Inductive Risk, Deferred Decisions, and Climate

Science Advising1
Joyce C. Havstad and Matthew J. Brown

Abstract
This chapter discusses the philosophical viability of Ottmar Edenhofer and Martin
Kowarsch’s proposed pragmatic-enlightened model of science advising. Edenhofer
and Kowarsch’s (2015) model makes central use of a cartographic metaphor—one
in which scientists and policy-makers craft and consider different scientific routes
to various value-laden ethical, political, and social destinations. But the argument
from inductive risk poses a significant challenge to the viability of the metaphor,
and hence, to the workability of the model.

Introduction
The argument from inductive risk (Rudner 1953; Douglas 2000, 2009) holds that
scientists must set standards of evidence for inferential decisions in science
according to the possible consequences of error—including both false positives and
false negatives—at least in those cases where there are predictable social, political,
or ethical consequences of such error. Another way to put this point is to say that
scientific practice requires non-epistemic value-laden judgments even at those so-
called “internal” stages of science traditionally considered clear of non-epistemic
values. Evidence, logic, and epistemic values can tell us something about the
strength of support for some claim, but they alone cannot compel a scientist to
make the choice to assert, infer, accept, or endorse that claim.2 Thus, according to
the argument from inductive risk, the scientist qua scientist must make classically
normative judgments.
One major line of response to this argument holds that, in these kinds of
cases, scientists can (and ought to) defer value-laden decisions on standards of
evidence, and thus whether to accept or reject hypotheses—by simply
communicating the evidence plus its attendant probabilities and leaving it for
others (such as policy-makers) to choose the “correct” evidential standards (see
Jeffrey 1956; Mitchell 2004; Betz 2013). This deferred-decision response
acknowledges the value-ladenness of decisions to accept or reject hypotheses (and
other decisions), but attempts to take those value-laden decisions out of the hands
of scientists. For reasons already articulated philosophically by Rudner (1953) and

1
Authorship is joint; the authors are listed reverse-alphabetically.
2
So long as the claim is a non-trivial, empirical claim that amounts to an ampliative
inference.
expanded upon by Douglas (2009), this line of response to the argument fails. Very
briefly, those reasons include: that the assertion of claims about “the evidence plus
its attendant probabilities” is itself subject to inductive risk; that the evidence thus
presented is the result of inductively risky processes of data collection,
characterization, and aggregation; and that even if scientists attempt to arbitrarily
defer certain inductively risky decisions toward the end of their assessments, these
decisions are inevitably preceded by a series of other inductively risky decisions
requiring non-epistemic value judgments. Scientific practice necessarily
incorporates a complex series of judgments whose complete deferral is
unattainable. Furthermore, the technical complexity of many of those decisions
makes the elimination of expert judgment impractical and undesirable.
Despite these objections, the deferred-decision style response to the
argument from inductive risk continually reappears in new proposals for models of
science advising, including models that have recently been designed for and
applied to the case of climate science. For instance: Roger Pielke, Jr.’s (2007)
honest broker of policy alternatives and Edenhofer and Kowarsch’s (2015)
pragmatic-enlightened model (PEM) are two examples of newly proposed models of
science-policy interaction, both of which are meant to reform climate science
advising. Yet because these newly proposed models each presume that deferral of
value-laden scientific decisions is a viable response, they both fail to appropriately
accommodate the problem of inductive risk. These ideals adopt a posture of
neutrality, but they must in practice present a narrow, greatly simplified space of
options. The presentation of options under the guise of neutrality serves to obscure,
rather than to highlight, the value-laden series of decisions that precedes the
presentation of those same options.
In this chapter, we focus on the PEM. This model is especially worthy of
philosophical scrutiny because of its practical application: it was designed by the
leaders of, and has been at least partially put into practice by, Working Group III
(WGIII) of the Intergovernmental Panel on Climate Change (IPCC). The focus of
WGIII of the IPCC is on mitigating the effects of climate change—an area where the
likely social consequences of judgments of inductive risk are particularly salient.
Most notably, the designers of the PEM seem to accept most of the inductive risk
argument for the value-ladenness of science; but their response to this argument
makes the same mistaken appeal to deferral of value-laden decisions as have so
many others before it.
In what follows, we provide an initial sketch of the PEM. Then we
demonstrate how—although the PEM improves on traditional models of science
advising, such as the linear model and that of evidence-based policy—it
unfortunately incorporates the untenable deferred-decision response to the
argument from inductive risk. We thus argue that the PEM does not adequately
model how to integrate non-epistemic values with climate science and policy. Next
we suggest amendments to the PEM, in order that it might truly meet the challenges

2
of the argument from inductive risk, and begin to comprehensively confront the
value-ladenness of climate science. We conclude by summarizing our critique as
well as the ongoing value of the PEM, and in closing we assess the ongoing
popularity of deferred-decision style responses to the argument from inductive
risk—cautioning against further pursuit of this style of response.

The Pragmatic-Enlightened Model (PEM)

The PEM was recently proposed (in 2012) by Ottmar Edenhofer, director of WGIII
of the IPCC, and Martin Kowarsch, head of the working group on Scientific
Assessments, Ethics, and Public Policy at the Mercator Research Institute on Global
Commons and Climate Change.3 Edenhofer and Kowarsch’s “pragmatic-
enlightened” model attempts to carve out a conceptual space for scientific results
that is objective yet value-laden, while remaining true to the IPCC’s commitment to
providing science advice that is “policy-relevant and yet policy-neutral, never
policy-prescriptive.”4 This commitment of the IPCC—to providing neutral and
relevant but not prescriptive advice on climate policy—is a central one, and it’s
one that was negotiated quite early in the development of the organization’s
mission.5
It’s also a commitment that’s particularly hard to satisfy, given the argument
from inductive risk.6 Inquirers always need some kind of practical reason to accept
a certain level of evidential support as sufficient. Purportedly neutral attempts
remain value-laden; it’s just that the values involved are unsuitably veiled or
implicit, rather than properly reflective and explicit.7 This consequence of the
argument from inductive risk is why it poses such a powerful challenge to the
value-free ideal—an ideal that, for better or for worse, has historically
characterized and continues to characterize so much of the scientific ethos.

