You are on page 1of 14

TUM School of Management Technische Universität München

Term paper in seminar Introduction to Academic Work

How to communicate risk and probability concepts

TUM School of Management– Department of Marketing and Consumer Research

Presented By :
Mirela Alina Georgescu, 03621834
Gisela Sánchez, 03621771

Freising, January 31, 2011


Table of contents

1. Introduction 2
2. Risk communication as a multi-stage process 2
3. Effective risk communication – shaping and evaluating the message 4
3.1. Probability concepts used in risk communication 4
3.2. Evaluation of the message 7
4. Problems in risk communication 8
4.1. Problems using terms: Safety vs. Zero Risk 8
4.2. Problems in expressing probabilities 8
4.3. Small probabilities problems and graphs’ role 9
4.4. How to avoid sending “mixed messages” 10
4.5. Obstacles in risk communication 10
5. Conclusions 11
6. Reference list 12
Appendix 13

1
1. Introduction
Effective risk communication is necessary at all times and it's applicability is growing
more each day, whether we refer to communicating risks to patients, informing consumers about
certain products, informing the general public about environmental risks etc. There is a great
body of knowledge on the topic, that provides rules, recommendations and outlines for creating
successful messages and campaigns. Communicators should not dive into the process
blindfolded, but should always keep in mind some key points, such as who is his public, what
kind of risk he is communicating and how he should handle the unpredictable situations.
The purpose of our paper is to present to the reader the steps needed to be followed when
designing the risk communication process, advices about the format of the message and finally,
problems that can arise and recommendations of how to avoid them.
The first chapter of this paper presents the beginning of risk communication. In contrast
with the simplicity of those early times, Baruch Fishhoff's seven stages of the risk
communication process are presented. In the second chapter, several studies about the best
presentation format of probability information and key points in making the right choice are
discussed. The last chapter deals with problems that may arise in risk communication situations,
providing examples. Furthermore, one study about how to avoid sending “mixed messages” is
presented. Finally, we present our conclusions and recommendations for effective risk
communication that would lead to minimizing the possibility of misunderstandings.

2. Risk communication as a multi-stage process


Risk communication nowadays is a very complex and interesting process, that be defined
as the continuous exchange of information referring to risk evaluation in a certain domain,
between experts, policy makers, practitioners, experts, interest groups and the large public.
The development of the probability theory made it possible for risk communicators to
phrase their messages in various ways to get to their audience. Even though one might think that
numbers solely are enough to convince a person of the outcome of his/her actions and decisions,
over the past twenty years, risk communication researchers and practitioners have developed the
process beyond simple percentages and correct numbers.
In his paper, “Risk Perception and Communication Unplugged: Twenty Years of
Process”, Baruch Fischhoff presents a brief development process of risk communication,
organized around seven stages, as Table 1. shows.

2
Table 1. Developmental Stages of the risk communication process
Stage Description

Get the A company does not want to communicate anything to the public but in the same time it does
numbers right risk analysis and diligent technical work. This stage is mandatory in order to resolve any
conflict that may arise or to go on with the rest of the process.

Tell them the In the second developmental stage, the numbers are communicated in a form that is very close
numbers to how they were produced. This may not always have the desired outcome, because of the
recipients' lack of specific knowledge or misinterpretation, even mistrust of the numbers that
they don't understand and create a bias between the researchers and the recipients.

Explain what “When the numbers do not speak for themselves, explaining them is important. Clearly
we mean by the communicating any number is a complicated task. Therefore, one should focus on those
numbers numbers that really matter. All too often, however, communications about risk involve a gush
of issues, with little selection.” (Fischhoff, 1995). As Lynn Frewer states, it is important to
communicate uncertainty in an explicit and understandable way that is focused on the
information needs of target audiences, because “the uncertainties associated with technical
risk assessments, upon which risk management decisions are founded, will increasingly be
subject to public and stakeholder scrutiny”. This stage also makes a call for transparency.

Show them they Risk comparisons are situations in which an unfamiliar risk is contrasted with a more common
have accepted one. Risk comparisons can be somehow informative, but this is however, not always efficient,
similar risks in because people find no reason to accept any avoidable risks, unless there are some
the past compensating benefits.

