable economic loss, but IC 2 only refers to threats of harm to human health or the environment, and does not take, e.g., economic loss into account.

I argue that interpreting the Wingspread PP in such an absolutist sense is not very plausible, and propose to replace IC 2 by the following new commitment:

C 2:: When an activity raises threats of harm to human health or the environment, then, pro tanto, precautionary measures should be taken even if some cause-and-effect relationships are not fully established scientifically. [medium] [replaced IC 2 at Step B₃]

Making the claim pro tanto means that it is still respected that threats to the environment and/or human health warrant precaution, but it also takes into account that this claim sometimes can be overridden. I thus argue that it is still close enough to the original input commitment to be seen as respecting it. While the RCPP does not account for C 2, it is at least consistent with it, which leads to a slight increase of the agreement between commitments and current system.

The next current commitment the RCPP conflicts with is:

EC 15:: Precautionary measures should not introduce serious threats of their own. [low] [emerged at Step B₂]

The RCPP only refers to unacceptable threats, and if we assume that while all threats of unacceptable outcomes are serious, but not all serious threats are threats of unacceptable outcomes, then it might also happen that the precautionary measure that it endorses will introduce a serious threat. However, independently of the RCPP, demanding that a precautionary measure is never allowed to introduce serious threats is potentially paralyzing, as it is possible that there is no such option. I thus propose the following replacement for EC 15, and argue that the RE criterion of “respecting input commitments” is fulfilled:

C 3:: Precautionary measures should not introduce threats that are equally or more severe than the threats they are aimed at, i.e., threats that have the same or a greater potential for harm. [medium] [replaced EC 15 at Step B₃]

I argue that the remaining two conflicting commitments cannot defensibly be adjusted:

EC 19:: The price of precaution should be proportional to the seriousness and the plausibility of the threat, given the available alternatives. [low] [emerged at Step B₂]

The RCPP conflicts with this because once no probabilities are available, all plausible outcomes are treated the same, i.e., additional comparisons of plausibility do not play a role. However, I think that it makes sense to somehow take into account our epistemic state concerning the possible outcomes, and that there can be comparisons made even if we have no probability information available. I thus argue that rejecting this commitment would not fulfill the RE criterion of “respecting input commitments”.

EC 20:: The target PP applies to plausible and serious threats and prescribes measures that are proportional to the severity and plausibility of the threat. [medium] [emerged at Step B₂]

The RCPP does not apply to all serious threats, given that not all of them will also be threats of unacceptable outcomes. The RCPP also fails to account for the part of EC 20 that concerns plausibility: all we have to know about outcomes in order to apply the RCPP is that they are “reasonable”, but we do not know their probability. Further information is not considered, neither about threats nor about precautionary measures against them—the RCPP does not include any comparisons of plausibility of outcomes. Consequently, the proportionality of measures to the plausibility of a threat does not play a role: the RCPP conflicts with the constraints EC 20 puts on the target PP. Should we reject EC 20 on this basis? I argue that the answer should be no, based on a similar argument as for EC 19: comparisons of plausibility are possible and sensible even if we have no probability information available.

The problem for the RCPP that is bigger than conflicting commitments is the broad range of commitments on which it remains silent because it does not apply. It is unable to account for any commitments on cases in which probabilities are available, or in which more than one of the available courses of action has an acceptable worst case, or in which no available course of action does have an acceptable worst case. Instead of zooming in on each of the particular, case-specific commitments, for each of these groups I discuss whether or not it makes sense to exclude them and thereby increase the RCPP’s ability to account for the current commitments.

(b) Excluding Probabilities from Subject Matter?

The RCPP does not apply to cases where probabilities are available. Consequently, commitments concerning such cases are not in its application set. However, it is not uncommon to restrict the scope of PPs to situations in which we know the possible outcomes, but not their probabilities—i.e., to restrict it to situations of decision-theoretic uncertainty. The idea behind this seems to be that as long as we do not only know the range of possible outcomes, but also their probabilities—i.e., as long as we are in a situation of decision-theoretic risk—we do not need something like a PP because we already know what to do: we can calculate expected utilities and choose the course of action that maximizes expected utility.

However, I think that there is a strong argument that there are at least some cases in which we should take precautionary measures even in situations of decision-theoretic risk. Choosing the course of action that maximizes expected utility might only make sense when potential losses are insurable, when we know that we can try several times, and/or when potentially bad outcomes are not extremely severe.

