Review of Michael Bishop and J.D. Trout, Epistemology and The
Psychology of Human Judgment
Jonathan M. Weinberg
The conjunctive title of Michael Bishop and J. D. Trout's Epistemology and The Psychology of Human
Judgment accurately conveys that the book will find at least two audiences:
epistemologists on the one hand, and psychologists and philosophers of
psychology on the other. (Of course
these are not mutually exclusive audiences!)
With regard to the latter group of readers, Bishop and Trout (B&T)
have much useful material to offer. Much
of their discussion of the follies and foibles of human judgment, and
especially its lousiness in comparison to scientifically-derived statistical
prediction rules, will prove a valuable introduction to that material for
philosophers not already familiar with it.
Moreover, they have significant insights to share even with those
already well-acquainted with such results, especially those who have been
following the 'rationality wars' (see in particular Chapter 8, "Putting
Epistemology into Practice: Normative Disputes in Psychology"). B&T's frequent, easy, and exceptionally
clear presentation of actual examples from the empirical literature is one of
their book's chief pleasures, and in this regard is a model for
empirically-engaged philosophizing.
But it is to the epistemologists that B&T offer the
greatest promise -- or, perhaps, the greatest threat. For they aim to deploy those psychological
results to do nothing less than launch a twofold revolution
in normative epistemology. First and
negatively, B&T offer mean to overturn "standard analytic
epistemology" (SAE), with such unapologetically fierce accusations as
SAE’s having "jumped the shark", with "its goals and methods
beyond repair" (22). Second, and
more positively, they aim to replace (what they argue is) the fruitless,
normatively inert SAE with a more empirically-informed, satisfyingly normative
epistemology. Unfortunately, the first,
negative project fails badly, because B&T do not take the epistemological
tradition sufficiently seriously. And
consequently, they do not appreciate several ways in which their own proposed
account cannot perform necessary epistemological work that SAE at least tries
to do, and thus the second meta-epistemological promise stands unfulfilled as
well.
Their argument against the normative status of SAE goes by
in a breezy 14 pages. (Full disclosure:
one of my papers is discussed in that section.)
SAE starts from a "descriptive core" of "the considered
epistemic judgments of philosophers" (109). And the problem, B&T contend, is that SAE
has no way to move out from that descriptive core and on to anything actually
normative. And normative guidance as to
belief formation and revision is, plausibly, a major payoff we were hoping to
get from our investment in epistemology.
They object, in a move that seeks to turn the table on traditional
arguments against naturalized epistemology, that SAE starts from the
"armchair anthropology" (184) of merely documenting that philosophers
have the particular intuitive judgments they do about particular cases – and
can then offer no reasons to infer from that bit of descriptive anthropology to
any claims as to how we should go
about forming our beliefs.
They do consider the possibility that philosophers might
claim the mantle of experts in
matters epistemic, and thereby lend their intuitions the normative force that
should go along with such expertise.
"But," they argue, "if proponents of SAE are experts
about epistemological matters, then it is reasonable to suppose that they have
some kind of documented success" (114).
Yet epistemology, they claim, has radically failed to converge in the
way that one would expect were it a fruitful cognitive enterprise.
But there are other ways that philosophers might count as
experts on cognitive success, even in the absence of runaway progress in
epistemology. First, those who are the
most achieved practitioners in a
domain need not be those who are experts on
it. We know better, for example,
than naively to ask scientists how science is done. Thus epistemologists may be experts on cognitive success more broadly, even when they
have produced less of it than one might like.
The particular questions they choose to focus on may just be harder
questions, for example, and so their expertise has not had a chance to manifest
itself in a set of answers to easier questions.
Moreover, B&T miss the larger set of gains that epistemology has
made in the last century, in terms of agreed rejection of theories and constructs. Just to list a few major points of
near-consensus, philosophers have come almost universally to reject the
conflation of the analytic, the a priori,
and the necessary; or the positivistic notion of the falsification of
individual claims independent of their theoretical background. Finally, and perhaps most importantly,
B&T miss the fact that many epistemologists take themselves to be engaged
in an ‘always already’ normative enterprise.
We are not merely describing our judgments, but expressing them and
thereby simultaneously bringing to bear and bringing to light the normative
commitments we already endorse.
