Professional Documents
Culture Documents
2007 Global Environmental Change
2007 Global Environmental Change
net/publication/222397518
CITATIONS READS
422 861
7 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by B. Orlove on 29 April 2018.
Abstract
People process uncertainty information in two qualitatively different systems. Most climate forecast communications assume people
process information analytically. Yet people also rely heavily on an experiential processing system. Better understanding of experiential
processing may lead to more comprehensible risk communication products. Retranslation of statistical information into concrete
(vicarious) experience facilitates intuitive understanding of probabilistic information and motivates contingency planning. Sharing
vicarious experience in group discussions or simulations of forecasts, decisions, and outcomes provides a richer and more representative
sample of relevant experience. The emotional impact of the concretization of abstract risks motivates action in ways not provided by an
analytic understanding.
r 2006 Elsevier Ltd. All rights reserved.
Keywords: Risk communication; Environmental decisions; Group decision making; Mental processing; Affective decision making; Climate information;
Uncertainty; Probability
1. Introduction next few decades. Let us use the example of a maize crop
farmer, who faces the decision of how much land to
To communicate is to transmit an idea so that it is allocate to the crop, selection of sowing date, genotype,
satisfactorily understood and, typically, used to guide crop density, and rate and time of fertilizer application.
action. The successful communication of climate informa- Advance information of the climatic conditions might give
tion, be it a seasonal forecast or the prediction of a long- farmers an indication of the likelihood of low or high soil
term warming trend, is complicated by the uncertainties moisture and soil temperature (frost) at sowing, flowering,
attached to the forecasts. A forecast might assign a and harvesting. Climate information can influence land
probability of 45% to higher-than-normal rainfall in the assignation, production schedules, and commercialization
next growing season and a probability of 15% to lower- strategies—decisions that are made long before the
than-normal rainfall, or it might state that an increase of at sowing season starts (Bert et al., 2006). When the recipient
least 1 1C in global temperature is highly probable over the of such climate information acts as though the unusual
season or warming trend is certain to occur, or, at the
Corresponding author. Tel.: +1 212 854 8760; fax: +1 212 854 3609. opposite extreme, entirely rejects the forecast because it is
E-mail address: [email protected] (S.M. Marx). uncertain, the probabilities have either not been adequately
0959-3780/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.gloenvcha.2006.10.004
ARTICLE IN PRESS
48 S.M. Marx et al. / Global Environmental Change 17 (2007) 47–58
communicated or have been ignored for other reasons, for climate information by examples drawn from one’s own
instance, economic and financial situation, lack of access to personal experience and that of others may be accom-
labor and machinery, production objectives, managerial plished.
skills, or psychological reasons such as risk aversion.
Communication of climate uncertainty may be improved
by better understanding how people learn and reason 1.2. Experiential and analytic processing: a brief overview
about uncertainty and how climate-related decisions are
influenced by uncertainty. The last 15 years have seen Experiential processing relates current situations to
considerable progress in understanding how uncertainty memories of one’s own or others’ experience. Some high-
and probability or likelihood information is understood by level abstract concepts may be involved in experi-
individuals with the help of two very different processing ential processing, but others are neglected, particularly
systems, one experiential, the other analytic. More recently, those that involve thinking about an ensemble of
insights from this research are being applied to the special different experiences together. While the concept ‘‘aver-
case of climate uncertainty, examining different methods of age’’ can be incorporated relatively easily in experiential
communicating uncertain forecasts.1 While behavioral processing by bringing to mind a typical situation
decision research has concentrated mostly on individuals, (more likely a mode than a true mean), concepts such as
many climate-related decisions are made by groups or by relative frequency and sample size tend to be neglected
individuals after group discussion. Thus it seems appro- because they cannot easily be imagined correctly. Analytic
priate to extend research on judgment and decision-making processing, by contrast, includes mechanisms that relate
under uncertainty beyond the individual to include the the current situation to processed ensembles of past
effects of group processes and group-related goals. relevant experience and thus can easily and naturally
We acknowledge that there are many interconnected express statistical constructs such as probability and
variables that affect individual and group decisions under sample size (Stanowich and West, 1998). Experiential
climate uncertainty, such as the framing of information, processes are akin to the ‘‘concrete operations’’
incentives, material and non-material goals. Here we focus (understanding of cause and effect) described by
on two ways in which people process information, while analytic processes are an example of ‘‘formal
experiential versus analytic processing, a distinction central operations,’’ (abstract thinking, i.e., operations on ensem-
to understanding the problem of communicating uncertain bles of concrete experiences).
climate information. Past experiences often evoke strong feelings, making
them memorable and therefore often dominant in proces-
sing (Slovic et al., 2002; Loewenstein et al., 2001). Strong
1.1. Outline of the paper
feelings such as pleasure and pain, fear, anger, horror, joy,
and awe involve activation of paleocortical brain structures
The paper is organized as follows. After defining
that are evolutionarily older than neocortical structures
experiential and analytic processing as part of the Section
and found in all vertebrates. By contrast, analytic
1, Section 2 reviews, in more detail, evidence showing that
processing involves the neocortex, a structure found only
personal experience or vivid descriptions often dominate
in mammals and in expanded form only in primates and
statistical information, even though the latter typically
especially humans. The extent to which analytic processes
provides more—and more reliable—information, and
occur in lower animals is a subject of active investigation;
explains this as the result of competition between the
but it seems clear that some processes, including those that
experiential and analytic processing systems. Section 3
underlie the syntactic structures of human language and
discusses several experientially based heuristics and some
the use of extended chains of logic, are uniquely human.
biases that may result. Section 4 concludes with implica-
Table 1 summarizes the major attributes of the two
tions for research and policy. Throughout this paper, we
processes (Piaget and Inhelder, 1962; Inhelder and Piaget,
will refer to a wide range of examples of decisions, in
1958).
particular two climate-related research projects currently
There is not a sharp separation between experiential and
underway at the Center for Research on Environmental
analytic processing. Even simple reflexes can be influenced
Decision (CRED): one project looks at the impact of
by neocortical processes and analytic reasoning can lead
(vicarious) concrete, experiential information, as exempli-
to strong feelings of pleasure, fear, anger, joy and awe.
fied in the movie The Day After Tomorrow, on climate
A decision process always integrates both kinds of
change risk perception and policy; the other project
processing. The role of analytic processes in the under-
examines the comprehension of statistical information
standing of (climate) uncertainty and in decisions
through experiential retranslation by Ugandan farmers’
involving such information, however, has often been
groups, illustrating how the concretization of statistical
overestimated and the role of experiential processes
1
For example, see the research conducted at the Center for Research on has been ignored. A better appreciation of experiential
Environmental Decisions (CRED) at Columbia University, www. processing may point us towards improved communication
cred.columbia.edu. strategies.