3
As far as we know, the PEM was first sketched by Edenhofer & Kowarsch in a working
paper from 2012, called “A Pragmatist Conception of Scientific Policy Advice.” An updated
version of the model was recently published, in 2015, in a paper titled “Cartography of
Pathways: A New Model for Environmental Policy Assessments.” Ongoing references to the
PEM will be to the model presented in the published version of the paper.
4
Quoted from: [Link]
5
For more on the negotiation of the IPCC’s commitment to relevant yet neutral, non-
prescriptive climate policy, please see Brown and Havstad, “The Disconnect Problem,
Scientific Authority, and Climate Policy” (2017).
6
For more on the meaning of the IPCC’s commitment to relevant yet neutral, non-
prescriptive climate policy, please see Havstad and Brown, “Neutrality, Relevance,
Prescription, and the IPCC” (forthcoming in Public Affairs Quarterly).
7
This is not a causal-explanatory claim, what Daniel Steel describes as the “descriptive
interpretation” of the argument from inductive risk (Steel, forthcoming). Rather, even if no
conscious or unconscious value judgment guides the decision, there is always a practical
decision to accept a certain amount or type of evidential as sufficient, and as such the
decision implies a practical reason (value). In this sense, all decisions subject to the
normative argument from inductive risk are, as a matter of descriptive fact, value-laden.

3
 Though the norms of objectivity, neutrality, and the value-free ideal may
effectively guard against bias and corruption in many parts of science, as the
argument from inductive risk shows, these norms may actually be hiding bias and
encouraging illicit importation of values.8 This is because the value judgments are
essential to the scientific practice, yet when they’re made they’re hidden, in a
mistaken effort to maintain the appearance of allegiance to the value-free ideal, or
implicit, in a heads-in-the-sand way. In a sense, the deep challenge of the
argument from inductive risk is to appreciate what the norms of objectivity,
neutrality, and the value-free ideal have done for science while simultaneously
constructing a method for functioning without them in cases where they simply
cannot apply—like those cases covered by the argument from inductive risk.9
This is the challenge faced by Edenhofer and Kowarsch (hereafter E&K) in
their attempt to (on the one hand) create a model of climate science advising that
produces relevant but neutral, non-prescriptive policy advice, while they also (on
the other hand) acknowledge that the argument from inductive risk forcefully
applies in this arena. As they do much to try and meet this bivalent challenge, there
is much to like in E&K’s model.
On the latter hand, E&K seem to unreservedly admit that scientific
knowledge is thoroughly value-laden. This admission is expressed in at least three
distinct ways. First: the designers of the PEM cite a variety of approaches from the
philosophy of science to confronting the value-ladenness of science—including
Douglas’ inductive risk approach and Putnam’s “thick ethical concepts” approach
(E&K 2015, 57). E&K especially seem to fully accept what Putnam (2002) calls “the
entanglement of fact and value” (E&K 2015, 59). Second: E&K espouse a
thoroughgoing pragmatism, and their model relies heavily on not just the neo-
pragmatist philosophy of Putnam, but also on the classical American pragmatism of
John Dewey (E&K 2015, 58). Third, and perhaps most radically: in a truly Deweyan
vein, E&K insist on significant stakeholder engagement as part of the scientific
process (E&K 2015, 61).

8
Douglas (2000, 577–8) divides science into four areas: (1) where there is low uncertainty,
or chance of error; (2) where making a wrong choice, or erring, has no real impact outside
of research; (3) where the science will likely be useful, but non-epistemic consequences of
error are hard to predict; and (4) where there are predictable non-epistemic consequences
of error. In the first two areas, the argument from inductive risk is not a salient
consideration, and scientists generally will not need to consider non-epistemic values. In
the third area, the argument from inductive risk is likewise not so salient, as the scientists
cannot make value judgments about consequences they cannot reasonably foresee. This is
a so-called “gray-area,” and in this area the matter will have to be decided on a case-by-
case basis. (Such situations do suggest the need for scientists to develop their moral
imagination as far as reasonably possible). In the fourth area non-epistemic values are,
Douglas argues, a necessary part of scientific reasoning.
9
This challenge is taken up by Douglas in “The Irreducible Complexity of Objectivity”
(2004) and in her Science, Policy, and the Value-Free Ideal (2009), especially chapter 6.

4
 Elaborating on the second point: advocates of the inductive risk argument
against the ideal of value-free science are generally committed to a very basic form
of pragmatism, insofar as they deny that one can always and completely separate
thought and action. In other words, adherents of the argument are at least in some
form committed to rejecting the claim that one can, in general and correctly,
evaluate beliefs or decide whether to accept, infer, assert, or endorse a hypothesis
without taking into account the implications of that belief or hypothesis for
decision-making and behavior. But E&K do not merely make this broad and vague
commitment to pragmatism. Rather, E&K go much further and commit themselves
to a fully pragmatist, Deweyan theory of inquiry.10
For E&K’s purposes, the key points of the Deweyan theory are that: (a)
inquiry is a means of resolving problematic situations; (b) the evaluation of a
hypothesis is primarily done in terms of its practical problem-solving success
(potential and then actual); (c) the results of inquiry are monitored and judged after
the fact, in their implementation; and (d) the entire collection of practical
objectives (both those whose problematic execution spurred inquiry in the first
place, as well as any standing objectives incidentally affected by proposed
hypotheses) are to be considered as inputs to, and evaluated for their validity, in
the process of inquiry.11
E&K furthermore subscribe to the pragmatist meta-ethical commitment of
the “interdependency of objectives and means” (2015, 58), articulated by Dewey
and others as the “continuum of ends-means” (Dewey 1939; Waks 1999; Anderson
2014). On this view, there are no a priori, self-justifying, ends-in-themselves.
Rather, we find ourselves in our activities with certain ends-in-view—themselves
actual or potential means to further ends—as well as with ideas about the means to
those ends. As inquiry is spurred by problems, in trying to reach those ends, not
only the means but the ends too may themselves be revised. As a result and on this
view, in the course of policy-relevant scientific inquiry many components of the
processes—facts, data, hypotheses, plans, policies, values, goals, etc.—are all put
up for grabs.12
Finally, with respect to the third point: the PEM commits itself to significant
stakeholder participation not just at the very end of the pipeline, when it comes
time to evaluate policy proposals, but at every stage of the inquiry—including the
definition of the problem requiring scientific analysis and policy amelioration. This
accords well with Deweyan conceptions of democracy as well as with Douglas’

10
Elaborated on page 58 of E&K (2015).
11
For more on Dewey’s theory of inquiry, see Hickman (2007), Dorstewitz (2011), and
Brown (2012).
12
To use E&K’s cartographic metaphor: when the map doesn’t help us navigate where we
want to go, it is not only our choice of map but also our choice of destination that is up for
grabs, depending on further facts of our situation. More on E&K’s cartographic metaphor
very shortly.

