Erschienen in: The Sage handbook of risk communication / Cho, Hyunyi et al. (Hrsg.). - Los Angeles, Calif. [u.a.

] : Sage, 2015.
- S. 10-23. - ISBN 978-1-4522-5868-3

Risk Perception
Nicolai Bodemer and Wolfgang Gaissmaier

Introduction can refer to the number of fatalities, which can be

measured through the probability of death, expected
Natural hazards, health hazards, terrorist attacks, life-years lost, deaths per person exposed, or total
new technologies, transportation—all of them deaths (Fischhoff, 2009, p. 943). More generally,
represent risks in our life. We face some of these most definitions of risk have in common that risk
risks daily, others rarely, if ever. Some risks con- comprises two factors: (1) the probability of harm
stitute a threat to individuals, some to the entire and (2) the magnitude of harm, where harm refers
society. We overestimate some risks, while under- to threats to humans and things they value
estimating others. Some risks trigger a strong (Hohenemser, Kates, & Slovic, 1985). In this sense,
emotional response, others are perceived more risks are measureable and risky situations are condi-
“cold” and rational. tions in which outcomes (i.e., the harm) and prob-
In this chapter, we start with a definition of the abilities are known. For instance, one can access the
concept of risk and how it differs from the concept risk of losing in the casino, the risk of an adverse
of uncertainty. We then outline major theories, event of a specific treatment, or the risk of a car
models, and mediators that influence our percep- accident for a given population. Risky situations
tion of risk. Although the models stem from differ- have to be distinguished from uncertain situations
ent research programs, highlight different that are conditions in which either both, the out-
mechanisms, and are often discussed in isolation comes and probabilities, or at least the probabilities
from each other, they serve as psychological expla- are unknown (see Meder, Le Lec, & Osman, 2013).
nations for how we perceive risks in our daily life. This distinction goes back to Knight (1921) who
defined uncertainty as immeasurable in contrast to
measureable risks. However, more recent
Defining Risk and approaches broaden the definition of uncertainty as
Risk Perception they consider probabilistic parameters such as stan-
dard deviations, expert confidence ratings, or
Risk is a highly interdisciplinary concept and its ranges as (quantifiable) indicators of uncertainty
measurement differs across and even within (e.g., Politi, Han, & Col, 2007). Throughout this
disciplines. For instance, in a health context, risk chapter, we mainly refer to risks, that is, situations

10

Konstanzer Online-Publikations-System (KOPS)

URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-0-287262
 11

in which outcomes and probabilities are measure- remember that risk perception always takes place
able. Furthermore, when talking about uncertainty, in a social and cultural context (see Part I,
we mean that probabilities are not measureable, and Section 2). The social amplification of risk frame-
we use the term ambiguity to designate situations in work is an integrative model of risk perception
which probabilistic parameters are used to quantify that is grounded on the assumption that “risk . . .
uncertainty of risk estimates. is not only an experience of physical harm but
Risk perception refers to the “subjective assess- the result of processes by which groups and indi-
ment of the probability of a specified type of acci- viduals learn to acquire or create interpretations
dent happening and how concerned we are with the of risk” (Kasperson et al., 2003, p. 13). Hence,
consequences” (Sjöberg, Moen, & Rundmo, 2004, risk perception has to be understood as a com-
p. 8z). Pidgeon, Hood, Jones, Turner, and Gibson munication process along a chain from the
(1992) defined risk perception as “people’s beliefs, sender to the receiver, with different stations in
attitudes, judgments and feelings, as well as the between that may amplify or attenuate risks.
wider social or cultural values and dispositions that Such stations can be social (e.g., news media),
people adopt, towards hazards and their benefits” individual (e.g., attention filter), or institutional
(p. 89) and thereby stressed that risk perception is (e.g., political and social actions).
not simply an individual process but has to be The way we perceive and react to risks is fur-
understood against the societal and cultural back- ther shaped by our values. In their cultural the-
ground (see also Kasperson, Kasperson, Pidgeon, & ory of risk, Douglas and Wildavsky (1983)
Slovic, 2003). distinguished between different cultural world-
Furthermore, evaluating a risk from different views. The major dimensions are individualism
perspectives has different implications for the (e.g., defending individual freedom) versus com-
evaluation of such a risk, as well as consequences munitarian (e.g., supporting collective action)
for behavior. Imagine a woman who knows that and hierarchical (e.g., leaving important deci-
of 100,000 women like her, 15 will have cervical sions to experts) versus egalitarian (e.g., striving
cancer. She might decide not to participate in for equality). For instance, people with hierarchi-
prevention, such as screening with a pap smear cal–individualist worldviews value markets and
(a test to identify early stages of cervical commerce and feel uneasy when these activities
dysplasia), because her baseline risk is rather low. are restricted. On the other hand, people with
Now imagine a health policy maker: Pap smear egalitarian–communitarian worldviews are
screening reduces the annual incidence of cervi- ambivalent about markets. The interaction of
cal cancer in Germany by a total 10,400 women. personal values and risk perception idea is also
In this case, a national program to implement implemented in the cultural cognition of risk
pap smear screening might be appreciated thesis.
(example from Neumeyer-Gromen, Bodemer,
Müller, & Gigerenzer, 2011). Thus, depending on
whether a policymaker or an individual layper- The Psychometric Paradigm
son decides about the screening, the benefit will
be evaluated differently. One of the earliest and most influential models in
risk perception is the psychometric paradigm
(Slovic, 1987). Based on the assumption that risks
Social and Cultural are quantifiable and predictable, the psychometric
Components of Risk Perception paradigm uses psychophysical scaling and multi-
variate techniques to explain laypeople’s reactions
Although the focus of this chapter is on risk per- to hazards. In their seminal study, Fischhoff,
ceptions of individuals, it is important to Slovic, Lichtenstein, Read, and Combs (1978)
12—