ARTICLE IN PRESS
S.M. Marx et al. / Global Environmental Change 17 (2007) 47–58 49
‘‘retranslation’’ of concrete experience by analysis may be disease’’ elicit greater fear than reports about incidences
accompanied by affect: strong emotions may be evoked by of bovine spongiform encephalitis or Creutzfeld-Jacob
analytic processes. A related sort of retranslation occurs disease, a more abstract, scientific label for the same
when probabilities are given a causal interpretation disorder (Sinaceur and Heath, 2004).
(Tversky and Kahneman, 1974), which tends to make the Retranslation by modifying concrete categories may also
relevance of statistical information salient, by providing be promoted through analogies between novel probabilistic
people with a story of why statistical base rate information assertions on the one side, and situations that arise in
needs to be considered. Following up on our previous everyday life where people already take chance into
example of the cab companies in a city: relative frequency, account in their planning on the other side. Such analogies
for instance ‘‘15% of the taxis in a city are operated by the play an important role both in classroom teaching of
Blue Cab company,’’ is mostly ignored when people are statistical thinking and in communication of probabilistic
asked to judge the probability that a ‘Blue Cab’ caused an information relevant to decisions. The following two
accident. However the same relative frequency information examples illustrate how analogies can be used to convey
is integrated with the case-specific evidence when it is stated probabilistic climate forecasts. Phillips and Orlove (2004)
in a way that lends itself to causal retranslation, namely conducted an exercise that helped Ugandan farmers
‘‘15% of the taxi accidents in a city involve the Blue Cab evaluate forecasts for seasonal climate conditions provided
company.’’ The causal interpretation combines readily with by the Ugandan Department of Meteorology. The exercise
other concrete thinking. Further evidence of the impor- compared the meteorological forecasts with traditional
tance of causal schemas in judgment is given by Tversky predictions and forecasts based on a cultural system of
and Kahneman (1982). In the context of climate variability, ‘‘signs’’ that are empirically observable (Orlove and
the role of causal schemas can be seen when a seasonal Kabugo, 2005). Farmers were presented with a common
forecast that assigns a 45% probability for La Niña prediction scenario: they were asked to consider a forecast
conditions is translated into ‘‘there is a 45% chance that of the sex of an unborn child based on a common local
the climatic conditions of the next season will contribute to sign. They were then asked to indicate the chances that the
lower than normal rainfall in the Argentine Pampas and child would turn out to be of the sex indicated by the sign,
cause crop failures.’’ by distributing ten dots on a piece of paper under two
To use another example of how the two processing pictures, one of a boy, the other of a girl, according to the
systems operate together, consider the decision process of likelihood of the outcome (see Fig. 1). A large majority of
someone who lost an early-planted maize crop three years discussants (77% pooling data from four focus groups
ago, when the rains, having begun, halted for three weeks ðn ¼ 252Þ) allocated dots to both pictures, showing that
just after planting. On contemplating early planting this they understood that the prediction, though deterministic
year, the person may react with anxiety and be inclined to (a certain shape of a woman’s belly indicates a son, for
wait. On the one hand, analysis of climate records or example), was uncertain, and its forecasted outcome might
forecasts might well reinforce this experientially based
decision—the chance of such a dry spell might be 10% or
20%, and this may seem too high to risk the investment of
seed and labor. However, if statistical analysis suggests that
what happened three years ago was really quite rare, early
planting might make sense analytically, yet the negative
affect stemming from the previous year’s experience may
nonetheless prevail. One sees this response in extreme form
in phobic reactions, where people know analytically that
their avoidance behavior is unnecessary or even harmful,
but they cannot bring themselves to suspend it.
Even in seemingly ‘‘objective’’ contexts such as financial
investment decisions, subjective and largely affective
factors have been shown to influence perceptions of risk.
For business executives studying in a University of Chicago
program, emotional reactions (e.g., worry or dread) to
investment opportunities of health and safety related
activities were just as important as statistical variables
(e.g., outcomes and their probabilities) in predicting
perceptions of risk in both the financial and the health/
safety domain (Holtgrave and Weber, 1993). Likewise, if
risk perceptions were driven exclusively by statistics, they
would not be influenced by the way a particular hazard is Fig. 1. Explaining the prediction ‘‘Game’’, Kumi, Uganda, July 2004
labeled. Yet, reports about incidences of ‘‘mad cow (Photo credit: Jennifer Phillips/Ben Orlove).
ARTICLE IN PRESS
S.M. Marx et al. / Global Environmental Change 17 (2007) 47–58 51
not take place. In the case of rainfall forecasts (based on experience is processed by the experiential system and the
e.g., observations of shifts in wind direction or arrival of generated affect is an effective motivator of action. More
migratory birds) the percentage of those who indicated the pallid statistical information is processed by the analytic
uncertain nature of the forecast was even higher, with 82% system, whose output tends to have less weight in actions
pointing to the possibility of more than one outcome. or decisions, unless decision makers have been trained to
Presented with a government-issued seasonal forecast pay conscious attention to statistical information and its
that mentioned a most probable outcome, 88% of the implications. Here, we focus specifically on the way in
participants allocated dots to more than one outcome. which information about the likelihood of events is
In their research with farmers in Zimbabwe, Suarez and acquired, comparing the experiential route (involving
Patt (2004) introduced conditional and joint probabilities affective and other associative processes) to the analytic
by linking seasonal forecast uncertainty (whether El Niño route (involving statistical and other abstract information
would weaken or gain strength over the next few months) processes).
with the more familiar example of betting on a soccer In daily life, decision makers often learn about outcomes
game. Analogous to the benefits of improved El Niño and their probabilities as a function of their profession or
information and higher-skilled forecasts becoming avail- role. Doctors, for example, learn about health outcomes of
able if one waited a month or two (until October or treatment decisions in a different way than the lay public.