5
(2009) insistence on a democratic analytic-deliberative process for policy-relevant
science.
And yet, in the former hand lies the fundamental tool of the PEM—E&K’s
response to the other half of the bivalent challenge. This is where E&K’s models
differs from generally pragmatist models and standard inductive risk accounts: in
the idea of “mapping out” a set of alternative, scientifically viable “policy
pathways” that scientists, policy-makers, and stakeholders can then consider and
adjudicate among (E&K 2015, 60).13 These pathways are to be built by combining
and mutually revising scientific evidence in concert with various policy means,
objectives, and value judgments into potential policy solutions (E&K 2015, 60–1).
Furthermore, once policies are implemented, the PEM requires that their
consequences be carefully monitored, and the cartography of policy pathways re-
applied, based on the analysis of those consequences (E&K 2015, 61 and 63). So
these alternative policy pathways are also intended to respect the interdependency
of objectives and means—and thus they are meant to include not only policies,
along with the relevant scientific information, but also the objectives associated
with those policies, and the expected (social) consequences of the implementation
of those polices. In other words, the pathways incorporate value judgments both in
the sense of policy goals and side-constraints on acceptable science and policy.
For instance: E&K are fundamentally concerned, as we are, with the
development of viable policy pathways for addressing the problem of global
climate change. They instantiate their theoretical model of pathway-creation with a
proposal for climate science advising that imagines the “differential costs, risks,
climate impacts as well as co-effects” (E&K 2015, 62) that might be associated with
a 1.5°C, 2°C, or 3.5°C rise in global temperature. Considering different temperature
outcomes gets assessment going down several different pathways. Considering not
just impacts but also costs, risks, and co-effects ensures that assessment combines
scientific predictions with predictions about the social, political, and ethical
implications of going down particular pathways. And finally, considering the
consequences of taking various pathways requires monitoring, revision, and
adjustment. E&K are quite clear on this last point: “Mapping policy pathways in
assessments is an iterative exercise that frequently requires adjustment if new forks
in the road, alternative destinations, pitfalls and uncharted territories turn up”
(2015, 63). More on E&K’s proposed instantiation of the PEM in the case of global
climate science and policy follows in the next section.
But returning now to their theoretical apparatus: providing a range of
options is supposedly pragmatic; allowing stakeholders, etc. to choose among these
options is meant to be democratic. The role of values in science is purportedly

13
Note the cartographic terminology. E&K speak, throughout their (2015) paper, of
mapping, pathways, routes, and the like. Much of the intuitive appeal of the PEM lies in the
suggestive force of this metaphor. As such, we will employ and extend E&K’s cartographic
metaphor in our own critique of the model (developed in the next section).

6
respected, but those doing the scientific advising on policy alternatives need not
commit themselves to or apply controversial values; rather, those advising with
respect to science and policy merely provide alternative assessments, taking
various alternative values into account, and others more appropriately positioned
to choose amongst the candidate value-laden options can then do so. The
cartography of policy pathways putatively avoids the problem that Pielke (2007)
refers to as “stealth issue advocacy”—by offering a variety of options—while
simultaneously incorporating a more realistic and normatively adequate account of
the role of values in science than does Pielke’s own model of the honest broker of
policy alternatives (at least according to E&K 2015, 57). To be clear, each
“pathway” is supposed to instantiate responsible, value-laden science in the sense
the inductive risk argument demands; the PEM’s novelty comes in providing a
range of different policy pathways, given different possible objectives and values,
and deferring the decision between them to the policy-makers.
We hypothesize that the creation of this value-incorporated cartography of
alternative policy-pathways is E&K’s way of attempting to accommodate the
argument from inductive risk while simultaneously following the IPCC’s mandate to
adopt a model of climate science advising that produces relevant but neutral, non-
prescriptive policy advice. In order to avoid the charge of policy prescriptiveness,
and retain the IPCC’s stated aim of policy neutrality, the PEM recommends charting
a variety of alternative policy pathways. The role of values in assessment of policy
alternatives is admitted by the PEM—but where values are controversial, the model
requires building in alternative pathways according to alternative sets of values.
This is where the deferred-decision response to the argument from inductive risk
kicks in; it is also where the PEM falters.

The Impossibility of Deferring Value-Laden Decisions

On the face of it, the PEM offers a happy compromise: values are incorporated into
climate science advising; yet it’s not the (unelected, unrepresentative) scientists
who are making the non-epistemic value judgments. The scientist-advisers are
merely offering a selection of value-incorporated scenarios for those (elected,
appointed) public representatives who can properly choose among the attendant
values to do so, and in an informed way that guides policy-making according to
which non-epistemic values are the chosen ones and in combination with the
relevant scientific facts.
But the PEM fails, in philosophical and practical terms, in both its
attempts—to avoid policy prescription and to respect the role of non-epistemic
values in those scientific cases where the argument from inductive risk applies. In
philosophical terms, the nature of this joint failure is best understood in terms of the
inductive risk argument and the relationship it reveals between uncertainty, choice,
consequences, responsibilities, and normative value judgments.