Show them that The fifth stage refers to the risk-benefit trade-off, “the public’s right to compensation for
it’s a good deal risk”. A health program may seem more attractive when described in terms of the lives that it
for them will save, rather than the lives that will still be lost. On the other side, if not properly
expressed, it can lead to suspicions of manipulation, in the choice of frame or it can lead to
instability in preferences, as frames vary over time.

Treat them nice The sixth stage introduces the question of the communication skills. “Even with a perfect
message, an inappropriate delivery can exact a toll”. Also, lack of substance or inappropriate
messages can be compensated by the smoothness of the message delivery.

Make them The process of risk communication must work both ways – from sender to receiver and back.
partners Involving people could shed light on individual perceptions, independent perspectives that
might interfere in the correct understanding of the message. “Partnerships are essential to
creating the human relations needed to damp the social amplification of minor risks-as well as
to generate concern where it is warranted.”
Source : based on Baruch Fischhoff (1995)

3
For the last stage especially, the communicator should know that “if a distrusted source
provides information that appears to promote its own vested interest, the information will
influence people’s attitudes in the opposite direction to that being promoted in the first place. [..]
The receiver of the information will become more opposed to the messages promoted by the
information source than they were before receiving the information“. (Frewer, 2003, p. 393)
In summary, an effective risk communication should be transparent, based on numbers
and arguments, individualized to each public for better understanding, creating durable relations
and taking into consideration all the implications of the actions and of the sent messages.

3. Effective risk communication – shaping and evaluating the message


In the context of potentially harmful situations, risk is technically defined as “a
combination of the probability or frequency of occurrence of a defined hazard and the magnitude
of the consequences of the occurrence'' (Royal Society, 1992, cited in Yeung and Morris, 171).
Thus, risk has two components: a probability and a magnitude of the consequences. Presenting
the public with a very low probability of a risk could lead to people ignoring it and considering it
nil, thus not even taking into consideration the consequences, even if severe or hazardous.
In their study on accident probabilities and seat belt usage, Slovic et al. provide their
subjects with different information : either that the chance of experiencing at least on disabling
injury while driving without a seatbelt is 0,00001 for each trip or that the probability of that
happening over 50 years of driving is 0,33. The result showed that subjects in the second
situation were much more prone to say that they would wear seatbelts in the future (Slovic,
Fischoff, Lichtenstein, 1980, cited in Stone et al., 1994, p. 387). This result should convince the
risk communicators that obtaining the numbers and just communicating them in a raw,
impersonal form is not enough to get their message to the public. The form in which information
is presented to one person influences the outcome and perception of the message.

3.1. Probability concepts used in risk communication


Visschers et al. did a literature review of the studies concerning probability concepts used
in risk perception and risk communication separating them in seven categories: (1) frequencies,
percentages, base rates, and proportions, (2) cumulative risk information, (3) absolute and
relative risk information, (4) verbal probability information, (5) numerical versus verbal
probability information, (6) graphs, and (7) risk ladders (Visschers et. al, 2009). In the following
we will respect their category system, present some of their findings and make additions,
4
emphasizing the special case of low-probabilities.

A. Frequencies, percentages, base rates and proportions - Based on the papers reviewed,
Visschers et al. suggest that the same denominator should be used when more probabilities are
expressed in one message, “so that people who neglect the denominator can still compare the
probability information” (Visschers et. al, 2009, p. 270). Another suggestion is doing step-by-
step descriptions of probability calculations because in general, they are relatively easy to
understand and could contribute in adequate risk estimates (Visschers et. al, 2009, p. 272).
Yamagishi reported in the case of frequencies, the higher the first number of the probability
estimate, the more risky the hazard was perceived to be. In his study respondents thought that a
mortality of “1,286 out of 10,000” represented a higher risk than a mortality of “24.14 out of
100.” (Yamagishi,1997 cited inVisschers et. al, 2009, p. 270). Also, in another cited study, it was
concluded that rates (e.g. 6 out of100) are easier to understand and compare than proportions (e.g.
1 in 166), (Grimes, Snively, 1999, cited inVisschers et. al, 2009, p. 271).