November), Suarez and Patt suggested that the national Consider the decision whether to vaccinate a child against
soccer teams from Zimbabwe and Argentina would be diphtheria, tetanus, and pertussis (DTaP). Parents who
facing each other in a game, yet one of the star players was research the side effects of the DTaP vaccine on the
supposedly injured and no one knew for sure whether it National Immunization Program web site will find that up
was Argentina’s or Zimbabwe’s star. Clearly, the likely to 1 child out of 1000 will suffer from high fever and about
winner would depend on which player was injured. A poll 1 child out of 14,000 will suffer from seizures as a result of
of the workshop showed that most participants thought immunization. Although doctors have these same statistics
Argentina would win in either case, however, if the at their disposal, they also have access to information not
Argentine star player Diego Maradona was injured many easily available to parents—namely, the personal experi-
people thought Zimbabwe had a better than even chance of ence, gathered across many patients, that vaccination
winning. Based on the following conditional probabilities: rarely results in side effects; few doctors have encountered
80/20 Argentina if Maradona plays, and 60/40 Zimbabwe one of the unusual cases in which high fever or seizures
if the Zimbabwean star Peter Ndlovu plays, workshop follow vaccination. If the importance assigned to rare
participants calculated the overall probabilities for either events differs as a function of how one learns about their
Argentina or Zimbabwe winning as 60/40 Argentina. The likelihood, then doctors and patients might well disagree
climate forecast in Zimbabwe was not dissimilar: fore- about whether vaccination is advised (Ball et al., 1998).
casters simply did not know yet what role El Niño would Several papers have examined how people combine
play. Assuming a probability of 50/50 for intensifying/ statistical and experiential information about likelihoods
weakening El Niño conditions, workshop participants were (Birnbaum and Mellers, 1983; Tversky and Koehler, 1994;
asked to develop a forecast of the impact of such Jenni and Loewenstein, 1997; Clemen and Winkler, 1999;
conditions on Zimbabwe’s climate. Most farmers under- Kraemer and Weber, 2004; Fagerlin et al., 2005). The
stood very well the analogy of the lack of knowledge about bottom line is that experiential information overwhelms
whose star player would be injured, on the one side, and statistical information, unless statistical information is re-
the lack of information about El Niño’s increase of expressed (visually, narratively, or otherwise) in ways that
decrease at the time of the seasonal forecast available in can be combined with personal experience.
October on the other side. Farmers saw the value of Related to the distinction between analytic and experi-
waiting for additional information and were able to ential processing is a recent distinction between decisions
provide the calculations for the November forecast. made from description vs. decisions made from experience
In summary, research on people’s processing of statis- (see Table 2). An example of a description-based decision is
tical information suggests that analytic information is best a choice between two lottery tickets, where each ticket is
understood when it is used to recategorize or recontextua- described by a probability distribution of possible out-
lize the decision maker’s current situation. In turn, this is comes (i.e., statistical summary information). In contrast,
best accomplished when the analytic results can be when people decide whether to back up their computer’s
translated into concrete images, strong emotions, or hard drive, cross a busy street, or invest in a new water
stories. system to irrigate their crops, they often do not know the
complete range of possible outcomes, let alone their
2.2. Experiential vs. analytic processing of probabilities probabilities. Instead people typically decide based on past
personal experience. Research has shown that the weight
The two processing systems described above affect the given to small-probability events differs dramatically in the
processing of all sorts of information to influence all sorts two processing systems (with much greater weight given to
of behaviors. Concrete, personal or vicariously related small-probability events when small probabilities are
ARTICLE IN PRESS
52 S.M. Marx et al. / Global Environmental Change 17 (2007) 47–58
2.3. Effect of affective/experiential vs. statistical evidence in 3. Consequences of the experiential processing of (climate)
groups risk information
Research on risk amplification (Pidgeon et al., 2003) and Most decisions require the evaluation of events that are
emotional contagion (Barsade, 2002) suggests that known to occur only probabilistically. These are referred to
groups—just as individuals—are more motivated by as decisions under risk, when the likelihood of different
concrete (affective and experiential) information than events is known precisely, e.g., the probability of getting a
statistical information when they decide which risks to ‘‘head’’ when tossing a fair coin, or as decisions under
pay attention to and when to take protective action. uncertainty, when the likelihoods themselves are uncertain,
We may, however, expect to see differences in the way e.g., the probability of precipitation tomorrow. Decisions
that groups (compared to the average individual) utilize from description, described in the previous section,
statistical information. First, groups sometimes include typically specify the probability of events as precisely as
ARTICLE IN PRESS
S.M. Marx et al. / Global Environmental Change 17 (2007) 47–58 53
possible; decisions from experience are typically made statistical risk of the situation reduce our feeling of risk, use
under some uncertainty. More important than this rather of the affect heuristic can result in biases, i.e., in
technical distinction between risk and uncertainty, how- inappropriate willingness to take on such risks. Klos
ever, is the differentiation, made in the previous section, et al. (2005) show, for example, that people are generally
between an analytic vs. experiential basis for the detection willing to take greater risks in situations where they have
of a risk, e.g., the risk of changing or varying climatic greater familiarity with the risky options (e.g., the home
conditions, and the decision to take defensive or mitigating bias observed in investors in financial markets), which can
action. be explained by the fact that familiarity with a risk reduces
When risk information is not processed analytically, the feeling of being at risk. Use of the affect heuristic also
processing often takes the form of simplifying heuristics. A predicts that misattributions of the sources of feeling at risk
heuristic is a rule of thumb, a replicable approach for will affect decisions and actions. Thus, Hirshleifer and
directing one’s attention in problem solving. In this section, Shumway (2003) found a positive relationship between
we describe three heuristics that all utilize concrete images morning sunshine at a country’s leading stock exchange
and emotional reactions. Heuristics influence judgments, and market index stock returns that day at 26 stock
choices, and decisions. While heuristics are generally exchanges internationally from 1928 to 1997, a finding that
beneficial, i.e. provide pretty good results most of the time is difficult to reconcile with fully rational price-setting, but
while minimizing processing effort (Payne et al., 1992), perhaps indicative of emotional mediation.