7
 The inductive risk argument relies on the endemic and pervasive
uncertainties in science to show the need for non-epistemic value judgments
throughout the course of scientific research. The fundamental requirements of
empirical adequacy and logical consistency alone cannot compel scientists to do
things one way rather than another—to choose this over that methodology, this
characterization of ambiguous data over that, this or that conceptual framework, a
higher or lower standard of evidence, etc. Rather, the scientific process involves a
series of unforced choices which lead to results that, while significantly constrained
by logic and evidence, are still highly contingent on the set of prior choices made.
Such choices often incorporate, either directly or indirectly, non-epistemic as well
as epistemic value judgments.
As the last half-century or more of philosophy of science has shown, these
value-laden choices are generally made on the basis of a mix of background
assumptions, methodological conventions, tacit knowledge, research tradition, etc.,
but they are nonetheless choices, in the most basic sense, that could be made
differently than they are.14 While many proponents and detractors of the inductive
risk argument focus only on the final choice in the course of a scientific inquiry—
that of accepting, believing, or certifying a hypothesis or theory on the basis of the
evidence available—in fact, scientific inquiry consists in a cascading series of such
value-laden choices.
The import of the argument from inductive risk is often limited: sometimes
uncertainty is so low that the chance of error is vanishingly small; sometimes the
chance of error is socially negligible because the relevant science has no ethically
salient aspects; and sometimes the socially significant consequences of error simply
cannot be anticipated (Douglas 2000, 577–8). But in those cases of significant
uncertainty where readily foreseeable social consequences will pertinently follow
from those innumerable value-laden choices made throughout the scientific
process, the argument from inductive risk does have significant import for the
choices made throughout the process.
Climate science is obviously a case in point: there is significant uncertainty,
especially with respect to the upcoming effects of anthropogenic global climate
change; it is undeniable that some of these upcoming effects will have social
consequences; and some of these looming social consequences are readily
foreseeable. As a result, the choices made by climate scientists (and climate
science advisors) throughout the scientific (and political) processes are covered by
the argument from inductive risk.

14
Notably, researchers may often not see themselves as choice-makers. They can, however,
be made to recognize this feature of their work, and this awareness can improve the social
and ethical responsibility of the science in question. Erik Fisher’s program of Socio-
Technical Integration Research (STIR) consists in a process he calls “midstream modulation”
aimed at doing just that—raising consciousness among scientists of their role as socially
responsible choice-makers (Fisher et al 2006; Fisher 2007; Fisher and Schuurbiers 2013).

8
 One of the premises of the argument is that “scientists have the same moral
responsibilities as the rest of us” (Douglas 2000, 563). The argument (elaborated in
Douglas 2009) reminds us that scientists are also moral agents—they don’t shed
their ordinary moral responsibilities when they step into the lab, the field, or the
conference room. Insofar as climate scientists make choices, and the social
consequences of those choices are reasonably foreseeable, climate scientists have a
responsibility to weigh those consequences when making their choices, lest they be
morally negligent.15 And this process of weighing social consequences is just what
making a value judgment consists in.16
Putting these points together—points about the integration of choice
throughout the scientific process, the kinds of cases where the argument from
inductive risk applies, the obvious fact that climate science is one of these cases,
and the presumption that scientists are not excused from their normal moral
responsibilities simply because they are scientists—raises a philosophical problem
for the PEM. E&K seem to think that scientific advisors can and should defer their
value-laden decisions throughout the scientific and advisory processes relating to
anthropogenic global climate change; but this is impossible. Such decisions are
inexorably laced throughout the relevant processes, and it does a disservice to
scientific practitioners and practice alike to pretend that such decisions neither can
nor should be made.
In practical terms, although E&K pay lip service to the philosophical point
that scientific work is thoroughly value-laden, they are much less precise when it
comes to how making such value judgments actually works in science. This lack of
precision becomes a significant problem when they get down to the dirty work of
constructing the alternative policy pathways that the PEM is supposed to offer.
Consider E&K’s set of three potential climate policy pathways, which happen to
correlate with the aforementioned trio of potential raises in global temperature:

15
This holds only, of course, when the choice is genuine and unforced. Insofar as evidence
and logic constrain a scientific (or other kind of) choice, weighing consequences of the
(forced) choice is just to fall into wishful thinking or outright deception.
16
We prefer to read social “consequences” here in a maximally broad way, as there are
competing value systems on offer, and utilitarianism is certainly not the only option—it is
not the case that all the candidate options are broadly consequentialist, even. On such a
broad reading of the term, even something like violating a person’s rights counts as a
negative social “consequence.”

9
 Figure 1: A reproduction of E&K’s Figure 3 “Potential key dimensions of
future IPCC WG III assessments” (2015, 62). In addition to demarcating
three potential temperature outcomes, E&K also identify four different
“assessment dimensions” and delineate five groups of affected parties for
each outcome. Reproduced by permission.
Offering this set of three potential climate policy pathways presents the role of
values in climate science as somehow limited to the question of which of these
three global temperature change outcomes scientists ought to consider when
assessing climate research and designing policy responses. It also raises more
questions than it answers—about how E&K or anyone else could determine who
might be affected by such global temperature changes, and how. It is unclear why
E&K identify the assessment dimensions that they do, and whether the potentially
affected parties ought to be demarcated as E&K have done so here. Yet what should
be clear is that, in considering possible temperature outcomes as well as
assessment dimensions and groups of affected parties, scientists don’t just face just
one or even a small number of choices. Instead—as E&K themselves at times admit,
and as we stressed during our philosophical critique of the “cartographic” aspect of
the PEM—scientists face a complex series of amplifying and interrelated choices
with respect to everything from what the potential temperature outcomes are to
who ought to be grouped together as affected by such outcomes.
This problem of choice amplification and interrelation is masked by the fact
that E&K discuss these policy pathways, based in varying long-term global
temperature rise targets, in only the most general and abstract of terms, rather than
delving into the concrete details of these choices. The climate projections that form