Randall (2011) identified relevant cases in which precaution is relevant—i.e., that a PP should be able to account for—even though probabilities are available.

For example, let us compare two courses of action. Option a, “widely dispersed”, has extremely good but also extremely bad possible outcomes. The outcome distribution is normal, i.e., symmetric around the mean, median, and mode; and the expected value is positive. Option b, “compact”, has the same expected utility as a, but its outcomes are much more narrowly dispersed, i.e., its best and worst cases are much less extreme, while the mode—the most likely outcome—is much more likely than in a. The two alternatives are illustrated in Fig. 7.3.

And things get worse when outcomes are not distributed along a normal distribution, but in some disproportional and/or asymmetric way. The set of possible outcomes of a course of action involves a disproportionate threat of harm when it includes possibilities of harm that, compared with the expected value of the course of action or with its most likely outcome, are disproportionately bad. An example is shown in Fig. 7.4a, where three discrete outcomes are possible, with a modest net gain being the most likely outcome, closely followed in likelihood by a modest net harm. Even though this extreme harm is unlikely, its possibility is non-trivial. And in comparison with what we can gain, it seems clearly disproportionate.

Asymmetric threats of harm is illustrated in Fig. 7.4b by a continuous distribution of outcome possibilities that is truncated at outcomes that are clearly bad but not disproportionately so (Randall 2011, 114). “Asymmetric” refers to the distribution of likelihoods, meaning that substantial (even though not necessarily disproportional) harm is more likely than it would be were the distribution of outcome likelihoods symmetric.

A threat is disproportionate and asymmetric when a disproportionately bad outcome is asymmetrically likely (Randall 2011, 114). This is illustrated in Fig. 7.4c.

All in all, while there might be cases where we know the probabilities of the possible outcomes and no special precaution is required—like in the case of the compact normal distribution—this is not true for all situations in which we have probabilities. Excluding them from the subject matter can thus not be defended.

(c) Excluding Cases with More Than One/without Any Acceptable Worst Case?

Because of condition 2 of the RCPP, “decision-makers care relatively little for potential gains that might be made above the minimum that can be guaranteed by the maximin approach”, the worst case selected by the maximin-rule—i.e., the best worst case—has to be at least acceptable. But according to condition 3, “the courses alternative to the one selected by maximin have unacceptable outcomes”. This means that the RCPP cannot account for any commitments where more than one course of action has an acceptable worst case, nor for commitments about cases where none of the courses of action has an acceptable worst case.

The latter is for example the case in case 3, R&D into SRM, Two Kinds of Research, where none of the three available courses of action has a worst case that is above the minimally acceptable level. The specific details of Case 3 can be found in the appendix, here only the distribution of outcomes is important, which can be seen in Fig. 7.5. “CS” refers to climate sensitivity, and the negative side effects refer to whether or not solar radiation management would work as intended as a climate engineering technology.

We could, e.g., try to argue that if there is no “safe” option, in the sense that even if everything goes wrong, we still end up with an acceptable outcome, then precaution is no longer an option and we might have to look to other decision criteria. But at this moment, it seems a bit premature to exclude these cases without even trying to explore whether there is another candidate system that can account for them.

As for whether to exclude cases where more than one worst case is acceptable, this seems even less plausible to me. At the very least, the target PP should leave it open to us in such a situation which course of action we want to choose.

So I argue that neither of the identified classes of commitments that are not accounted for by the current system should be excluded from the subject matter.

(d) Clarifying Commitments

It is unclear how the RCPP relates to IC 1 and IC 24, but arguably this is due to the lack of clarity of those commitments:

IC 1:: Where there are threats of serious or irreversible damage, lack of full scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degradation. (Principle 15 of the Rio Declaration) [low]
IC 12:: The claim that the measles/mumps/rubella (MMR) vaccine might cause autism is not a reason not to vaccinate your child. [medium]

I propose to adjust them in the following way, which helps to determine whether or not the RCPP can account for them:

C 1:: When there is a plausible threat of serious or irreversible harm to the environment, then uncertainty of the harm must not lead to postponing cost-effective measures to prevent it. [medium] [replaced IC 1 at Step B₃]
C 4:: The claim that the measles/mumps/rubella (MMR) vaccine might cause autism does not constitute a plausible threat. [medium] [replaced IC 12 at Step B₃]

These adjustments are defensible as respecting the input commitments, since they are intended as clearer formulations replacing the original, more vague, ones.