asked participants to rate 30 activities and tech- across participants and across hazards. In this
nologies (e.g., fire fighting, nuclear power, pesti- case, up to 80% of the variance of perceived risk
cides, motor vehicles, smoking, food coloring) on can be explained with the two factors, however, if
nine different dimensions they had previously one considers only one specific hazard each time
identified in the literature: (1) voluntariness of the explained variance drops to about 20%
risk, (2) immediacy of effect, (3) knowledge about (Gardner & Gould, 1989; Sjöberg, 1996, 2002b).
risk by scientist, (4) knowledge about risk by Hence, the original aggregated analysis of per-
exposed person, (5) control over risk, (6) newness, ceived risk in the psychometric paradigm does
(7) catastrophic potential, (8) dreadfulness, and not give information about individual variation
(9) severity of consequences. Each dimension was as well as intraindividual perceptions across dif-
assessed on a 7-point scale. A factor analysis based ferent hazards. Sjöberg (2002a) challenged the
on the correlation of the mean ratings revealed assumption that experts differ in their underly-
two orthogonal factors: dread risk and unknown ing processes and assess risks via annual fatalities
risk (sometimes called novelty). High dread risk only. In fact, he found that experts’ and laypeo-
means that perceived lack of control, dread, and ple’s risk perceptions were rather similar.
catastrophic potential and fatal consequences are However, one possible explanation is that lay-
high (e.g., nuclear reactor incidents); high people and experts follow different definitions of
unknown risk means that the hazard is unobserv- risk, as experts primarily refer to an objective,
able, unknown, new, and delayed (e.g., chemical numerical assessment, whereas laypeople are
and DNA technologies; Slovic, 1987). Note that more affect driven (Slovic, 1999). In sum, the
unknown risk is different from uncertainty and psychometric paradigm has been an influential
ambiguity as defined above, as unknown risk is a model for describing risk perceptions and has
psychological construct and refers primarily to the yielded important insights and impulses. At the
fact that the risk is novel, yet it could, in theory, be same time, its application to predict individual
objectively quantified. The two factors (dread and risk perception is limited and it does not provide
unknown risk) have been observed in various a process model that explains the cognitive
studies across countries to predict laypeople’s reac- mechanisms underlying risk perception.
tions to hazards and have therefore been assumed
as robust across different cultures and environ-
ments (Boholm, 1998; Slovic, 2000). The factor Dread Risk: The Role of the
dread risk was found to better predict laypeople’s Social Circle
risk perception and desire for risk regulations
compared with the factor unknown risk (Slovic, As proposed by the psychometric paradigm,
1987). It is important to keep in mind that the dread is an important predictor in people’s per-
model was primarily intended to explain risk per- ception of and reaction to hazards. The dread
ception in laypeople, because they give relatively hypothesis further proposes that people have a
little weight to actual risk assessment (Covello, strong tendency to avoid risks that kill many
von Winterfeldt, & Slovic, 1987). Experts, in con- people at once, compared with risks that cause
trast, defined risks strictly in terms of annual the same number of fatalities over a longer
fatalities, at least, according to findings from the period of time (Slovic, 1987). For instance, the
psychometric paradigm on small expert samples terrorist attack of 9/11 represents a dread risk. As
(Slovic, Fischhoff, & Lichtenstein, 1979). a consequence of the attack, many people avoided
The psychometric paradigm is not without its airplanes and switched to cars instead, which
critics (Sjöberg, 2002b, 2003). For instance, the increased fatalities in car accidents (Gaissmaier &
model primarily predicts risk perception on an Gigerenzer, 2012; Gigerenzer, 2004, 2006). At the
aggregated level—that is, using average ratings same time, the fact that between 44,000 and
 —13

98,000 patients die in the United States annually kills a number of people immediately and (2) it
due to preventable medical errors is not per- reduces the number of future offspring by reduc-
ceived as a dread risk. At least four possible ing the number of their potential parents. Hence,
explanations can account for the fact that people a risky event that kills young people—potential
tend to fear dread risks more than continuous parents for future generations—strongly influ-
risks, even if both cause the same number of ences population size. Due to the fact that dread
fatalities (Galesic & Garcia-Retamero, 2012; risks kill many people once and often affect
Gigerenzer, 2006). First, dread risks are perceived younger generations also, it takes longer for the
as less controllable. The findings from the psy- population to recover from dread risks than from
chometric paradigm suggest that lack of control equivalent continuous risks.
loads highly on the factor dread risk (Slovic,
1987). Whereas we may assume high control
when driving, which yields in low risk percep- The Role of Affect: Risk-as-
tion, we may assume low control over earth- Feelings and the Affect Heuristic
quakes and terrorist attacks, which yields in high
risk perception. Second, people are not aware of Affective reactions provide important signals
the actual underlying statistical information. In about how we perceive and “feel” about our envi-
particular, knowledge about continuous, every- ronment (see also Chapter 3, this volume). Affect
day risks may be underestimated as they are less emerges automatically and quickly, often before a
salient compared with dread risks that, in turn, cognitive and conscious evaluation of the situa-
may be overestimated. Third, from an evolution- tion takes place. Affective reactions allow evalua-
ary perspective, an event killing many group tion of the target, guide what information we
members once imposed a substantial threat to search for and focus on, motivate behavior, and
the survival of the group when still living in allow comparison of different events and situa-
hunter–gatherer societies. Hence, people might tions on a common level (Peters, 2006). Risks can
be prepared to fear particularly those risks that trigger emotions in two dimensions: (1) immedi-
threaten survival of their group. Galesic and ate emotion, when one is confronted with the
Garcia-Retamero (2012) examined this hypoth- risk as integral emotions that are caused by the
esis and tested to what extent risks that affect a risk itself or incidental emotions that are caused
number of people corresponding to the typical by other, unrelated factors yet influence risk per-
size of our social circle (e.g., family, friends) are ception, and (2) anticipated emotions, which are
perceived as more relevant. They defined the expected to be experienced in the future.
social circle as a group of up to 200 people. Two major approaches have been proposed to
Results of nine experiments consistently showed study the role of affect in risk perception. The
that people perceived a risk killing 100 people as risk-as-feelings hypothesis (Loewenstein, Weber,
more dreadful and frightening than a risk killing Hsee, & Welch, 2001) assumes a dual-process
10 people. However, a risk killing 1,000 people model according to which people assess risks
was rated equal to a risk killing 100 people, sug- cognitively and emotionally. In general, the emo-
gesting that the number of people corresponding tional appraisal is considered as stronger than the
to the social circle is crucial in defining the dread cognitive appraisal. The affect heuristic describes
potential and related fear. Fourth, and related, is how affective reactions influence risk perception.
the finding that dread risks have a stronger According to this heuristic, people consult their
impact on the population size over time than “affect pool” as a cue about the judgment of a risk
continuous risks causing the equivalent number (Finucane, Alhakami, Slovic, & Johnson, 2000;
of fatalities (Bodemer, Ruggeri, & Galesic, 2013). Pachur, Hertwig, & Steinmann, 2012). For
The reason is that a fatal event strikes twice: (1) it instance, when comparing two risks, a person
14—