10
these pathways are based in enormously complex computer models, both climate
models and “integrated assessment models” (IAMs) (which include not only
physical processes, but model technical and human systems and impacts, see IPCC
2014, 51). As Eric Winsberg (2012) points out, “climate modeling involves literally
thousands of unforced methodological choices” (130), a result of the fact that such
models are highly idealized and incredibly complex, and doubly so when we
consider IAMs that include human factors like economic systems and agricultural
development. Such choices might include decisions about different possible
parameterizations and model structures, particular parameter values, choice
between different approximation methods, decisions about which climate forcings
to include in the model or exclude as insignificant or approximate with a simple
parameter, choice of higher or lower model resolution (or grid size), decisions
about aggregating ensembles of models, and so on.
By unmasking the complexities of climate change modeling, and
combining these complexities with the unresolved issues of outcome and
assessment demarcation, it becomes easy to see that E&K’s “cartographic” project
quickly runs into a multiplying effect. For every new climate modeling choice—
how to model cloud formation, say, or the effect of melting sea ice17—there are
multiple options, multiple ethical and political considerations relevant to those
options, multiple dimensions of uncertainty about just what the consequences will
be, multiple ways of grouping sets of such consequences according to different
dimensions, multiple ways of considering who will be affected by such
consequences, and so on. By carefully considering what it means to really get into
these permutations, we can see just how flawed the cartographic project actually
is.
Applying and extending E&K’s own metaphor: it is not as though there are
only three routes up, say, Mount Everest.18 Rather, there are many potential routes
that any given climber can take—perhaps a handful of well-supported ones—and
on any given day—depending on the season, the weather, how many people are
going up the route, how skilled the guides are, how much and what kinds of
assistance the climber accepts, and many more (some unknown) variables—the
“route” taken might have a wildly variable chance of successfully getting the
climber up the mountain, and may or may not be “safe.”
So what does the “map” look like now? Instead of three neat “pathways,”
there are dozens, each with different degrees of solidity, each "route" with different
chances of leading to multiple “destinations.” And if there are any outcomes that
the climber absolutely wants to avoid (say, death while attempting to summit) or
the climber absolutely insists must result (say, making the summit), then this is
drastically going to affect the selection of potential routes genuinely available to the

17
See Biddle and Winsberg (2010) and Winsberg (2012).
18
Also known as Sagarmāthā (in Nepal) or Chomolungma (in Tibet).

11
climber, and the risk that they're willing to assume—of false negatives or false
positives—with respect to the likelihood of those outcomes as projected by their
“pathways.”
Now reconsider the main application of the PEM that E&K pursue. Our
figure 1 (their figure 3) represents E&K’s attempt to apply the PEM to the case of
climate science advising—specifically, to their work on potential mitigation of the
effects of global climate change, via policy recommendation by WGIII in the
current cycle of IPCC reports (AR5).19 As shown in the figure, and discussed above,
E&K have chosen to consider three possible mitigation targets in terms of global
temperature rise (1.5°C, 2°C, and 3.5°C). But of course there are more than three
potential outcomes, even using global temperature change as the only relevant
dimension. There is a gradual range of possibilities here: global temperature alone
could change by anything between a reasonable lower bound (considering
mitigation alone, probably not 0°C, at this point) to some unknown but large upper
bound. In the Synthesis Report for the previous IPCC cycle (AR4), for instance, six
different potential temperature outcomes were considered, given in ranges (2.0–
2.4°C, 2.4–2.8°C, 2.8–3.2°C, 3.2–4.0°C, 4.0–4.9°C, and 4.9–6.1°C).20 So, why do
E&K think their three potential outcomes are the relevant ones for WGIII in this
IPCC cycle (AR5)? Furthermore, why consider global temperature change as the
relevant dimension at all? (More on that latter question in the next, penultimate
section).
Perhaps it is unfair to pick on E&K for suggesting policy pathways
associated with just these three targets for global temperature rise. After all, E&K are
simply trying to give us a sense of how the assessment process would work
according to the PEM. Actual application would involve a more complicated
picture. But it is precisely what happens when we move from E&K’s vaguely
specified example of three temperature targets to a more realistic picture that shows
how the PEM is unworkable. The increasing complexity, and the associated value-
relevance of each decision, show that crafting alternative policy pathways that
defer all significant value judgments to policy-makers is either unworkable,
because of the problem of the multiplying effect, or a wretched subterfuge, because
the vast majority of relevant value judgments have been made prior to presenting a
small number of policy pathways to decision-makers. The overwhelming variety
and particularity of the value judgments that must be made in the course of this

19
In their contribution to the IPCC’s Fifth Assessment Report (AR5), WGIII’s mission is
rather narrowly defined as providing “a comprehensive and transparent assessment of
relevant options for mitigating climate change through limiting or preventing greenhouse
gas (GHG) emissions, as well as activities that reduce their concentrations in the
atmosphere” (IPCC 2014, vii).
20
See Table 5.1 in the Synthesis Report (IPCC 2007, 67) from the IPCC’s Fourth Assessment
Report (AR4).

12
kind of inquiry makes the PEM unworkable in the case of climate change and, we
suspect, in many other important environmental assessment situations.
To demonstrate this point, let’s pretend that E&K’s three temperature-based
target outcomes are the obvious and only relevant ones. For each of these three
potential results, E&K present four kinds of factors (“assessment dimensions”) that
are relevant sources of scientific information for each target: (i) available policy
instruments, institutions, and technologies; (ii) costs and risks of mitigation and
adaptation policy options; (iii) climate impacts and vulnerabilities; and (iv) what
E&K call co-effects on additional policy objectives, which we can think of as side
effects or social consequences of climate policy measures not directly related to
climate or environment. Even presuming that these four are the only candidate
sources of relevant scientific information, what the philosophical critique of the
PEM reminds us is that, for each of these four factors, there are presumably dozens
if not hundreds of key value-laden choices to be made. Even imagining that there
are only something like ten such choices for each of the four streams of scientific
information, and given that E&K also identify five groups (once again, a likely
underestimate) of stakeholders (international; national; states; cities; actor groups)
for whom relevant consequences and values must be chosen, those “low” numbers
result in at least (4 x 10 x 5 =) 200 permutations of the relevant choices or
considerations per each of the three “cartographic” options. That’s 600 potential
“routes”—and we know that this is an underestimate along every relevant
dimension.
E&K tell us that “the PEM-guided cartography of the political solution space
is clearly an immense and time-consuming effort” (2015, 63), but the situation is far
worse than that. It is an impossible fiction.