they typically also have some downside. In the following Let us turn to a more detailed example of how the affect
pages, we describe three heuristics of the experiential heuristic applies to climate forecast use among Ugandan
processing system, drawing on examples of climate farmers. Farmers’ concern with seasonal variability is
information use ranging from Ugandan farmer commu- particularly strong because of its effects on agricultural
nities to the general public in the United States and livelihoods. In rain-fed agriculture, a rainy season that is
Europe. At the end of this section we also point to some either scanty or excessive can lead to economic hardship.
limitations that can make the use of heuristics problematic. Past research has shown a tendency for climate profes-
sionals and the lay public alike in the region to translate
3.1. The affect heuristic or ‘‘risk-as-feelings’’ probabilistic forecasts into deterministic information,
leading to potential increases in risks associated with
Many, if not most decisions and actions under risk and responses; this translation is often due to the affect
uncertainty are driven by how we feel about the situation heuristic, granted the great concern over crop failure in
(Slovic et al., 2002, 2004; Loewenstein et al., 2001). Slovic the region.
et al. (2004) define affect as a ‘‘‘faint whisper of emotion’, In an ongoing project, Orlove and Roncoli (2006) are
referring to the specific quality of goodness or badness studying group processing of climate forecasts, a project
(1) experienced as a feeling state (with or without motivated by an interest in improving the communication
consciousness) and (2) demarcating a positive or negative of probabilistic seasonal climate forecasts to rural com-
quality of a stimulus. Affective responses occur rapidly and munities in Nakasongola and Kumi districts in Southern
automatically—note how quickly you sense the feelings Uganda (Phillips and Orlove, 2004). Phillips and Orlove
associated with the stimulus word ‘treasure’ or the word project developed and broadcasted a series of radio
‘hate.’ We argue that reliance on such feelings can be programs in local languages, including background in-
characterized as the affect heuristic.’’ It is important to formation on climate and climate variability and presenting
note that affect does not equal bias. the seasonal forecast for the upcoming season. Semi-formal
If something worries us, we try to reduce the source of groups of farmers were established to listen to the
worry. If something scares us, we may leave or avoid the programs and discuss them, often with a facilitator from
scene of the scare. Similarly, negative feelings can greatly the project. A ‘‘farmer-listening-group’’ program, includ-
influence risk perceptions and behavior. For example, a ing broadcasting of forecasts, interviews with farmers
national (US) survey found that negative feelings (affect) before and after the forecast meeting, as well as household
about global warming were stronger predictors of public surveys, are currently being repeated in a second district,
risk estimates and policy preferences than sociodemo- Rakai (Orlove and Roncoli, 2006).
graphic variables, values, or political variables (Leiserowitz, Based on preliminary analysis of interviews and group
2006). discussions, the affect heuristic, or risk-as-feelings, appears
Likewise, another study found that differences in worry in the ways that farmers anticipate and respond to the
about global warming were associated with differences in scientific seasonal forecasts, which is presented in analytical
willingness-to-pay more for gasoline, if such price increases terms as a set of probabilities. Farmers voice their fear that
would result in less harm to the environment (Hersch and unusually scarce or heavy rains could cause hunger for
Viscusi, 2006). The affect heuristic describes the fact that them and their households, evidenced by statements
our emotional reactions to a situation strongly influence that are strongly experiential. After hearing the probabil-
the way we subsequently act, i.e. that we use our affect as istic forecasts, individuals mention similar years that they
short-cut guide to action. If factors unrelated to the recall, describe the coming agricultural season that they
ARTICLE IN PRESS
54 S.M. Marx et al. / Global Environmental Change 17 (2007) 47–58
anticipate, and express relief or fear in relation to expected of 1967, the year when a new president abolished the
production outcomes. Farmers sometimes draw directly on traditional royal courts and associated institutions that had
personal experience, often echoed by other farmers. They continued under the British colonial rule and in the first
also make strongly analytical statements, including discus- years of independence. In both these cases, the affective
sion of the relative merit of different proposed courses of charge of these experiences was probably underscored by
action (choosing certain crop varieties or planting dates, its symbolic significance, since it is an established belief in
allocating labor to reduce the risk of landslides from Uganda, as in many other parts of the world, which a
nearby hills, reallocating labor from social and leisure stable legitimate ruler can assure not only social and
activities to field preparation and planting during critical political order, but cosmological and natural order as well.
weeks, requesting specific forms of assistance from Farmers mark certain years as particularly salient if
government agencies). Thus, the affect heuristic may direct extraordinary climatic conditions co-occurred with other
attention to a topic, but does not necessarily block affectively charged events. In turn, the affective charge of
analytical processing. the year may make the year more readily available for
recollection.