might infer that the risk that evokes a stronger received either a high anchor (motor vehicle
emotional reaction (in terms of dread) is more accidents cause 50,000 deaths per year) or a low
prevalent. anchor (electrocution causes 1,000 deaths per
Affective reactions also account for how peo- year). Participants’ absolute risk judgments were
ple simultaneously perceive benefits and risks. influenced by the anchor. When the high anchor
Benefits and risks of technologies or medical was provided, risk estimates were about two to
treatments are usually positively correlated in five times higher than when the low anchor was
our environment: A greater benefit goes along provided. When it comes to relative judgments,
with a greater risk. Yet people’s perceptions of participants performed better, and more frequent
benefits and risks are usually negatively corre- risks were generally identified as more likely
lated. Technologies with high benefits are per- when compared with a less frequent risk.
ceived as less risky, and high-risk technologies However, two biases were obtained. First, partici-
are perceived as less beneficial. Alhakami and pants overestimated low frequencies and under-
Slovic (1994) explained this pattern with partici- estimated high frequencies (primary bias).
pants’ affective evaluation of technologies. Second, participants assigned different ratios to
Favorable affective evaluations result in high- different pairs of causes of death even when the
benefit and low-risk perceptions, whereas unfa- ratio was the same (secondary bias). The avail-
vorable affective evaluations result in the opposite ability heuristic was proposed to explain these
pattern (see also Slovic, Finucane, Peters, & biases: People base their estimates on recalled
MacGregor, 2002). instances. If the recall of available instances in the
mind is proportional to the actual frequency of
the event, then people correctly assess the risk.
Availability Heuristic However, when the recall of instances is not pro-
portional to the actual frequency, then people
Another strategy is to judge risks via the avail- may misjudge risks that could explain the pri-
ability heuristic, that is, to judge “the frequency mary bias.
of a class or the probability of an event by the ease Two major criticisms with respect to the avail-
with which instances or occurrences can be ability heuristic have been raised (e.g., Hertwig,
brought to mind” (Tversky & Kahneman, 1974, Pachur, & Kurzenhäuser, 2005). First, the heuris-
p. 1127). Whether and when this heuristic leads tic has often been used as a post hoc explanation
to accurate risk perception depends on the struc- rather than to predict risk perception patterns.
ture of the environment. Assuming that more Second, the original definition does not distin-
frequent events are easier to recall, risk percep- guish between ease of recall and actual number
tion should be quite accurate. However, factors of recalled instances. Hertwig and colleagues
such as memorability, imaginability, or dispro- (2005) addressed the two criticisms and com-
portional media coverage of an event can bias pared several models that specified these pro-
risk perception. In their seminal work, cesses precisely. Their results were similar to
Lichtenstein, Slovic, Fischhoff, Layman, and those from Lichtenstein et al. (1978). Moreover,
Combs (1978) used two different methods to the authors compared two different versions of
elicit participants’ frequency judgments. In one the availability heuristic: First, availability-by-
method, they presented participants pairs of recall operationalized by the number of instances
causes of death and asked them first to state one recalls from one’s social circle (e.g., family
which causes a higher number of deaths and sec- members, friends, colleagues). Second, fluency
ond to estimate the ratio of their frequencies. In operationalized by the anticipated ease of recall
another method, participants had to estimate the with which instances can be brought to mind.
frequency of an event and prior to estimation Fluency was modeled in two ways: (1) the speed
 —15

with which instances come to mind and (2) the ratings with average population risks. This is due
occurrence of events in the media. Availability- to the fact that one has usually no knowledge
by-recall better predicted people’s choice than about the actual risk of an individual.
fluency, and it suggests that, at least in this con- Different explanations for optimism bias have
text, availability may be primarily defined by the been proposed. Shepperd, Carroll, Grace, and Terry
number of recalled instances. However, it should (2002) identified four broad categories: (1) the
be noted that there are situations in which avail- desired end state of comparative judgments (e.g.,
ability-by-recall is not applicable, namely, when a self-enhancement), (2) cognitive mechanisms (e.g.,
risky event has not (yet) occurred in one’s social representativeness heuristic), (3) information about
circle. Yet the social environment seems to play self versus target (e.g., person–positivity bias; dis-
an important role in the evaluation of risks. counting background information), and (4) under-
Furthermore, the author also tested other possi- lying affect (e.g., mood congruency; for an overview,
ble mechanisms that could explain the data. For see also Chambers & Windschitl, 2004).
instance, regressed frequency assumes that peo- But is thinking to be better than average
ple monitor the occurrence of risks but tend to always a bias? Studies found that the majority of
overestimate small and underestimate low risks people believe that they drive more safely than
as a consequence of a regression-to-the-mean the average (Johansson & Rumar, 1968; Svenson,
effect. In a later study, Pachur et al. (2012) also 1981)—which is considered a bias as it is simply
compared whether availability-by-recall or the not possible that the majority are better than
affect heuristic better described people’s risk average. However, a closer look at the distribu-
judgments and found that availability-by-recall tion of car accidents shows that most drivers
was a stronger predictor than affect for frequency have few accidents, and few drivers have many
estimates, although the affect heuristic described accidents. In such a nonsymmetric distribution,
participants’ value of a statistical life and per- the median and the mean are not identical
ceived risk similarly well. (Gigerenzer, Fiedler, & Olsson, 2012). Hence, the
majority of drivers (i.e., more than 50%) have
actually less accidents than the average. For
Optimism Bias instance, of 7,800 drivers in the United States,
80% had fewer accidents than average; and of 440
When asked about their risk in comparison to German drivers, 57% had fewer accidents than
the average risk, people often show unrealistic average. Hence, in this case, the better-than-
optimism: They believe themselves to be better average effect is no bias.
off and less likely to experience negative life
events (or more likely to experience positive
events) than others (Weinstein, 1987). Optimism Representation of Risk:
bias, also termed the above-average effect and Fuzzy-Trace Theory
comparative optimism, serves as an explanation
of why people often do not take precautions and The way we perceive risks strongly depends on
instead simply discount their personal risk (“It how we mentally represent them. Fuzzy-trace
won’t happen to me”). For example, when theory addresses this issue and distinguishes
Weinstein (1987) presented participants with 32 between two kinds of representations: verbatim
different life hazards, participants on average and gist (Reyna, 2008; Reyna & Brainerd, 1995).
rated their chance of experiencing the hazard to Verbatim representations encode the stimulus
be below average on 25 of them. However, it is objectively—that is, as it actually happened. For
important to note that optimism bias is primarily example, imagine that mammography screening
defined on a group level as it compares average reduces breast cancer mortality for women aged
16—