Amending the PEM

E&K as much as recognize the failure of their “cartographic” model, in their own
presentation of the PEM, when they talk about narrowing the selection of policy
pathways:
…the scope of possible future pathway analyses has to be narrowed down
because of the vast range of environmental policy pathways and related
consequences and the limited resources available for assessment
processes. However, there is a danger of being biased in this selection of
pathways… (2015, 60)
This “narrowing down” is how E&K keep the PEM from recommending that the
IPCC’s WGIII adopt the absurd position of “mapping” 600 or more potential
“pathways.” But it also transforms E&K’s valiant effort to meet the bivalent
challenge—of combining the IPCC’s mandate of policy-neutrality with genuine
accommodation of the argument from inductive risk—into a failure on both counts.
While the narrowing down process is still supposed to result in options that “reflect
several politically important and disputed objectives, ethical values and prevalent

13
policy narratives” (E&K 2015, 60), there is no plausible account of the “narrowing
down” process that avoids making any non-epistemic, policy-prescriptive value
judgments. Neither has the argument from inductive risk been properly
accommodated, since the process makes non-epistemic, policy-prescriptive value
judgments in such a disguised, implicit, and unreflective way.
All is not lost, however. The means to amend the PEM are suggested by
E&K’s own presentation of the pragmatist principles behind the design of their
model. Recall our discussion of E&K’s commitment to the notion that science is
thoroughly value-laden—especially, the third way in which E&K expressed this
commitment. Taking seriously the Deweyan idea of the continuum of means-ends,
E&K purportedly designed the PEM with significant stakeholder participation in
mind: not just at the very end of the pipeline, when it comes time to evaluate
policy proposals, but at every stage of the inquiry. According to E&K’s own
presentation of the overall process and their discussion of what should ideally
occur, stakeholders and policy-makers ought to be involved in, for instance, the
process of selecting alternatives to narrow down the list of possible pathways. We
suggest that actually implementing stakeholder participation at this and all other
stages of the process of climate science and policy advising is the crucial factor that
can rescue the PEM—although this implementation will result in significant
attendant amendments to E&K’s model as presented.
What would such a stakeholder-engaged process look like? Again,
consider a figure from E&K’s own work:

Figure 2: A reproduction of E&K’s Figure 2 “The PEM model” (2015, 61).

Reproduced by permission.
In the figure, both researchers and stakeholders are present at the initial stage in the
process, that of problem analysis (upper-left box). In the caption to the figure, E&K
attempt to add depth to the process depicted:

14
 The PEM as a model for solution-oriented assessments suggests that after
researchers and stakeholders have jointly framed the problem, they explore
the objectives, means and consequences. The two white boxes indicate
steps in the policy process that are outside the assessment-process per se,
such as public debate on alternative policy pathways, as well as policy
decisions and implementation by policymakers. Next, there is a scientific
ex-post evaluation of the actual means-consequences, which is also the
starting point for a new assessment cycle. (2015, 61)
Although researchers and stakeholders are identified as working together in many
aspects of the process as E&K have depicted and described it, it is less clear how
they are meant to work together and what such an engaged process would look
like, post-engagement. Neither the simplistic flow of this diagram, nor the blithely
optimistic caption really takes into account what genuine stakeholder involvement
at this and other parts of the process might entail or produce.
For one, it seems naïve to presume that stakeholders will likely think that
the relevant outcomes of global climate change will come in a set of three distinct
options—particularly, options having to do with a 1.5°C, 2°C, or 3.5°C increase in
global temperature. Those are relative measurements that are, presumably, loosely
and uncertainly correlated with a set of actual phenomena that stakeholders are
much more likely to care directly about—such as loss of coastal property value due
to rising sea levels, increased chance of tropical storms, rising fuel prices,
constraints on energy use for personal consumption and economic development,
higher food costs, national and international instability, loss of environmental
resources and diminished biodiversity, etc. Some of these phenomena may be
inexorably incurred by certain global temperature changes; others may be
individually mitigated despite the overall temperature change occurring; still others
may be more or less readily adapted to. Because these phenomena can come apart
from one another and from overall changes in global temperature, it seems highly
unlikely that stakeholders (or anyone, really) would want to only, or even primarily,
pursue unequivocal policy solutions—solutions which respond merely to the
relative measure as opposed to the many diverse effects of the comprehensive
phenomenon of global climate change. E&K’s choice of outcomes seems motivated
by technical features of the modeling process rather than genuine value judgment
or stakeholder interest.
For another, it’s not just the contents of any one box that are up for grabs—
rather, it’s the shape of the whole diagram. Commitment to genuine, integrated
stakeholder participation ought to call into question any presuppositions with
respect to what the shape of the relevant inquiry will look like. Stakeholders should
be involved in a discussion of the process in which they will participate;
stakeholder participation should be expected to change the shape of the process as
it goes. It’s unclear why E&K think they know what the process will look like,
before any substantive incorporation of stakeholder participation. Conceivably,
stakeholder participation might call into question the notion that policy

15
implementation really is “outside the assessment-process per se” or that satisfactory
problem analysis in terms of large-scale global climate change could really occur
entirely prior to evaluation of local efforts to craft and implement successful small-
scale and context-specific mitigation or adaptation strategies.
Or, stakeholder participation might undermine the presupposition that
public debate with respect to properly narrowed down cartographic options ought
to occur only at one, intermediate stage in the process of inquiry. Especially taking
into account the inductive risk argument and the deeply pragmatist commitments
of the PEM, the way that E&K focus their model on producing multiple policy
options at one particular stage seems unprincipled and arbitrary. Why are only
these options subject to public debate? Perhaps stakeholder participation will
reveal that the really crucial value judgments and choices, from the public’s point
of view, are ones that play a role much earlier in the process—in the problem
analysis, or data characterization, or narrowing of options. Having admitted that it
is not possible to fully defer the decisions about value judgments and incorporate
all controversial value-laden questions into pathways in the cartographic project,
the solitary placement of three policy options for public debate at the center of the
process seems to be a rather haphazard result, and one that is unlikely to survive
significant stakeholder involvement.
The need for stakeholder involvement raises crucial practical questions
about what stakeholder engagement processes would look like. Detailed discussion
of such processes is beyond the scope of the present essay, and while E&K
emphasize the importance of stakeholder engagement, they likewise are unable to
provide details about how such engagements would work.21 There is, however, a
rich and developing literature on public and stakeholder engagement and
participation in science and science-based policy, which offers many proposals
and raises many problems (see Stern and Fineberg 1996; Douglas 2005; M.B.
Brown 2009, Chs. 9-10; Douglas 2009, Ch. 8; Elliott 2011, Ch. 5; M.J. Brown
2013). What is most relevant to our discussion is not the particular form of the
public and stakeholder engagement process, but rather the role that engagement
plays in the scientific inquiry and assessment. Namely, the public and stakeholders
must be the ultimate source of the objectives, outcomes, and values that must
guide the process. Because of the complexities of the process, a variety of
appropriate processes may be expected to play a role—some more direct, others
more representative, some more and less deliberative, etc. We suspect any
guidance here to be fairly context-dependent. Certainly, as Douglas (2003) argues,
scientists will often have to face the burden of making value judgments themselves,
as they are the only ones with the technical competencies to do so. Stakeholder
engagement in some broad sense is necessary for the legitimacy of those