A further problem arises when they consider SAE's commitment
to a "stasis requirement", which they interpret as a commitment in
Jaegwon Kim's famous (1988) essay on naturalized epistemology. They state it thus: "if an epistemic
theory forced us to radically alter our considered epistemic judgments ... then
ipso facto that theory is unacceptable" (9). But Kim’s idea, I take it, is not meant to be
terribly radical – if a theory of knowledge had the result that we know that 2
+ 2 = 5, but do not know that we have hands, then that would just be a
nonstarter as a theory of knowledge.
Whatever knowledge is, we cannot be that
wrong about its instances. B&T,
however, interpret Kim as meaning that no changes in our normative judgments
are possible, period. But it is not
obvious that SAE practitioners have such a commitment to so thoroughgoing a conservatism. Rather, it seems more likely that
epistemologists could approve of all sorts of ways in which judgments may be
improved over time, as we learn more and more about what forms of reasoning are
more reliable. Such changes would be
entailed, for the epistemologist, by other of our normative commitments – in
particular, commitments to principles along such lines as “generally shape your
beliefs to accord with the evidence” or “do not put evidential weight on
sources of evidence that have been shown to be unreliable”. Kim is worried about a massive overturning of
our centrally-held judgments, such as would be entailed by a skeptical
epistemology, but he (and SAE types more generally) are not at all worried
about local, piecemeal adjusting in accord with such discoveries.
B&T further object that, at best, SAE might avoid an
unsustainable conservatism only by yielding to an equally unbearable
vacuity. If SAE produces no cognitive
improvements of its own, but can merely render itself consistent with them,
then SAE can seem about as useful as a broker who offers advice no more
specific than “Buy low and sell high” (144).
There is, I suspect, something to this complaint, but the SAE practitioner
may well respond that epistemology stands to practical reasoning advice as
theoretical physics stands to engineering – its impact on our lives is highly
mediated and indirect, but substantive nonetheless. But, because they seem committed to the idea
that epistemology must have direct practical consequences, B&T do not even
consider such a rejoinder.
* * * * *
For all that, even if SAE is not as dead in the water as
B&T claim, their own preferred replacement
methodology could still win out as comparatively better. This methodology, which they call “Strategic
Reliabilism”, involves performing cost-benefit analyses for different reasoning
strategies: how accurate is each strategy, across how robustly divergent a
range of conditions, incurring what costs in terms of time and effort, and
weighted in terms of the significance of the discovered truths? We can use such analyses to guide our
selection of different reasoning strategies in different circumstances. It is a methodology with obvious normative
force – who could reasonably prefer to pick the
less-accurate/more-expensive/less-significant strategy? – and they illustrate
it with numerous examples of reasoning strategies from the diagnosis of
psychotic patients to Gigerenzer et al.’s
work on the recognition heuristic to research on the risks of talking on a cell
phone while driving. B&T are
admirably clear about several of the theoretical burdens they must shoulder, to
make their case. For example, they acknowledge
that they owe us an account of how we are to determine epistemic significance,
and devote a chapter to attempting to develop an account in terms of object
reasons to apportion one’s cognitive resources in a given way.
Nonetheless, at the end of the day Strategic Reliabilism
cannot be considered a real rival to SAE.
The fundamental reason that this is so, is that
Strategic Reliabilism cannot give an account of the full breadth and variety of
our epistemic practices. There are only
a very select and limited set of cognitive tasks for which we can even begin to
perform the kind of cost-benefit calculations that B&T invite us to engage
in. The most obvious example of this
shortcoming is scientific reasoning itself.
Consider such extant debates as the proper role (if any) of
adaptationist hypotheses in evolutionary biology and psychology; or the
appropriateness (if any) of adopting a fully extended/embedded/embodied
explanatory framework in cognitive science.
It is decidedly unclear how, even in principle, Strategic Reliabilism
could help us conduct such debates.
B&T's framework can only apply when we have at least a faint idea of
how to individuate strategies, and how to measure their products in terms of
reliability and significance. But even
if we grant that, say, the 'extended mind' theorist and her rivals really have
different reasoning strategies (instead of locating their disagreement in
other, perhaps metaphysical terms), we have no way of gauging their comparative
robust reliabilities. What works well
for measuring neatly delineated statistical prediction rules in reasonably
operationalizable domains will fail us just where we need the most philosophical assistance.
