Purely theoretically speaking, a completely unman-ageable number of risk classes could be constructed out of the eight criteria. If only the two alternatives of the ‘normal’ and ‘transitional’ case are distinguished, eight variables lead to 28combinations. Such a diver-sity of cases would run counter to the purpose of clas-sification, namely to present a clear-cut matrix of risk classes. In reality, however, some of the criteria are closely coupled to each other, while other combina-tions, although theoretically conceivable, have little or no empirical grounding. Moreover, to apply the
‘guard rail corridor’ concept developed by the Coun-cil it suffices if the transitional case is reached for one criterion alone, regardless of whether the other crite-ria additionally fall into the extreme range. An allo-cation procedure was therefore applied under which individual risks were assigned to that class where they reach or overstep to a particularly striking de-gree one of the possible extreme values. As the first and third criteria each have two sub-categories, ten theoretically conceivable cases result in which the transitional range can be reached or crossed. These cases are listed in Table C 4-1.
The first case is not relevant to global risks, as a damaging event with a probability approaching 1 is either locally contained or, on the other hand, would certainly cross one of the guard rails established by the Council in previous reports (as it is certain that the consequences would occur). Major damage po-tentials with a probability that approaches 1 will
scarcely be acceptable. Such risks are very rare. It is precisely a characteristic of most anthropogenic risks that the extent of damage correlates negatively with the probability of occurrence. Usually, the larger the damage the lower the probability. Case 1 can there-fore be dismissed from further analysis unless it is as-sociated with a major delay effect, in which event case 9 applies. Case 2 can similarly be dismissed from the analysis.A probability of occurrence approaching 0 gives no cause for concern as long as the associated magnitude of potential damage is not considerable.
The special case of a small probability associated with a very large magnitude of potential damage is already covered by case 3. All other cases are requi-site to characterize global risks.
The case 6 in Table C 4-1 refers to globality of ef-fects. This case need not be further highlighted here, as the Council anyway only considers risks that have global effects or require global action (Section C 2.4-2). It may further be noted that ubiquity correlates closely with persistency (for chemicals, ubiquity is a function of persistency and mobility).
In the following, the classes of risk resulting from cases 3 to 10 are described using ideal type tables.
These ideal type tables list, for each class, the relevant criteria and their properties. Each table includes probability of occurrence, extent of damage and the confidence intervals for these two criteria. Where requisite to characterize the risk class, one or several other criteria are included in the table. The
confi-Table C 4-1 Extreme cases of the evaluation criteria selected.
Source: WBGU
Criterion Extreme property Case
Probability of occurrence P High (approaching 1) 1
Low (approaching 0) 2
Extent of damage E Approaching infinity 3
Certainty of assessment Of probability P: low 4
Of extent E: low 5
Ubiquity Global effect 6
Persistency Very long removal period 7
Irreversibility Damage not reversible 8
Delay effect Very long time lag 9
Mobilization potential High psychological and political relevance 10
dence intervals of probability of occurrence and ex-tent of damage express the certainty of assessment in each instance.
C 4.1 Damocles
The third case in Table C 4-1 is of great relevance, both theoretically and empirically. Many sources of technological risk have a very high catastrophic po-tential, although the probability that this potential manifests itself as damage is extremely low. Nuclear power plants, large chemical facilities, dams and me-teorite impacts are typical examples. This is why the Council has chosen this case as one of the classes of
risk to be studied. A prime characteristic of this class of risk is its combination of low probability with high magnitude of damage. Theoretically the damage can occur at any time, but due to the safety measures im-plemented this is scarcely to be expected.
We call this type ‘Damocles’ (Renn, 1990). In Greek myth, Damocles was once invited by his king to a banquet. However, he was obliged to take his meal under a razor-sharp sword hanging on a fine thread. For Damocles, opportunity and danger were closely linked, and the ‘Sword of Damocles’ has be-come a byword for a happy situation overshadowed by danger. The damage potential of the risk taken by Damocles was the highest possible, namely the loss of his life. On the other hand, the probability of occur-rence was extremely low, for according to the myth Box C 4-1
Terms used in the ideal type risk class tables The tables contain information in five dimensions:
1) The two classic risk factors, probability of occurrence P and extent of damage E.