3.2. Availability heuristic The availability heuristic can play a large role in judging
the risk posed by climate variability (for instance El Niño),
Similar to the affect heuristic, personal experience also because people can usually recall unusually good or bad
plays a role in the availability heuristic, which is a rule of seasons. The response to long-term climate change
thumb that allows people to solve problems based on what information is different because most of us, especially the
they remember and how easily their memory is retrieved, younger generation, do not (yet) have experiences that we
how readily available that memory is. People have been associate with climate change, and cannot bring examples
found to employ the availability heuristic when asked for (whether frightening or pleasant) to mind. The availability
probability or frequency judgments, often of a comparative heuristic makes us assume that the future will be similar to
type (Tversky and Kahneman, 1983). When asked to judge what we have experienced so far (Sunstein, 2006).
whether the probability of thunderstorms is greater for July However, in the absence of personal familiarity people
or August, people will try to recall storms that they might draw on second-hand experience. A study of the
remember occurring in either July or August (‘‘just after movie ‘‘The Day After Tomorrow’’ suggests that movies,
the 4th of July,’’ ‘‘on my birthday, which is August 22’’). which integrate vivid imagery, strong emotions, and
Whichever category provides more available concrete vicarious experience, speak directly to the experiential
examples or for which it feels easier to generate examples processing system, in some cases bypassing critical analysis
is the one that is judged to be more likely. It is quite likely altogether (Leiserowitz, 2004). The vicarious experience
that the opposite process is also at work. When given the provided through cinema can, at times, have a significant
numerical probability of a climate event, farmers may effect on how people perceive, interpret and respond to
generate concrete examples of this event from their own risk. Movies often provide vivid, emotionally laden and
experience. The ease or frequency of generating such memorable images that can easily be brought to mind,
examples will then influence their intuitive understanding which drives the availability heuristic. For example,
and interpretation of the numeric probability information. anecdotal evidence suggests that the intense images and
To draw on our example of Ugandan farm communities theme music from the movie Jaws (1975) continue to affect
again (Orlove and Roncoli, 2006), the availability heuristic people’s intuitive assessments of the risk of shark attacks.
can also contribute to the tendency to translate probabil- Several other studies have addressed the power of images,
istic forecasts into deterministic information. When faced versus verbal or other information formats, in eliciting
with statistical information about general properties affective responses (e.g., Golding et al., 1992; Loewenstein,
(timing, amount) of precipitation, Ugandan farmers often 1996; Slovic et al., 1998; Jenkins-Smith, 2001; Leiserowitz,
turn to the familiar pattern of the two annual agricultural 2006), yet this is not the place for an extensive review of
seasons. To understand how a particular forecast may that literature.
relate to personally significant outcomes, farmers draw on
their personal experience with previous climate scenarios, 3.3. Recency heuristic
agricultural strategies, and food security impacts. An
interesting interaction of availability heuristic and affect In decisions based on summary statistical information
heuristic can be seen, for instance, in the recollection of people often overweight rare events, i.e. give them more
specific years. For example, older farmers in Nakasongola weight than they deserve by probability alone (Kahneman
district remembered the drought of 1961, because there was and Tversky, 1979). Yet in decisions based on personal
considerable tension in the region between ethnic groups experience, people tend to underweight rare events, i.e. give
during that year, a time when Uganda was moving towards them less weight than they deserve based on their
independence from British rule; they spoke of the flags that probability of occurrence. This results from the fact that
flew at the time, a vivid image still associated with the the experiential processing system gives a lot of weight to
drought. Farmers in other districts mentioned the droughts recent observations. Since rare events have generally not
ARTICLE IN PRESS
S.M. Marx et al. / Global Environmental Change 17 (2007) 47–58 55
occurred recently, they are underweighted in subsequent experiential information, though there is evidence that
decisions. Such underestimation of the risks of rare events vicarious experience is not as powerful as personal
based on experiential processing may contribute, for experience (Schotter and Sopher, 2003), probably because
example, to the neglect of flood control infrastructure by it fails to elicit the strong emotional reactions experienced
the federal government in recent decades. By the same in the face of positive or negative outcomes of one’s own
process (i.e. recency weighting) if the statistically rare event actions.
has occurred in the very recent past, people tend to
overreact to it. This makes decisions from personal 3.4. Limitations of heuristics of the experiential system
experience far more volatile than decisions based on
analytic processing of statistical information. The terrorist While ways of presenting information about (climate)
attacks of 11 September 2001 and the devastation caused risks that engage the experiential processing system have
by Hurricane Katrina in the Gulf of Mexico are examples clear benefits, they also have their downsides that argue for
where recent rare events may have led people to over- careful usage. As concern about one type of risk increases,
estimate the likelihood of subsequent similar events. In worry about other risks frequently decreases, as if people
summary, when people base their decisions on personal had a limited budget to spend on worry (Linville and
experience with a risky option, recent outcomes strongly Fischer, 1991). Hansen et al. (2004) found evidence for a
influence the evaluation of the risky option. As a result, finite pool of worry among farmers in the Argentine
low-probability events generate less concern than their Pampas. As concern with climate risk increased in the
probability warrants on average, but more concern than course of a decision simulation exercise that provided
they deserve in those instances where they do occur. climate information, concern with political risk went down
O’Connor et al. (2005) provide evidence of the operation even though the level of political risk had not changed.
of both the recency heuristic and the affect heuristic in a Concern and worry seemed to be a finite resource also in
survey of water managers in two eastern US states. The the case of each individual farmer. Those who stated
survey was motivated by concern that weather and climate greater worry about political risk worried less about
forecasts are being underutilized in the risk management of climate risk. Knowing that people’s capacity for worry or
community water systems. O’Connor et al. found that concern is finite places important limitations on attempts
managers who reported feeling at risk with respect to to raise greater concern and thus motivate more protective
weather events were far more likely to indicate use of or mitigative action against some risk, for example by
weather and climate forecasts. Feelings of risk, in turn, providing concrete images of possible damages. Raising
resulted from recent personal experience, i.e. the experience concern about some aspect of a situation comes at the cost
of adverse weather events over the past 5 years. Similarly, of potentially reducing concern about another. The costs of
Orlove and Roncoli (2006) found in their work with worry also appear to be cumulative, i.e. the pool of worry
farmers in the Ugandan Rakai district that farmers appears to take time to regenerate. Emotional numbing as
describe climate variability largely in terms of the previous the result of repeated exposures to emotionally draining
5 years or so, though, as indicated above, farmers also situations is a commonly observed reaction in individuals
recall a few years from previous decades that were marked living in war zones or being subjected to repeated hurricane
by other events. Farmers in Rakai are particularly threats within a season—especially in the modern media
concerned by recent years in which the rains started environment where people experience a bewildering num-
unusually early or late, contrasting with their beliefs about ber and diversity of vicarious experiences each day, ranging
the normal seasonal cycle; they are far less likely to report from news stories, television dramas, educational pro-
unusual onsets further in the past, though these had also grams, confessional talk shows, movies, novels, video
occurred. Other factors may have also reinforced this games, etc. One may be moved by a movie one day, but
recency effect, particularly the strong emphasis that NGOs the feeling may be replaced by a new vicarious experience
and government environmental agencies have placed on the following day.