50 and older by 1 in 1,000 (i.e., from 5 in 1,000 frequency of such events. However, if media cov-
without screening to 4 in 1,000 with screening) erage is disproportional to the actual frequency of
(Gøtzsche & Nielsen, 2011). The verbatim repre- a hazard—for instance, if dramatic, low-
sentation would encode the exact wording and probability events are covered more often—the
numerical information from the statement. In media could contribute to people’s tendency to
contrast, gist representations are more fuzzy and misrepresent risks. In an early study, Combs and
encode the information subjectively, that is, they Slovic (1979) found that the frequency of newspa-
interpret the information. For instance, a risk per reports about causes of death correlated more
reduction of 1 in 1,000 could be translated into a highly with laypeople’s estimates than with the
“small” effect. Although individuals have differ- actual frequency. In contrast, Freudenburg,
ent gist representations due to experience, knowl- Coleman, Gonzales, and Helgeland (1996) sys-
edge, and emotional reactions, they do not differ tematically analyzed whether the media primarily
in their verbatim representations. Moreover, in exaggerated risks and found only little support:
contrast to other dual-process approaches, fuzzy- The objective severity of events (i.e., number of
trace theory states that the two representations casualties) predicted media content. Hence, the
are encoded, stored, and retrieved in parallel frequency of reports in the media is not necessar-
rather than sequentially. ily biased; just as the media’s reaction to events is
Fuzzy-trace theory has been applied to a wide often quick and dramatic, so, too, is the rate at
range of tasks in judgment and decision making which such instances diminish in the media and
and risk perception to explain framing effects, get substituted by other news.
denominator neglect, and the role of emotions in Does the frequency of reporting influence risk
the encoding of risk information (Reyna & perception? In the above-mentioned study by
Brainerd, 2011). Generally, findings suggest that Hertwig and colleagues (2005) on the availability
people rely more heavily on gist than on verba- heuristic, fluency—operationalized through the
tim representation. This tendency increases with occurrence of instances in the media—did not
age and expertise. The advantage of gist repre- predict participants’ risk judgments of societal
sentations is that they are more stable and less risks compared with, for instance, availability-
prone to interference, whereas verbatim repre- by-recall. Although the media are one source
sentations are more error prone and can be easily providing information about risks and its occur-
forgotten. Errors in gist representations can rence in the environment, it is only one of many
occur (because gist reflects understanding). So a factors influencing risk perception, and effects of
woman might interpret reducing breast cancer selective and short-term intensive media cover-
with mammography as reducing cancer mortal- age may only be temporary (Sjöberg et al., 2000;
ity. However, reliance on gist is generally associ- Wahlberg & Sjöberg, 2000).
ated with lower levels of unhealthy risk taking However, not only does the frequency of
and more developmentally advanced decision information in the media matter (i.e., which
making (e.g., Reyna et al., 2011). information is provided) but also the format of
the information (i.e., how the information is pro-
vided). Media coverage is often biased as it pres-
Media and Risk Perception ents incomplete and nontransparent information.
For instance, newspaper and Internet reports
The media are commonly perceived as an impor- about the human papillomavirus vaccine lacked
tant mediator in the perception of risk (see also fundamental statistical information about its
Part II, Section 4). If the media covered hazards benefits and harms (Bodemer, Müller, Okan,
proportional to their actual occurrence, the media Garcia-Retamero, & Neumeyer-Gromen, 2012).
would represent a good proxy for the actual Studies covering participants from nine European
 —17

countries showed that a vast majority of people representations corresponding to the classic
overestimates the benefits of cancer screening, or notion of “rational” and deliberate thinking.
does not know (Gigerenzer, Mata, & Frank,
2009); those who acquired more information
about it did not know better, but even tended to Risk Perception in Older Adults
know less well. This suggests that many informa-
tion channels, including the media, often fail to Only few studies systematically investigated
provide adequate information about health risks. how older adults perceive risks. Hermand, Mullet,
and Rompteaux (1999) compared risk perception
across different age groups and included older
adults but found no support that older adults
Moderators in Risk Perception show higher risk perceptions than younger adults
across 91 hazards (only a slightly higher percep-
Besides the media, other factors moderate risk
tion of risks for surgery and radiation therapy).
perception. We briefly summarize these below.
When it comes to the understanding of risks,
older adults have difficulties in correctly inter-
preting health statistics compared with younger
Age adults. Yet adequate communication formats such
as graphical tools can overcome shortcomings
Risk Perception in Adolescents (Galesic, Garica-Retamero, & Gigerenzer, 2009).
The prevalence of risks and its perceptions is
also not constant, but changes across the life
span. Whereas some risks are more prevalent in Expertise and Risk Perception
young age (e.g., sexually transmitted diseases,
crimes, alcohol), others are more prevalent in Laypeople’s and experts’ risk perceptions are
older age (e.g., cardiovascular diseases, cancer). often discussed in a dichotomous manner: Experts
A widely held but unsupported belief is that ado- assess risk objectively, analytically, and wisely,
lescents engage in risky behavior because they whereas laypeople are more emotional and irratio-
feel invulnerable. In fact, the opposite may be nal in their risk perception (Slovic, 1999). However,
true. Quadrel, Fischhoff, and Davis (1993) found this may be a very general assumption as it may only
that adolescents are less inclined to optimism hold for some situations, but not all. Sjöberg (1998)
bias compared with adults. In addition, adoles- proposed three areas in risk perception to illustrate
cents provide higher risk estimates than adults a more fine-grained analysis of the commonalities
for various natural hazards and behavior-linked and differences between laypeople and experts:
outcomes (Millstein & Halpern-Felsher, 2002) (1) common, well-known risks (e.g., fatalities for
and overestimate their risk of dying in the near common diseases), (2) technological risks (e.g.,
future (Fischhoff, Bruine de Bruin, Parker, nuclear waste disposal), and (3) lifestyle and job
Millstein, & Halpern-Felsher, 2010). A second environment risks (e.g., domestic radon, smoking,
common belief is that adolescents are less alcohol). In line with the psychometric paradigm,
rational—however it may be defined—than laypeople may show higher risk perception than
adults. Yet studies applying fuzzy-trace theory on experts for technology risks. However, the pattern
risk perception obtained the opposite finding: might be reversed for lifestyle and job environment
Gist-based representation increased from risks. Furthermore, the two groups may, on average,
childhood through adolescence to adulthood have very similar perceptions in the case of common
(Reyna & Farley, 2006). In other words, adoles- and well-known risks, where both, laypeople and
cents base their decisions more often on verbatim experts, assess the frequency relatively precisely.
18—