For what concerns the base rates, which are the number or proportion of people that
experiences certain negative consequences in the population at hand, one study concluded that
“people who were confronted with tailored base rates of low-risk groups seemed to take this
information into account better than those presented with tailored base rates of high-risk groups
or general base rates.” (Greening et al., 2005, cited inVisschers et. al, 2009, p. 271)

B. Absolute risk reduction and relative risk reduction - Two important concepts in risk
information are the absolute risk reduction (ARR) and the relative risk reduction (RRR). “ARR
represents the difference between the percentage of victims when this preventive measure would
succeed and the percentage of victims in the current situation. RRR describes the absolute risk
reduction divided by the percentage of victims in the current situation”. Based on 11 studies
reviewed, Visschers et al. reported that “RRR increased people’s willingness to get treatment,
their willingness to advise treatment, and their willingness to pay to prevent the risk, compared to
ARR or other presentation formats.” (Visschers et al., 2009, p.272). In addition, the study by
Stone et al. concluded that in low-probability risk situations, the ARR is likely to be neglected
and considered essentially nil while the RRR is more effective, because the respondents do not
know the base rates, overestimating the risk reduction. Also, by communicating the RRR, there
was a higher increase in willingness to pay of the subjects, so attention must be paid to this two
concepts in risk communication to consumers.(Stone et al. 1994, p. 395).

5
C. Cumulative probabilities - Knauper et. al (2005) did a study on estimation of the cumulative
risk of infection with sexually transmitted diseases and came to the conclusion that people have
trouble estimating accurately, they either overestimate or they understimate the risk. Other
studies showed that probability information in a cumulative format increased perceived riskiness
compared to probability presented as a single event and that people understand better formats that
present the probability that an event will at least happen one time, than formats that present the
probability that an event will never happen (Visschers et al, 2009, p.275).

D. Verbal information - Research has shown that “there are large differences between
individuals in the numerical probabilities that were associated with the same verbal expressions” ,
that “the numerical ratings of verbal expressions differed between experts (e.g., physicians) and
the general public” and also that “the context in which a verbal expression is reported was found
to affect its interpretation” (Visschers et al,2009, p.276). Finding verbal expressions that will be
interpreted by people in the same way is difficult and they should be pretested.

E. Numerical versus verbal information – People prefer to receive information about the
probability of occurrence in the form of numerical statements, but prefer to use verbal statements
of probabilities themselves when communicating to others. That is because “numerical
information is better understood and trusted than verbal information”(Erev,Cohen, 1990 cited in
Visschers et al,2009,p.279). So, to increase homogeneity in people’s interpretation and to earn
their trust, verbal communication should accompany numerical statements.

F. Risk ladders - The target risk’s probability and its location on the ladder appears to
determine people’s risk perception. This type of display is very useful when making risk
comparisons. However, the subject of risk comparison is complex.”The simplicity and intuitive
appeal of comparisons of unrelated risks may be highly deceptive. Many factors appear to play a
role in determining whether such comparisons will be useful. Whether these kinds of
comparisons ultimately generate more light than heat will depend on the degree to which both the
context of risk communication and the content of the messages are sensitive to those factors.”
(Slovic et al., 1990, p.391). In one study on risk comparisons, Slovic et al. concluded that even if
initial risk comparisons of unrelated risks lower the mean risk judgment, in a adversarial context,
the effects of the comparisons on perceived risk would be fully offset by expert critique.

6
G.. Graphs – Probability information presented in graphs appears seems to have a higher
impact on the public than numerical information only, leading to higher probability estimates and
more risk aversion.(Visschers et al, 2009, p. 279). However, to be useful, graphs must
communicate different risk characteristics, such as: risk magnitude, relative risk, cumulative risk,
uncertainty and interactions among risk factors. (Lipkus and Hollands,1999, p. 150)
In the table 2 are presented several types of graphs, their main uses and effects.

Table 2. The visual communication of risk – possible formats


Type of table Uses Effects
Risk Ladder Describing Effectively helps people “anchor” a risk to upper- and lower-bound
and Related environmental hazards references points. Perceived risk is influenced by the location of risk
Formats e.g., radon or asbestos perhaps more than the actual numbers.