2) The certainty of assessment of these two factors. High certainty of assessment means that the statement of a specific probability that a particular damaging event oc-curs (or a certain magnitude of damage materializes) or the statement of a specific magnitude of damage for a particular probability can be made with great reliability.
A low certainty of assessment means that statements of the probability of a particular event or, conversely, state-ments of magnitude for a particular probability are sub-ject to considerable variance. If certainty of assessment is high, the error bars around a value on the magnitude-probability function are very small, if certainty of as-sessment is low the bars are very large.
3) The quality of uncertainty attaching to the various crite-ria. Uncertainty prevails if there is a lack of knowledge about either the probability (indeterminacy) or the po-tential magnitude (obliviousness) of damage. However, there must be at least reason to assume that damage is to be expected. Under uncertainty, the certainty of as-sessment is by definition extremely low (approaching 0).
Uncertainty is indicated in the tables separately for each criterion.
4) The risk criteria of ubiquity, persistency, irreversibility, delay effect and mobilization potential. All of these crite-ria are treated separately in the tables.
5) The range of the sources of risk within a type of risk.
Most of the tables for specific risk potentials in Section D are constructed for a type of risk (such as floods) or for a risk in a particular social context (such as BSE in England or Germany). The individual sources of risk within a type can have different properties for the vari-ous criteria. This range of sources within a type is indi-cated by grey to black shading in the horizontal bars of the tables. The lighter the shade, the less sources of risk are situated at this point in the continuum. Dark shading indicates the median of the risks within a type.
The properties of the criteria range from ‘low’ to ‘high’. The various meanings of ‘low’ and ‘high’ are briefly explained in the following:
• Unknown
Unknown means that available knowledge does not per-mit any specific rating in the spectrum from low to high, nor a meaningful statement of confidence intervals (e.g.
lies with a probability of 90% between x and y).
• Probability of occurrence P
‘Low’ means ‘highly improbable’ (approaching 0).
‘Tends to be low’ means ‘improbable’.
‘Tends to be high’ means ‘probable’.
‘High’ means ‘highly probable’ (approaching 1).
• Extent of damage E Self-explanatory
• Certainty of assessment of P or E
‘Low’ means ‘poor’ certainty of assessment.
‘Tends to be low’ means ‘still relatively poor’ certainty of assessment.
‘Tends to be high’ means ‘relatively good’ certainty of assessment.
‘High’ means ‘good’ certainty of assessment.
• Ubiquity
‘Low’ means ‘local’.
‘Tends to be low’ means ‘regional’.
‘Tends to be high’ means ‘transboundary’.
‘High’ means ‘global’.
• Persistency
‘Low’ means ‘short-term’ (<1 year).
‘Tends to be low’ means ‘medium-term’ (1–15 years).
‘Tends to be high’ means ‘long-term’ (15–30 years).
‘High’ means ‘several generations’ (>30 years).
• Irreversibility
‘Low’ means ‘restorable’.
‘Tends to be low’ means ‘largely restorable’.
‘Tends to be high’ means ‘only partially restorable’.
‘High’ means ‘irretrievable’.
• Delay effect self-explanatory
• Mobilization potential
‘Low’ means ‘politically not relevant’.
‘Tends to be low’ means ‘tends not to be politically rele-vant’.
‘Tends to be high’ means ‘tends to be politically rele-vant’.
‘High’ means ‘politically highly relevant’.
58 C Risk: Concepts and implications
the thread did not break. Modern society feels about many large-scale technologies as Damocles felt about the sword that could have fallen on him at any time while he was eating (although the thread was evidently so stable that this event never occurred).
Accordingly, a major mobilization effect upon the population is associated with this class of risk.