Risk Perception and Values (i.e., their perception of risk is lower), as they
value guns as part of their social roles and indi-
How we perceive risks also depends on our vidual virtues. This pattern was most dominant
personal and cultural values, as proposed by the among people with hierarchical and individualis-
cultural theory of risk (Douglas & Wildavsky, tic worldviews, which in turn was most promi-
1983). In one study, Kahan, Braman, Slovic, nent in a subgroup of white males. Hence, the
Gastil, and Cohen (2009) investigated laypeople’s distribution of worldviews in line with culture–
perception of the risk of nanotechnology. At the identity protection can explain why specific sub-
time of the study, the vast majority of partici- groups fear some risks more than others.
pants had never heard of nanotechnology, and
risk perception did not depend on cultural values
initially. However, after they received balanced The Role of Numeracy
information about this technology, participants in Risk Perception
with different cultural values interpreted the
information in different ways: Only a minority Individuals differ in their ability to deal with
(23%) of those with an egalitarian–communitar- numerical and statistical information, which in
ian worldview (people who tend to take environ- turn influences the perception of risks. For
mental risks seriously) now thought that the instance, people low in numeracy give higher
benefits of nanotechnology would outweigh the estimates of actual risks and treatment effective-
risks. In contrast, a large majority (86%) of those ness than people high in numeracy (Dillard,
with a hierarchical–individualist worldview McCaul, Kelso, & Klein, 2006; Schwartz,
(people who tend to dismiss claims of environ- Woloshin, Black, & Welch, 1997). Less numerate
mental risk) now thought that the benefits of people are also more sensitive to the way risks are
nanotechnology would outweigh the risks. framed, that is, whether a risk is presented in a
gain or loss frame, or in percentage or frequency
format (Garcia-Retamero & Galesic, 2010; Peters
The White Male Effect et al., 2006). What actually underlies this effect
and whether it is primarily due to less numerate
First observed by Flynn, Slovic, and Metz people having difficulties adequately interpreting
(1994), the white male effect states that white and using scales measuring risk perception
men fear and worry about risks less than women remain open questions (Reyna, Nelson, Han, &
or minorities. Socioeconomic status and educa- Dieckmann, 2009).
tion were proposed as possible mechanisms
underlying this effect; yet even when controlling
for both, the white male effect occurs (Finucane, Reflections for Theory
Slovic, Mertz, Flynn, & Satterfield, 2000). A pos- and Research
sible explanation for the effect stems from its
interaction with values, worldviews, and culture– With the rise of new technologies in our world,
identity protection (Kahan, Braman, Gastil, and new developments in cognitive psychology,
Slovic, & Metz, 2007). First, worldviews correlate models and theories have been developed and
with demographic variables. Second, culture/ applied to the perception of risk. The social ampli-
identity-protective cognition suggests that we act fication of risk framework considers risk percep-
in a way to protect identity, that is, our world- tion as a communication process within social and
views, interests, and in-group. For instance, peo- institutional contexts. Thereby, it suggests possible
ple who hold hierarchical and individualistic factors that influence how we perceive risks and
worldviews may consider guns as less dangerous how risk perception develops across different
 —19

stations; yet it does not make predictions on a cognitive strategies may be applied—the key
cognitive level. The psychometric paradigm question is when which is used. Third, most
focused on the identification of the psychological research on risk perception has focused on
mechanisms that influence risk perception by known risks, that is, situations in which the out-
highlighting two major factors: dread risk and come is known and probabilities can be esti-
novelty risk. However, the major level of analysis mated. However, our distinction at the beginning
was on an aggregated level, again neglecting the of this chapter highlights that in many situations
actual underlying cognitive processes and indi- we do neither know the outcomes nor the prob-
vidual differences. The risk-as-feelings hypothesis, abilities. Up to now, we know only very little how
affect heuristic, and availability, in contrast, aim at people deal with such truly uncertain situations.
describing how individuals mentally perceive and
judge risks and extends previous models by speci-
fying and testing cognitive process models. Finally, Recommendations for Practice
fuzzy-trace theory adapts a dual-process approach
to risk perception by distinguishing between ver- Based on the knowledge on how the mind per-
batim and gist representations. Our understand- ceives risks, we can design environments that
ing of risk perception further requires studying facilitate and improve risk perception. For
mediating factors such as age, gender, expertise, instance, a doctor may overestimate the preva-
values, and worldviews. lence of a disease, as she samples patients with the
Despite, or because of, the progress in recent disease disproportionally to the population. After
decades, we would like to stress three major the 9/11 terrorist attacks, many people in the
approaches for future research on risk percep- United States avoided flying and switched to their
tion. First, only limited effort has been spent on cars instead, resulting in an increase of highway
integrating different theories and models. fatalities (Gaissmaier & Gigerenzer, 2012;
Whereas most models have been postulated and Gigerenzer, 2006). Understanding the underlying
tested in isolation, future research may address cognitive processes allows improving people’s per-
commonalities and differences and bring ceptions of risk and design information in such a
together different views to constitute an integra- way to reduce potential flaws in risk judgments.
tive framework to study risk perception. Thus, One illustrative example of how psychological
the focus should not only be on psychological research can help understand and shape debates
theories, but it should also apply a multidisci- about public health risk issues was provided by
plinary view on risk perception by including Arkes and Gaissmaier (2012). They investigated
concepts from anthropology, sociology, commu- the furor that followed when the U.S. Preventive
nication research, and technology research. Services Task Force recommended against using
Second, and related, is the fact that most models the prostate-specific antigen (PSA) test to screen
are perceived as general models that describe and for prostate cancer. Several factors documented
predict risk perception in different domains (i.e., by psychological research may have contributed
for different risks). However, a more ecological to the public’s condemnation of the report, for
approach might be useful: Just as humans have a instance, that an anecdote or two can have a
wide range of tools in their adaptive toolbox to more powerful effect on decision making than a
make decisions and judgments (Gigerenzer, compendium of more reliable statistical data.
Todd, & the ABC Research Group, 1999; Todd, The information given by the U.S. Preventive
Gigerenzer, & the ABC Research group, 2012), Services Task Force that “no trial has shown a
different tools may exist to assess risks in differ- decrease in overall mortality with the use of PSA-
ent situations. Depending on knowledge, time, based screening through 11 years of followup”
expertise, and the risk to be judged, different will not have the same impact as information,
20—