Stick and facial To aid relative risk Stick figures or facial displays may affect behavioral change despite
Figures judgments the unappealing nature of the risk information.
Line Graphs Communicating trends Results regarding improvement of risk aversion in comparison to
in data numeric formats are not available
Dots fields and Communicating Increase risk aversion
marbles different probabilities
Pie charts Conveying proportions A pie chart showing joint probabilities is more effective than two pie
charts, each showing single-risk probabilities. In case of low
probabilities, they might be ignored by the receivers.
Histograms Expressing no. of cases People readily understand and find histograms helpful. They may
per different categories induce risk aversion compared with numbers alone
Source : based on Lipkus, Hollands, The Visual Communication of Risk, 1999

In regard to graphs, Visschers et al. describe the attention priority model, which
emphasizes the effectiveness of graphs, stating that “graphs draw people’s attention to the visual
contrast in the display, for example, the contrast between the number of people affected versus
the number of unaffected people.” (Visschers et al, 2009, p. 281)

3.2. Evaluation of the message


As we have seen above, there are multiple possibilities shaping risk information and
creating risk messages, with different outcomes and reactions from the public.
Weinstein suggests some criteria for evaluating the effectiveness of these risk messages:
comprehension, acceptance, dose-response consistency (do perceptions of risk and intentions
vary according to the magnitude of the risk, such that increasing levels of risk are perceived as

7
such within a hazard?), hazard-response consistency (do people facing a hazard that is higher in
risk perceive the risk as greater, show greater readiness to take action, or both than people
exposed to a hazard of lower risk?), uniformity in people's interpretation(do people with the same
exposure level interpret and react similarly to the information?), audience evaluation, direction
of communication errors. (Weinstein, Sandman, 1993)
Even so, after closely examining the issue and choosing the format that appears to be
technically right, problems may arise. These problems will be discussed in the next chapter.

4. Problems in risk communication


Risk communicators deal with risks that involve certain degrees of threat and physical
hazards (toxic substances, unfamiliar technologies, environmental pollutants). In consequence
misunderstandings, disagreements and apathy come to scene. We will analyze below specifics
problems that risk communications arise (Katherine E. Rowan p. 367).

4.1. Problems using terms: Safety vs. Zero Risk


As the word risk, the term safe is a complex concept. Most studies agree that the word
safety involves personal value judgments, making it a very individualized concept (Jardine,
Hrudey, 1997, p.491). For example, when people hear that something is considered as “no safe”,
public may immediately catalog it as unsafe. A clear example of this is how the World Health
Organization documents nickel air quality for the European guideline “no safe level was
recommended for nickel because of its cancinogenic properties”. If people consider that “no safe
level” is unsafe, it could lead to understanding that “safe” equals zero risk. This may generate a
problem because an unintentional message, “that any infinitesimal, nonzero risk prediction is
unsafe” can be sent to people (Jardine, Hrudey, 1997, p. 491). Other definitions can be applied to
this term, as the one by Dawson, “a safe level is one that you don’t need to worry about”.

4.2. Problems in expressing probabilities


There are three types of probabilities that are suggested by Kleindorfer et al: classical:
“The probability of getting exactly two heads upon three independent flips of an air coin is 3/8.”,
frequency: “The probability of dying in a car accident on a random day, trip and road in the
Unites States is 1 out of 4,000 000”, subjective: “The probability that the United States is still a
democracy in the year 2060 is 0.7”.

8
Using the first type of probability format is considered a bit unrealistic, because “most
realistic risk situations involve more complex predictions which will not allow the simplifying
assumptions needed for the classical analysis” (Jardine, Hrudey, 1997, p. 491).
The frequency probability can be applied in “the insurance industry, to predict the
magnitude of premiums which must be collected to cover the expected losses”. This type of
probability is considered closer to realistic applications in risk assessment, and even more in
“cases where relevant frequency data is available” (Jardine, Hrudey, 1997, p. 491).
The third kind of probability, subjective, is an “expression of confidence, derived from
one’s believe in the likelihood of a specified outcome”( Jardine, Hrudey, 1997, p. 491). But these
beliefs may come from a very good and detailed analysis of trends in society. The problem with
this kind of probability is that there is not too much adequate and relevant frequency data to
develop a strict approach.
Marketers or people in charge of risk communication should also be aware that expressing
a risk without giving a consequence or time frame could also lead to misunderstanding. For
example, “one in a million risk” rather than “one in a million lifetime cancer risk”. The first
statement can be confused with an annual risk or a risk of other outcomes, when the second
statement gives a specific time frame (Jardine, Hrudey, 1997, p. 492).
There is also a potential confusion between the words “probable” and “probability”.
When the world probability is discussed, some people may hear or think in the word probable.
This important because in risk assessments, “very low probabilities are usually estimated and
discussed and using the word probability may create an unintended impression of likelihood and
reality for some of the audience” (Jardine, Hrudey, 1997, p. 492).