21
E&K also acknowledge that such a discussion is important, but they punt for lack of
space. They do refer to the fact that “the IPCC WG III conducted an interesting multi-
stakeholder meeting” in Washington, DC in 2012 (E&K 2015, 62).

16
judgments, but scientists cannot avoid the burden of judgment—the decision
cannot be deferred.
In sum: we encourage E&K to revisit what is entailed by their commitment
to the utterly value-laden, deeply pragmatist, and stakeholder-enriched
consequences of the argument from inductive risk. Is it really possible to narrow
down a range of possible “pathways” without making non-epistemic value-laden
judgments? Can even just the proper dimension on which to represent possible
outcomes be determined without a thoroughly pragmatist interrogation of not only
the ends-in-view but also the interdependency of means with these ends? Would
most stakeholders even want non-prescriptive recommendations of policy?
We think that the answer to each of these three questions is likely no. But
even if we’re wrong about that, anyone committed to the value-ladenness of
science, the properly pragmatist nature of inquiry, and the necessity of stakeholder
involvement is also committed to the importance of asking these questions and not
presuming to know the answers for certain before acquiring sufficient data to
resolve the inquiry. In short, we’re asking for stakeholder participation to begin in
earnest—in order to address the comprehensively value-laden nature of climate
science, and with the possibility open of needing to revise expectations about how
the process of climate science advising is going to proceed once stakeholder
participation begins.

Concluding Remarks
Despite our objections to the deferred-decision style response to the argument from
inductive risk, we think that E&K’s PEM can still be a valuable tool. But this, we
have argued, requires initiating stakeholder participation.
In the case of climate science advising, this correspondingly entails a shift
in focus: from thinking of IPCC reports as a finished product to thinking of these
reports as intermediate stages in an ongoing process of science-informed
policymaking. In general, if science advisors are going to be consulting regularly
with stakeholders and policy-makers in the course of their assessment process
(because they will need to be considering values and re-evaluating policy
objectives in the course of this assessment process), then they will also require
some mode of communication about the options in place. Far from trying to
represent a choice between finished “policy pathways,” the PEM could be used to
represent the likely consequences for various choices in the middle of the
assessment process. Stakeholder input and political discourse could help decide
the best among those options, and the process could be iterated again with the next
set of key choices.
But this alteration in thinking requires a parallel adjustment to the PEM.
Even though E&K claim to think of the relationship between assessment and policy
as part of an iterative feedback loop, their purportedly pragmatist model still
contains a remnant of the old decisionist approach. On the decisionist model,

17
assessment does its job, hands over the information to the policy-makers, policy-
makers decide based on policy objectives and public values, and finally the chosen
policy is implemented. By thinking of the map of policy pathways as the end result
of the scientific assessment process, the PEM fails to integrate science and
policymaking to the degree necessary to make dynamic, scientifically-informed
policy choices in response to these problems. Yet as E&K themselves admit, the
process requires constant feedback and only comes to a close when a policy is
implemented and found, by subsequent monitoring, to be a stable solution to the
particular problem it was put in place to resolve.
Finally, our critique of the PEM implies that an adjustment to the archetypal
distinction between scientific experts and legislative policy-makers is also required.
E&K themselves call for “cooperative knowledge production and a role for mutual
learning between experts and decision-makers in environmental policy” (2015, 57).
But our discussion has shown that science advisors are decision-makers, too, and
as such are at least somewhat responsible to the public. Likewise, policy-makers
are (or ought to be) conceived of as experts in their own field—in the field of
responding to public concern and of putting in place regulations, laws, institutions,
and so forth to resolve those concerns. While scientists and policy-makers are used
to thinking differently about their responsibilities and working at different time
scales, they need to learn to work according to a model of interdisciplinary
collaboration, rather than in conceptually strict and opposing roles.
There is much to like about E&K’s approach: their acknowledgement of the
value-ladenness of science; their emphasis on post-implementation monitoring of
policy, driving a feedback look on the assessment process; their guarding against
misuse or misguided use of science as advocacy; and their attempt to forge a close
relationship between science advisors and policy-makers. But at the center of the
PEM as E&K are currently presenting it, there is an unstable and incoherent
compromise between the crucial insights inspiring the model, and the demand that
the work still draw a line between scientific assessment and policy processes—in
order, we presume, that climate science advising retain a superficial appearance of
value-neutrality.
At this point we can only reiterate something that Douglas herself, echoing
Hempel, said at the conclusion of her re-introduction of the argument from
inductive risk to philosophers of science: “the argument ‘I want X to be true,
therefore X is true’ remains a bad argument, both within and without science”
(2000, 578). This is a point that pertains not just to those wary of biases in science,
but also to those who think science ought to remain purportedly value-free, or that
scientists can and should simply defer their decisions in response to inductive risk.

References
Anderson, Elizabeth. 2014. "Dewey's Moral Philosophy." The Stanford
Encyclopedia of Philosophy, edited by Edward N. Zalta.