say, about the reader’s mail carrier’s older brother highly interdisciplinary constructs as they have
who had a positive PSA test, a biopsy, and a radi- different connotations in different domains. We
cal prostatectomy, and is now still alive. reviewed major findings and challenged the indi-
Psychological research has also developed more vidual theories and methods to illustrate their
effective means to represent statistical informa- advantages and limitations. The following chap-
tion about clinical evidence, including tabular ters illustrate the complexity and diversity of risk
and graphical formats, so that it can be easily perception and highlight theoretical approaches
understood even by laypeople. Arkes and to understand the interplay of the mind and the
Gaissmaier suggest that augmenting statistics environment and its implications for risk com-
with these representations might help commit- munication.
tees communicate more effectively with the pub-
lic about health risk issues and with the U.S.
Congress and could more generally be used to Suggested Additional Reading
educate the public and elevate the level of civic
discussion. More generally, providing “clean” Fischhoff, B. (1995). Risk perception and communication
information would be an important step toward unplugged: Twenty years of process. Risk Analysis,
a citizenship that deals with risks in an informed 15, 137–145.
way (Gigerenzer & Gray, 2011). Gigerenzer, G. (2002). Calculated risks: How to know
The other important building block for helping when numbers deceive you. New York: Simon &
Schuster.
the public understand risk is to teach the psychol-
Roeser, S., Hillerbrand, R., Sandin, P., & Peterson, M.
ogy and mathematics of risk. It is well documented
(2012). Handbook of risk theory. London, England:
that, for instance, low statistical numeracy in Springer.
health distorts perceptions of risks, impedes access
to treatments, and is associated with worse health
outcomes (Reyna et al., 2009). As Meder and
Gigerenzer (2014) put it, “Teaching statistical References
thinking should be an integral part of comprehen-
sive education, to provide children and adults with Alhakami, A. S., & Slovic, P. (1994). A psychological
study of the inverse relationship between perceived
the risk literacy needed to make better decisions in
risk and perceived benefit. Risk Analysis, 14, 1085–
a changing and uncertain world” (p. 127). A cur-
1096.
riculum that aims at improving people’s risk liter-
Arkes, H. R., & Gaissmaier, W. (2012). Psychological
acy should ideally already start early and target research and the prostate-cancer screening contro-
children when they develop their skills and habits versy. Psychological Science, 23, 547–553.
(Gigerenzer, 2014). An educated citizen knows Bodemer, N., Müller, S. M., Okan, Y., Garcia-
which questions to ask and where to get good Retamero, R., & Neumeyer-Gromen, A. (2012).
information and is more strongly protected against Do the media provide transparent health infor-
undue hopes and anxieties, including distorted mation? A cross-cultural comparison of public
perceptions of risk (Gigerenzer, Gaissmaier, Kurz- information about the HPV vaccine. Vaccine,
Milcke, Schwartz, & Woloshin, 2007). 30, 3747–3756.
Bodemer, N., Ruggeri, A., & Galesic, M. (2013). When
dread risks are more dreadful than continuous risks:
Comparing cumulative population losses over time.
Conclusions PLoS ONE, 8(6), e66544. doi:10.1371/journal.
pone.0066544
In this chapter, we provided an overview of the Boholm, A. (1998). Comparative studies of risk percep-
psychological theories and models in risk per- tion: A review of twenty years of research. Journal of
ception research. Risk and risk perception are Risk Research, 1, 135–163.
 ——21