However, they are not accounted for by the RCPP, which is the current system, but are both merely consistent with it: since the RCPP does not say anything about how to identify plausible threats, it does not tell us anything about C 4. And while the RCPP does not tell us that we should postpone measures in the face of uncertain harm, i.e., it does not conflict with C 1, neither does it tell us that measures must not be postponed. The RCPP sometimes can be used to comply with the demand of C 1, i.e., to identify measures to prevent serious or irreversible harm to the environment, but it does not itself account for this demand.

7.3.2 Searching for Further Relevant Commitments

When discussing whether cases with probabilities should be excluded from the subject matter (Sect. 7.3.1), I argued that there are cases where probabilities are available, but taking precautions is warranted. This revealed further relevant commitments, and these should be added as emerging input commitments:

EC 21:: The wider the dispersal of the outcome value distribution of a course of action, the more precaution is warranted by its negative outcomes. [medium] [emerged at Step B₃]
EC 22:: Pro tanto, cases that involve threats of disproportionate harm warrant precautionary measures, even if this harm is very unlikely. [medium] [emerged at Step B₃]

7.3.3 The Adjusted Set of Current Commitments, C₃

In Fig. 7.6, you can see the adjusted set of commitments compared with the account table from step A₃, Fig. 7.1. For a full list of the text of each commitment, you can, as always, refer to Appendix A at the end of the book. The commitments that have been adjusted with respect to the current system are marked with double border cells. These adjustments slightly increase the account value between RCPP and the current commitments from 72 to 78.

7.4 Step A₄: From RCPP to Maximin-PP

By adjusting some of the weaknesses of the RCPP, the “Maximin-PP for Combinations of Uncertainty and Incommensurability (Maximin-PP)” is introduced as an additional candidate system (based on Aldred 2013). In order to increase the competition, a slightly adapted version of the PP proposal by Steel (2015) is introduced as the “Tripartite Precautionary Approach (TPA)”. After comparing five candidate systems with respect to account (Sect. 7.4.1) and their theoretical virtues (Sect. 7.4.2), the Maximin-PP is chosen in the overall comparison (Sect. 7.4.3) against the TPA, which is a close runner-up.

To adjust the current system, I again compare different candidate systems. Even though the Principle 3-System (P 3) and the Utilitarian Uncertainty Principle (UUP) were rejected at step A₃, they are again considered as potential alternatives to the RCPP, this time with respect to the current commitments C₂. However, learning from the weaknesses of the current system, the RCPP, I also assess two further candidate systems. Firstly, I propose an adjustment of the RCPP that has a wider scope and should also be more determinate, the “Maximin-PP for Combinations of Uncertainty and Incommensurability (Maximin-PP)”. Secondly, I introduce a slightly adapted version of Steel’s (2015) proposal for a PP, which I call the “Tripartite Precautionary Approach (TPA)”.

Introducing the Maximin-PP as an Adjustment of the RCPP

The small scope, low determinacy, and low practicability of the Rawlsian Core Precautionary Principle (RCPP) remains a problem. A candidate that would score higher with respect to these virtues than the RCPP (that preferably includes the RCPP in its scope) is desirable. In order to find such a candidate, let us have a closer look at the weaknesses of the RCPP.

For the RCPP, the main problem with respect to determinacy is its condition 2, which demands that “decision makers care relatively little for potential gains that might be made above the minimum that can be guaranteed by the maximin approach”. It is unclear what this means, exactly: that outcomes should be distributed in a way that the best worst case is almost as good as the overall best case? As an empirical claim about preferences of decision-makers? That in all alternative courses of action, potential gains that are above the overall best worst case are outweighed by their negative outcomes?

Furthermore, even if the criterion that rejected alternatives that have “unacceptable” outcomes is supposed to be a thick ethical claim, i.e., that to determine whether or not it is met we need to make value judgments and cannot describe it in merely factual terms, it still seems that a bit more could be said on what exactly makes an outcome unacceptable.

The scope of the RCPP is a problem, too, especially because its small range of applicability prevents it from accounting for whole classes of commitments. At this point, it seems reasonable that the target PP should be restricted to cases in which only reasonable/plausible outcomes are considered, and in which outcomes are at least comparable (even if not necessarily commensurable). But it would be desirable if we could broaden the scope of the target system beyond the RCPP’s other restrictions to situations in which no probabilities are available, all courses of action alternative to the one selected by maximin include unacceptable outcomes, and “decision-makers care relatively little about potential gains that could be made above the best worst case”.