18
Betz, Gregor. 2013. “In Defence of the Value Free Ideal.” European Journal for
Philosophy of Science 3(2): 207–20.
Biddle, Justin B., and Eric Winsberg. 2010. “Value Judgements and the Estimation
of Uncertainty in Climate Modeling.” In New Waves in Philosophy of
Science, edited by P. D. Magnus and J. Busch, pages 172-197. Basingstoke,
Hampshire: Palgrave MacMillan.
Brown, Mark B. 2009. Science in Democracy: Expertise, Institutions, and
Representation. Cambridge, MA: MIT Press.
Brown, Matthew J. 2012. “John Dewey’s Logic of Science.” HOPOS 2(2): 258-306.
Brown, Matthew J. 2013. “The Democratic Control of the Scientific Control of
Politics.” In EPSA11 Perspectives and Foundational Problems in Philosophy
of Science, edited by V. Karakostas and D. Dieks, pages 479–491.
Dordrecht: Springer.
Brown, Matthew J., and Joyce C. Havstad. 2017. “The Disconnect Problem,
Scientific Authority, and Climate Policy.” Perspectives on Science 25(1):
67–94.
Dewey, John. 1939. Theory of Valuation. Chicago: The University of Chicago Press.
Dorstewitz, Philipp. 2011. “Dewey’s Science: A Transactive Model of Research
Processes.” In The Continuing Relevance of John Dewey: Reflections on
Aesthetics, Morality, Science, and Society, edited by L. Hickman, M.
Flamm, K. Skowronski, and J. Rea, pages 205-224. New York: Rodopi.
Douglas, Heather E. 2000. “Inductive Risk and Values in Science.” Philosophy of
Science 67(4): 559–579.
Douglas, Heather E. 2003. “The Moral Responsibilities of Scientists: Tensions
between Autonomy and Responsibility.” American Philosophical Quarterly
40(1): 59–68.
Douglas, Heather E. 2004. “The Irreducible Complexity of Objectivity.” Synthese
138(3): 453–473.
Douglas, Heather E. 2005. “Inserting the Public into Science.” In Exploring Novel
Forms of Scientific Advice in Political Decision-Making, edited by S. Maasen
and P. Weingart, pages 153–169. Dordrecht: Springer.
Douglas, Heather E. 2009. Science, Policy and the Value-Free Ideal. Pittsburgh:
University of Pittsburgh Press.
Edenhofer, Ottmar, and Martin Kowarsch. 2015. “Cartography of Pathways: A New
Model for Environmental Policy Assessments.” Environmental Science &
Policy 51: 56–64.
Elliott, Kevin C. 2011. Is a Little Pollution Good for You? Incorporating Societal
Values in Environmental Research. New York: Oxford University Press.
Fisher, Erik. 2007. “Ethnographic Invention: Probing the Capacity of Laboratory
Decisions.” NanoEthics 1(2): 155–165.

19
Fisher, Erik, Roop L. Mahajan, and Carl Mitcham. 2006. “Midstream Modulation of
Technology: Governance from Within.” Bulletin of Science, Technology &
Society 26(6): 485–496.
Fisher, Erik, and Daan Schuurbiers. 2013. “Socio-Technical Integration Research:
Collaborative Inquiry at the Midstream of Research and Development.” In
Early Engagement and New Technologies: Opening up the Laboratory,
edited by N. Doorn, D. Schuurbiers, I. van de Poel, and M. E. Gorman,
pages 97–110. New York: Springer.
Havstad, Joyce C., and Matthew J. Brown (forthcoming). “Neutrality, Relevance,
Prescription, and the IPCC.” Public Affairs Quarterly.
Hickman, Larry A. 2007. “Beyond the Epistemology Industry: Dewey’s Theory of
Inquiry.” In Pragmatism as Post-Postmodernism: Lessons from John Dewey,
chapter 12, pages 206–230. Bronx, N.Y.: Fordham University Press.
IPCC. 2007. Climate Change 2007: Synthesis Report. Contribution of Working
Groups I, II and III to the Fourth Assessment Report of the
Intergovernmental Panel on Climate Change, edited by Core Writing Team,
R.K. Pachauri, and A. Reisinger. Geneva: IPCC.
IPCC. 2014. Climate Change 2014: Mitigation of Climate Change. Contribution of
Working Group III to the Fifth Assessment Report of the Intergovernmental
Panel on Climate Change, edited by O. Edenhofer, R. Pichs-Madruga, Y.
Sokona, E. Farahani, S. Kadner, K. Seyboth, A. Adler, I. Baum, S. Brunner,
P. Eickemeier, B. Kriemann, J. Savolainen, S. Schlömer, C. von Stechow, T.
Zwickel and J.C. Minx. Cambridge: Cambridge University Press.
Jeffrey, Richard C. 1956. “Valuation and Acceptance of Scientific Hypotheses.”
Philosophy of Science 23(3): 237–246.
Mitchell, Sandra D. 2004. “The Prescribed and Proscribed Values in Science
Policy.” In Science, Values, and Objectivity, edited by P. Machamer and G.
Wolters, pages 245-255. Pittsburgh: University of Pittsburgh Press.
Pielke, Roger A., Jr. 2007. The Honest Broker: Making Sense of Science in Policy
and Politics. New York: Cambridge University Press.
Putnam, Hilary. 2002. The Collapse of the Fact/Value Dichotomy and Other
Essays. Cambridge, MA: Harvard University Press.
Rudner, Richard. 1953. “The Scientist Qua Scientist Makes Value Judgments.”
Philosophy of Science 20(1): 1–6.
Steel, Daniel (forthcoming). “Climate Change and Second-Order Uncertainty:
Defending a Generalized, Normative, and Structural Argument from
Inductive Risk.” Perspectives on Science: 696–721.
Stern, Paul C. and Harvey V. Fineberg, eds. 1996. Understanding Risk: Informing
Decisions in a Democratic Society. Washington, DC: National Academies
Press.

20
Waks, Leonard J. 1999. “The Means-Ends Continuum and the Reconciliation of
Science and Art in the Later Works of John Dewey.” Transactions of the
Charles S. Peirce Society 35(3): 595–611.
Winsberg, Eric. 2012. “Values and Uncertainties in the Predictions of Global
Climate Models.” Kennedy Institute of Ethics Journal 22(2): 111–137.

21