Chambers, J. R., & Windschitl, P. D. (2004). Biases in social Gaissmaier, W., & Gigerenzer, G. (2012). 9/11, Act II: A
comparative judgments: The role of nonmotivated fine-grained analysis of regional variations in traffic
factors in above-average and comparative-optimism fatalities in the aftermath of the terrorist attacks.
effects. Psychological Bulletin, 130, 813–838. Psychological Science, 23, 1449–1454.
Combs, B., & Slovic, P. (1979). Newspaper coverage of Galesic, M., & Garcia-Retamero, R. (2012). The risks we
causes of death. Journalism & Mass Communication dread: A social circle account. PLoS ONE, 7(4),
Quarterly, 56, 837–849. e32837.
Covello, V., von Winterfeldt, D., & Slovic, P. (1987). Galesic, M., Garcia-Retamero, R., & Gigerenzer, G.
Communicating risk information to the public. In (2009). Using icon arrays to communicate medical
J. C. Davies, V. Covello, & F. Allen (Eds.), Risk com- risks: Overcoming low numeracy. Health Psychology,
munication (pp. 109–134). Washington, DC: 28, 210–216.
Conservation Foundation. Garcia-Retamero, R., & Galesic, M. (2010). How to reduce
Dillard, A. J., McCaul, K. D., Kelso, P. D., & Klein, W. M. the effect of framing on messages about health.
P. (2006). Resisting good news: Reactions to breast Journal of General Internal Medicine, 25, 1323–1329.
cancer risk communication. Health Communication, Gardner, G. T., & Gould, L. C. (1989). Public perceptions
19, 115–123. of the risks and benefits of technology 1. Risk
Douglas, M., & Wildavsky, A. B. (1983). Risk and culture: Analysis, 9, 225–242.
An essay on the selection of technical and environ- Gigerenzer, G. (2004). Dread risk, September 11, and
mental dangers. Berkeley: University of California fatal traffic accidents. Psychological Science, 15,
Press. 286–287.
Finucane, M. L., Alhakami, A., Slovic, P., & Johnson, S. M. Gigerenzer, G. (2006). Out of the frying pan into the fire:
(2000). The affect heuristic in judgments of risks Behavioral reactions to terrorist attacks. Risk
and benefits. Journal of Behavioral Decision Making, Analysis, 26, 347–351.
13, 1–17. Gigerenzer, G. (2014). Risk savvy: How to make good deci-
Finucane, M. L., Slovic, P., Mertz, C. K., Flynn, J., & sions. New York, NY: Penguin Books.
Satterfield, T. A. (2000). Gender, race, and perceived Gigerenzer, G., Fiedler, K., & Olsson, H. (2012).
risk: The “white male” effect. Health, Risk & Society, Rethinking cognitive biases as environmental con-
2(2), 159–172. doi:10.1080/713670162 sequences. In P. M. Todd, G. Gigerenzer, & the ABC
Fischhoff, B. (2009). Risk perception and communica- Research Group (Eds.), Ecological rationality:
tion. In R. Detels, R. Beaglehole, M. A. Lansang, & Intelligence in the world (pp. 80–110). New York, NY:
M. Gulliford (Eds.), Oxford textbook of public health Oxford University Press.
(5th ed., pp. 940–952). Oxford, England: Oxford Gigerenzer, G., Gaissmaier, W., Kurz-Milcke, E., Schwartz,
University Press. L. M., & Woloshin, S. (2007). Helping doctors and
Fischhoff, B., Bruine de Bruin, W., Parker, A. M., patients make sense of health statistics. Psychological
Millstein, S. G., & Halpern-Felsher, B. L. (2010). Science in the Public Interest, 8, 53–96.
Adolescents’ perceived risk of dying. Journal of Gigerenzer, G., & Gray, M. (2011). Better doctors, better
Adolescent Health: Official Publication of the Society patients, better decisions: Envisioning health care
for Adolescent Medicine, 46, 265–269. 2020. Cambridge: MIT Press.
Fischhoff, B., Slovic, P., Lichtenstein, S., Read, S., & Gigerenzer, G., Mata, J., & Frank, R. (2009). Public
Combs, B. (1978). How safe is safe enough? A psy- knowledge of benefits of breast and prostate cancer
chometric study of attitudes towards technological screening in Europe. JNCI Journal of the National
risks and benefits. Policy Sciences, 9, 127–152. Cancer Institute, 101, 1216–1220.
Flynn, J., Slovic, P., & Mertz, C. K. (1994). Gender, race, and Gigerenzer, G., Todd, P. M., & the ABC Research
perception of environmental health risks. Risk Group. (1999). Simple heuristics that make us
Analysis, 14(6), 1101–1108. doi:10.1111/j.1539–6924 smart. New York, NY: Oxford University Press.
.1994.tb00082.x Gøtzsche, P. C., & Nielsen, M. (2011). Screening for
Freudenburg, W. R., Coleman, C.-L., Gonzales, J., & breast cancer with mammography. Cochrane
Helgeland, C. (1996). Media coverage of hazard Database of Systematic Reviews, (1), CD001877.
events: Analyzing the assumptions. Risk Analysis, Hermand, D., Mullet, E., & Rompteaux, L. (1999).
16, 31–42. Societal risk perception among children, adolescents,
22—