What we want is ideally a new candidate system that scores better than the RCPP with respect to the virtues of practicability, scope, determinacy, as well as with respect to the account value; while being equal or at least similar to it with respect to simplicity (or maybe even simpler than the RCPP, too).

Starting with the conditions of application of the RCPP, is there a way to broaden their scope while at the same time rendering them more precise?

Aldred (2013) has made a proposal that looks very promising: by drawing on decision theory analyzing decisions under uncertainty, i.e., where it is not possible to attach one discrete probability to each outcome, he not only provides support for the specific criteria of the RCPP, but also argues that those criteria are just one configuration of features a decision problem can have that will justify following the maximin rule. Put differently, he proposes a PP candidate that is a qualified maximin rule, like the RCPP, but the criteria of this maximin-PP are more “flexible”, i.e., they cover a broader scope. More specifically, he proposes a maximin-PP that is limited to specific combinations of uncertainty and incommensurability.

Since there are different interpretations of what it means for outcomes to be incommensurable, let me cite Aldred’s (2013, 133) understanding: “Outcomes are incommensurable when, even in conditions of certainty, their value cannot be precisely measured along some common cardinal scale. In contrast, outcomes are incomparable when they cannot even be ranked on an ordinal scale. Thus comparability is a necessary but not sufficient condition for commensurability.”

Maximin-PP for Combinations of Uncertainty and Incommensurability (Maximin-PP)

Select the course of action with the best worst case if you are either:

In a situation of decision-theoretic risk or uncertainty (or some combination), and the outcomes of the available actions can be ranked on an ordinal scale, and all courses of action alternative to the one selected by maximin have outcomes that are incommensurably worse than the best worst case; or
In a situation of (partial) decision-theoretic uncertainty, outcomes can be ranked on a cardinal scale, and all courses of action alternative to the one selected by maximin have negative outcomes that outweigh every potential gain that could be made above the level that can be guaranteed by maximin.

The Tripartite Precautionary Approach

Since P 3-System, RCPP, and UUP have already been assessed in detail, I am adding an additional candidate from the literature, which can be assessed and compared with them. I adapt the proposal from Steel (2015), which is an elaborate and comprehensive candidate. I call it the “Tripartite Precautionary Approach”, since it provides a framework consisting of three main elements, which allows us to formulate specific versions of a precautionary principle.^{Footnote 3}

The Tripartite Precautionary Approach (TPA) consists of (cf. Steel 2015, 9–10):

The Meta Precautionary Principle (MPP): Uncertainty must not be a reason for inaction in the face of serious threats.
The Precautionary Tripod: The elements that have to be specified in order to obtain an action-guiding precautionary principle version: If there is a threat that meets the harm condition under a given knowledge condition then a recommended precaution should be taken.
Proportionality: Demands that the elements of the Precautionary Tripod are adjusted proportionally to each other, understood as Consistency: The recommended precaution must not be recommended against by the same PP version, and Efficiency: Among those precautionary measures that can be consistently recommended by a PP version, the least costly one should be chosen.

The strategy to obtain a PP decision rule by adjusting the precautionary tripod: (1) select a desired safety target and define the harm condition as a failure to meet this target, (2) select the least stringent knowledge condition that results in a consistently applicable version of PP given the harm condition. To comply with the MPP, uncertainty must neither render the PP version inapplicable nor lead to continual delay in taking measures to prevent harm (cf. Steel 2015, 10).

7.4.1 Maximin-PP, TPA, P3, RCPP, and UUP: Account for Current Commitments

In Fig. 7.7, the results from assessing account for the five candidate systems with respect to the current commitments C₃ is summarized. As before, the full list of commitments can be found in Appendix A. The TPA reaches the highest account value, 150, followed by the P 3-System with 141.5, the Maximin-PP with 127.5, the RCPP with 78, and lastly the UUP with 42.

7.4.2 Theoretical Virtues of Maximin-PP, TPA, P3, RCPP, and UUP

Since the theoretical virtues of the P 3-System, the RCPP, and the UUP have already been assessed in step A₁, I focus now on the Maximin-PP and the TPA. The rankings of the candidates with respect to theoretical virtues are described starting on p. 24, right before the overall comparison with respect to both theoretical virtues and account (7.4.2).