Indeed, philosophy itself would remain epistemologically
inscrutable, in Strategic Reliabilism's terms.
B&T's own argumentation is, after all, not itself conducted via the
kind if cost-benefit analysis it proposes, but in terms of straightforward,
old-fashioned argumentation (albeit not so old-fashioned as to exclude frequent
appeal to scientific results). That the
form of appraisal the book supports cannot be extended to include itself is,
surely, a sign that the proposed theory is regrettably narrow in its
applicability. (At least, if it can be so extended – requiring, just for
starters, a way of cleanly distinguishing between different types of philosophical
argument, such that the Strategic Reliabilist apparatus can be applied to them
– B&T present us with no picture of how this might be so.)
That the authors do not really attend to the question of
whether Strategic Reliabilism may not apply to all or even most areas where we
might seek epistemological guidance, is perhaps the most disappointing
blindspot of the book. For example, at
one point they argue that even a successfully completed SAE of justification,
were we to have such a thing, would be of little use if we also had the results
of Strategic Reliabilism. Either the two theories would converge, or they wouldn't; if
they converged, then there'd be no work for the SAE theory to do; if they
diverged, then for any belief that SAE ruled as unjustified yet Strategic
Reliabilism recommended, it would be more rational to follow Strategic
Reliabilism (116). Put aside such
worries as why it would be SAE, and not Strategic Reliabilism, that is obviated
in the first case -- what I find most striking is that they don't even consider
the possibility of some beliefs' being considered justified by SAE but either ruled against or not evaluable at
all by Strategic Reliabilism. (If my
discussion above is correct, then some philosophical theses may be just such
cases.) It is just not a possibility
they address, and their intended dilemma is thus shy a horn.
Moreover, B&T do not consider that we might have good
reasons to adopt a form of epistemic scorekeeping that abstracts away from such
matters as ease of deployment. Perhaps
it is part of science's success that it can create social epistemic spaces that
are largely sheltered from such concerns.
(B&T present an oddly unsocial picture of epistemic agents, with
only two brief and incidental references to social epistemology in all of EPHJ.)
So, B&T offer no novel serious challenge to SAE; and
even if they did, their rival epistemology really cannot begin to replace
SAE. But I believe their fundamental
mistake is to hunt such big game in such a short book. Rather, what they have done successfully is laid the groundwork for a naturalistic
epistemology of judgment & expertise.
B&T describe a number of studies that make pressing the question:
when ought we trust experts over grounded reasoning
strategies? And they make a serious case for the answer: hardly ever! Yes, in
principle one might be able to track factors relevant to a given judgment
that some model might not have incorporated; this "selective defection
strategy" (47) would have us generally comply with some statistical
prediction rule, except (and it's a big exception!) in such circumstances where
we take ourselves to have reason to think we can improve on it. But in actual practice, human agents are much
more likely to misidentify such situations and/or misdeploy the information
about such factors, than we are to improve over the model by using the
information appropriately. The resulting
puzzle for epistemologists, then, is how to incorporate such a result into our
picture of human epistemic agency. There
would be plenty here for epistemologists to chew on, if B&T had only sought
to present these results to epistemologists -- instead of presenting them at them.
Epistemologists need to think about what this all means for the
Principle of Total Evidence, for example, or for our theory of epistemic
self-trust. The overconfidence
literature, which B&T discuss at some length (144ff), also could present
some interesting challenges for contemporary epistemologists. But making such challenges
in a way that epistemologists would have to attend to, would have required a
book that itself attended more closely to epistemology, instead of aiming
mostly to contend with it.
Austin, J. L., (1946/1979). "Other
Minds" in Austin’s Philosophical
Papers, ed. J.O. Urmson and G.J.
Warnock. Clarendon Press.
Bishop, Michael and Trout, J.D. (2005). Epistemology and the Psychology of Human Judgment.
Jackson, Frank (1998).
From Metaphysics to Ethics: A
Defence of Conceptual Analysis.
Kim, Jaegwon. (1988). "What is Naturalized
Epistemology?" in Philosophical
Perspectives 2, edited by James
E. Tomberlin, Asascadero, CA: Ridgeview Publishing Co:
381-406.
Leite, Adam. (2004). "Is Fallibility an Epistemological
Shortcoming?" Phil Quart, 54