4.3. Small probabilities problems and graphs’ role


As we mention in second chapter, there are certain situations where some kind of graphs
help to understand better the sent message. On the other hand, small probability events are harder
or more complicated to understand to the audience (e.g. 0.0003) and in a graph they may be
ignored. In these cases it is better to present the information as frequencies, 3 out of 10000.
“Frequency information conforms more readily to people‘s intuitive assessment of probabilistic
occurrence in nature” (Lipkus, Hollands,1999, p.160). It is also recommended, due to the fact that
most of the risk events during our lives have a small probability to happen during a certain month
of year, to communicate risk in a sum risk over time. An example for this is the car accident
example, presented in the previous chapter.

9
Furthermore, some studies present that familiarity and experience with graphical displays
make people understand the risk communication better (Visschers et al, 2009). On the other hand,
there are studies that “show that technical training has no effect on the audience’s performance
using graphs” (Lipkus, Hollands, 1999 p.160). What is certainly clear, as we mentioned above, in
risk communication it is important to focus on how well the graph used suits the task or message
that one wants to communicate.

4.4. How to avoid sending “mixed messages”


The first step in the process of avoiding sending mixed messages is “recognizing words
and phrases which may create unnecessary confusion or misunderstanding” (Jardine, Hrudey,
1997, p. 496). Jardine and Hrudey suggest that reviewing the information with a critical eye is
important to avoid mixed messages. Also, presenting the information to a pilot audience to get a
feedback, may lead to interesting findings.
After the words or phrases that could lead to a mixed message have been identified, some
strategies as, i.e. substituting the word that can be misunderstood for another one or giving
examples of the intended meaning of the term that is being used, can help to avoid mixed
messages. (Jardine, Hrudey, 1997, p. 496). Furthermore, keeping the consistency of the word, by
for example using the word “significant” only in a statistical context and use the word
“substantial” when it refers to something important, are also recommended.

4.5. Obstacles in risk communication


There are four major obstacles in effective risk communication (Covello, 2001, p.2-3).
Uncertainty, complexity, and incompleteness of the data drive the outcomes of most risk
assessments to be best seen as estimates, this justifying the interpretations of the data. A second
major obstacle is distrust, deriving from poor communication skills of the spokespersons,
disagreements between experts or organizations. A third obstacle is the selective reporting by the
news media, who tend to emphasize issues that play to the same “outrage factors” that the public
uses in evaluating risks. Lastly, psychological and social factors influence perception, such as
heuristics, overconfidence and unrealistic optimism, difficulty in understanding information that
is probabilistic in nature, is unfamiliar, or is presented in unfamiliar ways, etc.

10
5. Conclusions
Risk communication thus, can be described as a complex process, during which a high
number of potential problems may be encountered. It is important to avoid misunderstandings
and sending mixed messages. The communicator should go through the message with an
objective point of view, making his choices wisely and purposively, always pre-testing the
message before sending it to the general public, because even if technically, it is very good, there
might significant differences in people's perception, and so, it will not reach its goal.
Based on the studies presented, we have some recommendations. In choosing the format
of the message, it is important that communicators keep in mind certain details. Very low
probabilities (0,0001) might be considered nil. In expressing probabilities one should always use
the same denominator. Rates are easier to understand than proportions, but RRR and ARR should
be used with caution, because people might confuse them and so, misinterpret the rates. Verbal
info should be accompanied by numerical statements, because of the differences in people's
perception of verbal statements. If possible, according to the attention priority model, graphs
should be used. Also, one should be cautious when making risk comparisons with unrelated risks,
because critique from an expert could fully offset the effects of the risk comparison. For an
effective communication, the recommendations made by Jardine and Hrudey, mentioned in the
fourth chapter, should also be respected.