the link between deforestation and regional climate change Another suboptimal consequence of affective processing
in the course of environmental awareness campaigns is the single-action bias (Weber, 1997). It describes a
implemented in the last decade. propensity to take only one action to respond to a problem
This tendency to give recent events a lot of weight is, of in situations where a broader set of remedies is called for.
course, adaptive in non-stationary environments. More Taking the first action to respond to the problem at hand
recent observations are more predictive of future occur- seems to reduce or remove the feeling of worry or concern.
rences when the relationships and patterns in one’s With this affective marker, motivation for further action is
environment can be expected to change. In this sense, reduced. For example, malaria-control efforts in regions
people appear to be well equipped to detect and adapt to with wide-spread chloroquine resistance might focus
climate change, if we can provide them with sufficient on procurement of alternative drugs such as artemisinin-
personal experience. Group discussions and the sharing of based combination therapy, but omit other protective
personal experiences and outcome feedback among group actions such as insecticide-treated bednets or indoor
members is one way of broadening the exposure to residual spraying (Marx, 2006). Weber (1997) found that
ARTICLE IN PRESS
56 S.M. Marx et al. / Global Environmental Change 17 (2007) 47–58
Midwestern farmers engaged in only one of three plausible understanding of both probabilities and the consequences
protections against the negative consequences of climate of incremental change and extreme events, and motivate
change. Hansen et al. (2004) similarly found that farmers in contingency planning.
Argentina employed only one of several adaptations to Yet, while the engagement of experience-based, affective
climate variability and climate change. If they had the decision-making can make risk communications more
capacity to store grain, for example, they were less likely to salient and motivate behavior, experiential processing is
also irrigate and invest in crop insurance. also subject to its own biases, limitations and distortions,
A way to overcome some of these limitations associated such as the finite pool of worry and single action bias.
with affective processing in decision-making under un- Experiential processing works best with easily imaginable,
certainty, namely the finite pool of worry and single-action emotionally laden material, yet many aspects of climate
bias, is group discussion. For example, the Uganda case variability and change are relatively abstract and require a
(Nakasongola) shows that participation in group discus- certain level of analytical understanding (e.g., long-term
sion helped elicit a wider range of adaptive responses to trends in mean temperatures or precipitation). Ideally,
climate forecasts than when forecasts were presented to communication of climate forecasts should encourage the
individual farmers (Orlove and Roncoli, 2006). The total interactive engagement of both analytic and experiential
number of possible responses that were mentioned by all processing systems in the course of making concrete
the participants increased from four mentioned by decisions about climate, ranging from individual choices
individuals before the group meeting to ten when about what crops to plant in a particular season to broad
interviewed after the group session. This increase cannot social choices about how to mitigate or adapt to global
be explained alone by mechanisms of sharing or pooling of climate change.
information in groups (Stasser, 1992; Larson et al., 1998). One way to facilitate this interaction is through group
Moreover, the percentage of participants mentioning and participatory decision-making. As the Uganda exam-
specific responses increased. The percentage who re- ple suggests, group processes allow individuals with a range
sponded that they would alter the timing of planting of knowledge, skills and personal experience to share
increased from 16.0% to 24.8% between the two surveys. diverse information and perspectives and work together on
In the pre-program survey, none of the respondents a problem. Ideally, groups should include at least one
mentioned the selection of particular varieties or the use member trained to understand statistical forecast informa-
of fertilizer as responses; in the post-program survey, tion to ensure that all sources of information—both
13.3% and 3.7%, respectively, indicated these actions. experiential and analytic—are considered as part of the
While these results are preliminary and we do not know yet decision-making process. Communications to groups
if farmers followed their suggestions with real actions, this should also try to translate statistical information into
difference in the number of responses before and after formats readily understood in the language, personal and
group discussion indicates that farmers considered, or at cultural experience of group members. In a somewhat
least spoke about, a much larger number of alternatives iterative or cyclical process, the shared concrete informa-
after participation in a group. tion can then be re-abstracted to an analytic level that leads
to action.
4. Research and policy implications Risk and uncertainty are inherent dimensions of all
climate forecasts and related decisions. Analytic products
Based on the observation that experiential and analytic like trend analysis, forecast probabilities, and ranges of
processing systems compete and that personal experience uncertainty ought to be valuable contributions to stake-
and vivid descriptions are often favored over statistical holder decision-making. Yet decision makers also listen to
information, we suggest the following research and policy the inner and communal voices of personal and collective
implications. experience, affect and emotion, and cultural values. Both
Communications designed to create, recall and highlight systems—analytic and experiential—should be considered
relevant personal experience and to elicit affective re- in the design of climate forecasts and risk communications.
sponses can lead to more public attention to, processing of, If not, many analytic products will fall on deaf ears as
and engagement with forecasts of climate variability and decision makers continue to rely heavily on personal
climate change. Vicarious experiential information in the experience and affective cues to make plans for an
form of scenarios, narratives, and analogies can help the uncertain future. The challenge is to find innovative and
public and policy makers imagine the potential conse- creative ways to engage both systems in the process of
quences of climate variability and change, amplify or individual and group decision-making.
attenuate risk perceptions, and influence both individual
behavioral intentions and public policy preferences. Like- Acknowledgements
wise, as illustrated by the example of retranslation in the
Uganda studies, the translation of statistical information Preparation of this paper was facilitated by US National
into concrete experience with simulated forecasts, decision- Science Foundation Grants SES-0345840 and SES-
making and its outcomes can greatly facilitate an intuitive 0435622, and US National Oceanic and Atmospheric
ARTICLE IN PRESS
S.M. Marx et al. / Global Environmental Change 17 (2007) 47–58 57
Administration Grant NA06GP0308. We would like to as a Key to Understanding Adaptive Cognition in an Uncertain
acknowledge that valuable comments were provided by Environment. Cambridge University Press, New York, pp. 72–91.