adults, and elderly people. Journal of Adult Neumeyer-Gromen, A., Bodemer, N., Müller, S. M., &
Development, 6, 137–143. Gigerenzer, G. (2011). Ermöglichen Medienberichte
Hertwig, R., Pachur, T., & Kurzenhäuser, S. (2005). und Broschüren informierte Entscheidungen zur
Judgments of risk frequencies: Tests of possible cogni- Gebärmutterhalskrebsprävention? [Do media
tive mechanisms. Journal of Experimental Psychology: reports and public brochures facilitate informed
Learning, Memory, and Cognition, 31, 621–642. decision making about cervical cancer prevention?].
Hohenemser, C., R., Kates, W., & Slovic, P. (1985). A Bundesgesundheitsblatt—Gesundheitsforschung—
casual taxonomy. In R. W. Kates, C. Hohenemser, & Gesundheitsschutz, 54, 1197–1210.
J. X. Kasperson (Eds.), Perilous progress: Managing Pachur, T., Hertwig, R., & Steinmann, F. (2012). How do
the hazards of technology (pp. 67–89). Boulder, CO: people judge risks: Availability heuristic, affect heu-
Westview Press. ristic, or both? Journal of Experimental Psychology:
Johansson, G., & Rumar, K. (1968). Visible distance s and Applied, 18, 314–330.
safe approach speeds for night driving. Ergonomics, Peters, E. (2006). The functions of affect in the construc-
11, 275–282. tion of preferences. In S. Lichtenstein & P. Slovic
Kahan, D. M., Braman, D., Gastil, J., Slovic, P., & Mertz, (Eds.), The construction of preference (pp 454–463).
C. K. (2007). Culture and identity-protective cogni- New York, NY: Cambridge University Press.
tion: Explaining the white male effect in risk percep- Peters, E., Vastfjall, D., Slovic, P., Mertz, C. K., Mazzocco,
tion. Journal of Empirical Legal Studies, 4, 465–505. K., & Dickert, S. (2006). Numeracy and decision
Kahan, D. M., Braman, D., Slovic, P., Gastil, J., & Cohen, G. making. Psychological Science, 17, 407–413.
(2009). Cultural cognition of the risks and benefits Pidgeon, N. F., Hood, C., Jones, D., Turner, B., & Gibson,
of nanotechnology. Nature Nanotechnology, 4, R. (1992). Risk perception. In Risk analysis, percep-
87–90. tion and management: Report of a Royal Society
Kasperson, J. X., Kasperson, R. E., Pidgeon, N., & Slovic, P. study group (pp. 89–134). London, England: Royal
(2003). The social amplification of risk: Assessing Society.
fifteen years of research and theory. In N. Pidgeon, R. Politi, M. C., Han, P. K. J., & Col, N. F. (2007). Com-
E. Kasperson, & P. Slovic (Eds.), The social amplifica- municating the uncertainty of harms and benefits of
tion of risk (pp. 13–46). Cambridge, England: medical interventions. Medical Decision Making,
Cambridge University Press. 27(5), 681–695. doi:10.1177/0272989X07307270
Knight, F. H. (1921). Risk, uncertainty, and profit. Boston, Quadrel, M. J., Fischhoff, B., & Davis, W. (1993).
MA: Houghton Mifflin. Adolescent (in)vulnerability. American Psychologist,
Lichtenstein, S., Slovic, P., Fischhoff, B., Layman, M., & 48, 102–116.
Combs, B. (1978). Judged frequency of lethal events. Reyna, V. F. (2008). A theory of medical decision making
Journal of Experimental Psychology: Human and health: Fuzzy trace theory. Medical Decision
Learning and Memory, 4, 551–578. Making, 28, 850–865.
Loewenstein, G. F., Weber, E. U., Hsee, C. K., & Welch, N. Reyna, V. F., & Brainerd, C. J. (1995). Fuzzy-trace theory:
(2001). Risk as feelings. Psychological Bulletin, 127, An interim synthesis. Learning and Individual
267–286. Differences, 7, 1–75.
Meder, B., & Gigerenzer, G. (2014). Statistical thinking: Reyna, V. F., & Brainerd, C. J. (2011). Dual processes in
No one left behind. In E. J. Chernoff & B. Sriraman decision making and developmental neuroscience:
(Eds.), Probabilistic thinking: Presenting plural per- A fuzzy-trace model. Developmental Review, 31,
spectives (pp. 127–148). Dordrecht, Netherlands: 180–206.
Springer. Reyna, V. F., Estrada, S. M., DeMarinis, J. A., Myers, R. M.,
Meder, B., Le Lec, F., & Osman, M. (2013). Decision mak- Stanisz, J. M., & Mills, B. A. (2011). Neurobiological
ing in uncertain times: what can cognitive and deci- and memory models of risky decision making in
sion sciences say about or learn from economic adolescents versus young adults. Journal of
crises? Trends in Cognitive Sciences, 17, 257–260. Experimental Psychology: Learning, Memory,
Millstein, S. G., & Halpern-Felsher, B. L. (2002). and Cognition, 37(5), 1125–1142. doi:10.1037/
Judgments about risk and perceived invulnerability a0023943
in adolescents and young adults. Journal of Research Reyna, V. F., & Farley, F. (2006). Risk and rationality in
on Adolescence, 12, 399–422. adolescent decision making: Implications for theory,
 —23

practice, and public policy. Psychological Science in Sjöberg, L., Moen, B.-E., & Rundmo, T. (2004). Explaining
the Public Interest, 7, 1–44. risk perception: An evaluation of the psychometric
Reyna, V. F., Nelson, W. L., Han, P. K., & Dieckmann, N. paradigm in risk perception research. Trondheim,
F. (2009). How numeracy influences risk compre- Norway: Rotunde.
hension and medical decision making. Psychological Slovic, P. (1987). Perception of risk. Science, 236, 280–285.
Bulletin, 135, 943–973. Slovic, P. (1999). Trust, emotion, sex, politics, and science:
Schwartz, L. M., Woloshin, S., Black, W. C., & Welch, H. Surveying the risk-assessment battlefield. Risk
G. (1997). The role of numeracy in understanding Analysis, 19, 689–701.
the benefit of screening mammography. Annals of Slovic, P. (2000). The perception of risk. London, England:
Internal Medicine, 127, 966–972. Earthscan.
Shepperd, J. A., Carroll, P., Grace, J., & Terry, M. (2002). Slovic, P., Finucane, M., Peters, E., & MacGregor, D. G.
Exploring the causes of comparative optimism. (2002). The affect heuristic. In T. Gilovich, D. Griffin, &
Psychologica Belgica, 42, 65–98. D. Kahneman (Eds.), Heuristics and biases: The psy-
Sjöberg, L. (1996). A discussion of the limitations of the chology of intuitive judgment (pp. 397–420). New York,
psychometric and cultural theory approaches to risk NY: Cambridge University Press.
perception. Radiation Protection Dosimetry, 68, 219. Slovic, P., Fischhoff, B., & Lichtenstein, S. (1979).
Sjöberg, L. (1998). Risk perception: Experts and the pub- Weighing the risks. Environment, 21(5), 17–20,
lic. European Psychologist, 3, 1–12. 32–38. (Reprinted in P. Slovic (Ed.). The perception
Sjöberg, L. (2002a). The allegedly simple structure of of risk. London, England: Earthscan)
experts’ risk perception: An urban legend in risk Svenson, O. (1981). Are we all less risky and more
research. Science, Technology & Human Values, 27, skillful than our fellow drivers? Acta Psychologica,
443–459. 47, 143–148.
Sjöberg, L. (2002b). Are received risk perception models Todd, P. M., & Gigerenzer, G., & the ABC Research Group.
alive and well? Risk Analysis, 22, 665–669. (Eds.). (2012). Ecological rationality: Intelligence in the
Sjöberg, L. (2003). Risk perception is not what it world. New York, NY: Oxford University Press.
seems: The psychometric paradigm revisited. In Tversky, A., & Kahneman, D. (1974). Judgment under
K. Andersson (Ed.), VALDOR Conference 2003 uncertainty: Heuristics and biases. Science, 185,
(pp. 14–29). Stockholm, Sweden: VALDOR. 1124–1131.
Sjöberg, L., Jansson, B., Brenot, J., Frewer, L., Prades, Wahlberg, A. A. F., & Sjöberg, L. (2000). Risk perception
A., & Tönnesen, A. (2000). Radiation risk percep- and the media. Journal of Risk Research, 3, 31–50.
tion in commemoration of Chernobyl: A cross- Weinstein, N. D. (1987). Unrealistic optimism about sus-
national study in three waves (Rhizikon: Risk ceptibility to health problems: Conclusions from a
Research Report 33). Stockholm, Sweden: Center community-wide sample. Journal of Behavioral
for Risk Research. Medicine, 10, 481–500.