11
6. Reference list

1. E. Aakko (2004): Risk communication, risk perception, and public health, Wisconsin
Medical Journal, 1(1), 25-27
2. V.T. Covello, J. Mumpower (1985): Risk Analysis and Risk Management: An Historical
Perspective , Risk Analysis, 5(2), 103-120
3. V.T. Covello, P. Sandman (2001): Risk communication: Evolution and Revolution,
Solutions to an Environment in Peril, John Hopkins University Press, pp 164-178
4. I. Erev, B.L. Cohen (1990), Verbal versus numerical probabilities: Efficiency, biases, and
the preference paradox, Organizational Behavior and Human Decision Processes, 45, 1–18.
5. B. Fischhoff (1995): Risk Perception and Communication Unplugged: Twenty Years of
Process’, Risk Analysis, 15(2), 137-145
6. L. Frewer (2004): The public and effective risk communication, Toxicology Letters,149,
391–397
7. D.A. Grimes, G.R. Snively (1999): Patients’ understanding of medical risks: Implications
for genetic counseling, Obstetrics & Gynecology, 93(6), 910–914.
8. C.G. Jardine and S.E. Hrudey (1997): Mixed messages in risk communication, Risk
Analysis, 17(4), 489–498
9. B. Knauper, R. Kornik, K. Atkinson, C. Guberman,C. Aydin (2005): Motivation influenced
the underestimation of cumulative risk. Personality and Social Psychology Bulletin, 31(11),
1511–1523.
10. I.M. Lipkus, J.G. Hollands (1999): The Visual Communication of Risk, Journal of the
National Cancer Institute Monographs, 25, 149-163
11. K. Rowan (1994), Why rules for risk communication are not enough:A problem solving
approach to risk communication, Risk Analysis, 14(3), 365-374
12. P. Slovic, N. Kraus, V. T. Covello (1990): What Should We Know About Making Risk
Comparisons?, Risk Analysis, 10(3), 389-392
13. E.R. Stone, J.F. Yates, A.M. Parker (1994): Risk communication:absolute versus relative
expressions of low probability risks, Organizational Behavior and Human Decision
Processes, 60, 387-408
14. V.H.M. Visschers, R.M. Meertens,W. F. Passchier, N. K. de Vries (2009), Probability
Information in Risk Communication: A Review of the Research Literature, Risk Analysis,
29(2), 267-286
15. N.D. Weinstein, P.M. Sandman (1993): Some criteria for evaluating risk messages, Risk
Analysis, 13, 103–14.
16. K. Yamagishi (1997) When a 12.86% mortality is more dangerous than 24.14%:
Implications for risk communication, Applied Cognitive Psychology, 11, 495–506.
17. R.M.W. Yeung and J. Morris (2001): Food safety risk, Consumer perception and purchase,
behaviour, British Food Journal, 103(3), 170-186

12
Appendix

Glossary

Absolute risk reduction - the difference between the percentage of victims when this preventive
measure would succeed and the percentage of victims in the current situation.

Base rate - generally refers to the (base) class probabilities unconditioned on featural evidence,
frequently also known as prior probabilities. For example, if it were the case that 1% of the public
are "medical professionals" and 99% of the public are not "medical professionals," then the base
rates in this case are 1% and 99%, respectively.

Cumulative probabilities – are calculated when people are exposed to the same risk factors over a
longer time period or that when they are simultaneously exposed to several risk factors (e.g., both
smoking and radon)

Frequency - the number of occurrences of a repeating event per unit of time.

Percentage - a way of expressing a number as a fraction of 100

Relative risk reduction - the absolute risk reduction divided by the percentage of victims in the
current situation

Risk ladders - a tool commonly used to present a risk in comparison to other risks. It includes
several risks that are ordered from low to high probability, with or without reporting the
numerical probabilities. The risk to be evaluated is presented in between the other risks

Risk assessment - is the determination of quantitative or qualitative value of risk related to a


concrete situation and a recognized threat (also called hazard)

Verbal statements – verbal expressions that describe probabilities in numerical format

13

You might also like