Roberta Balstad and Kenneth Broad (co-directors of the Hirshleifer, D., Shumway, T., 2003. Good day sunshine: stock returns and
the weather. Journal of Finance, American Finance Association 58 (3),
Center for Research on Environmental Decisions, CRED). 1009–1032.
Furthermore, we thank two anonymous reviewers and the Holtgrave, D., Weber, E.U., 1993. Dimensions of risk perception for
members of the Krantz/Weber/CRED lab for constructive financial and health and safety risks. Risk Analysis 13, 553–558.
feedback and suggestions. Inhelder, B., Piaget, J., 1958. The growth of logical thinking from
childhood to adolescence. Basic Books, New York.
Jenkins-Smith, H., 2001. Modeling stigma: an empirical analysis
References of nuclear images of Nevada. In: Flynn, J., Slovic, P.,
Kunreuther, H. (Eds.), Risk, Media and Stigma. Earthscan, London,
Argote, L., Devadas, R., Melone, N., 1990. The base-rate fallacy: pp. 107–131.
contrasting processes and outcomes of group and individual judgment. Jenni, K.E., Loewenstein, G., 1997. Explaining the ‘‘identifiable victim
Organizational Behavior and Human Decision Processes 46, 296–310. effect’’. Journal of Risk and Uncertainty 14, 235–237.
Ball, L.K., Evans, G., Bostrom, A., 1998. Risky business: challenges in Kahneman, D., Tversky, A., 1979. Prospect theory: an analysis of decision
vaccine risk communication. Pediatrics 101, 453–458. under risk. Econometrica 47, 263–291.
Barsade, S.G., 2002. The ripple effect: emotional contagion and its Kerr, N.L., MacCoun, R.J., Kramer, G.P., 1996. Bias in judgment:
influence on group behavior. Administrative Science Quarterly 47 (4), comparing individuals and groups. Psychological Review 103 (4),
644–675. 687–719.
Bert, F., Satorre, E.H., Toranzo, F., Podestá, G., 2006. Climatic Kerr, N., Tindale, R., 2004. Group performance and decision making.
information and decision-making in maize crop production systems Annual Review of Psychology 55, 623–655.
of the Argentinean Pampas. Agricultural Systems 88, 180–204. Klos, A., Weber, E.U., Weber, M., 2005. Risk perception and risk
Birnbaum, M.H., Mellers, B.A., 1983. Bayesian inference: combining base behavior in repeated gambles. Management Science 51, 1777–1790.
rates with opinions of sources who vary in credibility. Journal of Koehler, J.J., 1996. The base rate fallacy reconsidered: descriptive,
Personality and Social Psychology 45 (4), 792–804. normative, and methodological challenges. Behavioral and Brain
Chaiken, S., Trope, Y., 1999. Dual Process Theories in Social Psychology. Sciences 19, 1–53.
New York, Guilford Publications. Kraemer, C., Weber, M., 2004. How do people take into account
Chu, Y.P., Chu, R.L., 1990. The subsidence of preference reversals in weight, strength and quality of segregated vs. aggregated data?
simplified and market-like experimental settings: a note. The American Experimental evidence. Journal of Risk and Uncertainty 29 (2),
Economic Review 80, 902–911. 113–142.
Clemen, R.T., 1996. Making Hard Decisions: An Introduction to Decision Larson, J., Christensen, C., et al., 1998. Diagnosing groups: the pooling,
Analysis. Duxbury Press, Belmont. management, and impact of shared and unshared case information in
Clemen, R.T., Winkler, R.L., 1999. Combining probability distributions team-based medical decision making. Journal of Personality and Social
from experts in risk analysis. Risk Analysis 19 (2), 187–203. Psychology 75 (1), 93–108.
Cohen, M., Jaffray, J.Y., Said, T., 1987. Experimental comparison of Leiserowitz, A., 2004. Surveying the impact of ‘‘the day after tomorrow’’.
individual behavior under risk and under uncertainty for gains and for Environment 46 (9), 23–44.
losses. Organizational Behavior and Human Decision Processes 39, Leiserowitz, A., 2006. Climate change risk perception and policy
1–22. preferences: the role of affect, imagery, and values. Climatic Change
Damasio, A., 1995. Descartes’ Error: Emotion, Reason, and the Human 77 (1–2), 45–72.
Brain. Harper Collins, New York. Linville, P.W., Fischer, G.W., 1991. Preferences for separating and
Epstein, S., 1994. Integration of the cognitive and the psychodynamic combining events: a social application of prospect theory and the
unconscious. American Psychologist 49, 709–724. mental accounting model. Journal of Personality and Social Psychol-
Fagerlin, A., Wang, C., Ubel, A.P., 2005. Reducing the influence of ogy Bulletin 60, 5–23.
anecdotal reasoning on people’s health care decisions: is a picture Loewenstein, G., 1996. Out of control: visceral influences on behavior.
worth a thousand statistics? Medical Decision Making 25, 398–405. Organizational Behavior and Human Decision Processes 65 (3),
Golding, D., Krimsky, S., Plough, A., 1992. Evaluating risk communica- 272–292.
tion: narrative vs. technical presentation of information about radon. Loewenstein, G., Weber, E.U., et al., 2001. Risk as feelings. Psychological
Risk Analysis 12 (1), 27–35. Bulletin 127 (2), 267–286.
Hamill, R., Wilson, T.D., Nisbett, R.E., 1980. Insensitivity to sample bias: Marx, S.M., 2006. Climate and decision making: the case of malaria
generalizing from atypical cases. Journal of Personality and Social control in sub-Saharan Africa. Center for Research on Environmental
Psychology 39, 578–589. Decisions and International Research Institute for Climate and
Hansen, J., Marx, S., Weber, E.U., 2004. The Role of climate perceptions, Society, Columbia University. Working paper.
expectations, and forecasts in farmer decision making: the Argentine Milch, K., 2006. Framing effects revisited. Master Thesis, Department of
Pampas and South Florida. IRI Technical Report 04-01. International Psychology, Columbia University.
Research Institute for Climate Prediction, Palisades, NY. Morgan, M.G., Henrion, M., 1990. Uncertainty. Cambridge University
Hersch, J., Viscusi, W.K., 2006. The generational divide in support for Press, Cambridge.
environmental policies: European evidence. Climatic Change 77 (1–2), Nisbett, R.E., Ross, L., 1980. Human Inference: Strategies and Short-
121–136. comings of Social Judgment. Prentice-Hall, Englewood Cliffs, NJ.
Hertwig, R., Ortmann, A., 2001. Experimental practices in economics: a O’Connor, R.E., Yarnal, B., Dow, K., Jocoy, C.L., Carbone, G.J., 2005.
methodological challenge for psychologists. Behavioral and Brain Feeling at risk matters: water managers and the decision to use
Sciences 24, 383–451. forecasts. Risk Analysis 25, 1265–1273.
Hertwig, R., Barron, G., Weber, E.U., Erev, I., 2004. Decisions from Orlove, B.S., Kabugo, M., 2005. Signs and sight in southern Uganda:
experience and the effect of rare events. Psychological Science 15, representing perception in ordinary conversation. Etnofoor 18 (1),
534–539. 124–141.
Hertwig, R., Barron, G., Weber, E.U., Erev, I., 2006. Rare risky Orlove, B.S., Roncoli, C., 2006. Integration of Climate Information from
prospects: different when valued through a window of sampled Multiple Sources through Group Discussion in Ugandan Farm
experiences. In: Fiedler, K., Juslin, P. (Eds.), Information Sampling Communities /www.cred.columbia.edu/research/projects/uganda/S
ARTICLE IN PRESS
58 S.M. Marx et al. / Global Environmental Change 17 (2007) 47–58
Paese, P.W., Bieser, M., Tubbs, M.E., 1993. Framing effects and choice Stasser, G., 1992. Pooling of Unshared Information During Group
shifts in group decision making. Organizational Behavior and Human Discussion. Sage Publications, Thousand Oaks.
Decision Processes 56, 149–165. Sunstein, C.R., 2006. The availability heuristic, intuitive
Payne, J.W., Bettman, J.R., Coupey, E., Johnson, E.J., 1992. A cost-benefit analysis, and climate change. Climatic Change 77 (1–2),
constructive process view of decision making: multiple strategies in 195–210.
judgment and choice. Acta Psychologica 80 (1–3), 107–141. Suarez, P., Patt, A., 2004. Cognition, caution, and credibility: the
Phillips, J., Orlove, B.S., 2004. Improving climate forecast communica- risks of climate forecast application. Risk Decision and Policy 9,
tions for farm management in Uganda. Final Report to NOAA Office 75–89.
of Global Programs. Tversky, A., Kahneman, D., 1974. Judgment under uncertainty: heuristics
Piaget, J., Inhelder, B., 1962. The Psychology of the Child. Basic Books, and biases. Science 185, 1124–1130.
New York. Tversky, A., Kahneman, D., 1981. The framing of decisions and the
Pidgeon, N.F., Kasperson, R.K., Slovic, P., 2003. The Social Amplifica- psychology of choice. Science 211, 453–458.
tion of Risk. CUP, Cambridge. Tversky, A., Kahneman, D., 1982. Evidential impact of base rates. In:
Schotter, A., Sopher, B., 2003. Social learning and convention creation in Kahneman, D., Slovic, P., Tversky, A. (Eds.), Judgment Under
inter-generational games: an experimental study. Journal of Political Uncertainty: Heuristics and Biases. Cambridge University Press, New
Economy 111 (3), 498–529. York, pp. 153–160.
Simon, H.A., 1956. Rational choice and the structure of the environment. Tversky, A., Kahneman, D., 1983. Extensional vs. intuitive reasoning: the
Psychological Review 63, 129–138. conjunction fallacy in probability judgment. Psychological Review 90,
Sinaceur, M., Heath, C., 2004. Emotional and deliberative reactions to a 293–315.
public crisis: mad cow disease in France. Psychological Science 16 (3), Tversky, A., Kahneman, D., 1986. Rational choice and the framing of
247–254. decisions. Journal of Business 59, 251–278.
Sloman, S.A., 1996. The empirical case for two systems of reasoning. Tversky, A., Koehler, D., 1994. Support theory: a nonextensional
Psychological Bulletin 1 (119), 3–22. representation of subjective probability. Psychological Review 101,
Slovic, P., MacGregor, D.G., Peters, E., 1998. Imagery, Affect, and 547–567.
Decision-Making. Decision Research, Eugene. Weber, E.U., 1997. The utility of measuring and modeling perceived risk.
Slovic, P., Finucane, M., Peters, E., MacGregor, D.G., 2002. The affect In: Marley, A.A.J. (Ed.), Choice, Decision, and Measurement: Essays
heuristic. In: Gilovich, D.G.T., Kahneman, D. (Eds.), Intuitive in Honor of R. Duncan Luce. Lawrence Erlbaum Associates,
Judgment: Heuristics and Biases. Cambridge University Press, Mahwah, pp. 45–57.
New York. Weber, E.U., Böckenholt, U., Hilton, D.J., Wallace, B., 1993. Determi-
Slovic, P., Finucane, ML., Peters, E., MacGregor, D.G., 2004. Risk as nants of diagnostic hypothesis generation: effects of information, base
analysis and risk as feelings: some thoughts about affect, reason, risk, rates, and experience. Journal of Experimental Psychology: Learning,
and rationality. Risk Analysis 24 (2), 311–322. Memory, and Cognition 19, 1151–1164.
Stanowich, K.E., West, R.F., 1998. Individual differences in rational Weber, E.U., Shafir, S., Blais, A.R., 2004. Predicting risk sensitivity in
thought. Journal of Experimental Psychology, General 127 (2), humans and lower animals: risk as variance or coefficient of variation.
161–188. Psychological Review 111, 430–445.