The consequences of damage are generally direct, but also, in the case of contaminant emissions, may not become injurious until some future time. By con-trast, both the probabilities and magnitudes of dam-age are sufficiently well known. Of course here, as in the other classes, uncertainties and possible unpre-dictable events remain. However, compared with other risks, the possibilities of damage occurring have largely been researched by scientific methods and their causal structures are understood (Table C 4.1-1).
C 4.2 Cyclops
The fourth case in Table C 4-1 refers to a constella-tion in which there is high indeterminacy in the as-sessment of the probability of occurrence, while the maximum damage is largely known. A number of natural events such as volcanic eruptions and floods
belong in this category, as does the outbreak of pan-demics wherever there is no information on their probability of occurrence or the information is con-tradictory. There is often too little knowledge about causal parameters, or too little observation time in which to identify cyclic regularities. This class of risk also includes such events or developments in which humankind modifies, through intervention in the ecosphere, the relative frequencies with which un-predictable natural processes occur, whereas the ef-fects of these processes are largely known and their magnitude can be assessed. Changes in ocean circu-lation brought about by human-induced climate change are a typical example. Similarly, a number of chemical or biological risks, where the maximum ex-tent of damage is known but the dose-response rela-tion is still unclear or controversial, can be grouped in this class. We call this type of risk ‘Cyclops’. Ancient Greek mythology tells of mighty giants who, for all their strength, were disabled by having only one sin-gle, round eye, which was why they were called
‘round eyes’ or Cyclopes. With only one eye, only one side of reality can be perceived and perspective is lost. When viewing risk, only one side can be ascer-tained while the other remains uncertain. Here it is often the case that risks are greatly underestimated whose magnitude can be grasped but whose proba-bility of occurrence is uncertain or continuously Table C 4.2-1
Ideal type table for the Cyclops risk class. Terms are explained in Box C 4-1.
Source: WBGU
Property
Criterion Low Tends to be low Tends to be high High Unknown
Probability of occurrence P Certainty of assessment of P Extent of damage E
Certainty of assessment of E Table C 4.1-1
Ideal type table for the Damocles risk class. Terms are explained in Box C 4-1.
Source: WBGU
Property
Criterion Low Tends to be low Tends to be high High Unknown
Probability of occurrence P Certainty of assessment of P Extent of damage E
Certainty of assessment of E
changes. The greater the time lag the more likely this is to happen. The mobilization potential is low. Con-sequences can be considerable if ubiquity and persis-tency are high and the expected damage is irre-versible (Table C 4.2-1).
C 4.3 Pythia
The fifth case in Table C 4-1 refers to a risk for which the potential magnitude of damage is unknown and the probability of occurrence also can not be ascer-tained with any accuracy. To that extent, we must as-sume for risks of this type that there is great uncer-tainty with regard to possible adverse effects and also with regard to the probability of ascertainable dam-age. We call this type of risk ‘Pythia’. When in doubt, the ancient Greeks consulted one of their oracles, among which the most famous was the Delphic Ora-cle with its blind seeress Pythia. Pythia intoxicated herself with gases, in order to make predictions and give advice for the future in a state of trance. Howev-er, Pythia’s prophecies were ambiguous. They re-vealed that a major danger might be impending, but not how high its probability or severity might be, nor the distribution or type of harm.
This class includes risks associated with the possi-bility of sudden non-linear climatic changes, such as the risk of self-reinforcing global warming or of the instability of the West Antarctic ice sheet, with far more disastrous consequences than those of gradual climate change. It further includes far-reaching tech-nological innovations in certain applications of ge-netic engineering, for which neither the precise level of risk nor the probability of certain damaging events occurring can be estimated at the present point in time. Finally, the Pythia class includes chemical or bi-ological substances for which certain effects are sus-pected, but neither their magnitude nor their proba-bility can be ascertained with any accuracy. The BSE risk is the best example of this (Table C 4.3-1).
C 4.4 Pandora
A number of human interventions in the environ-ment cause wide-ranging and persistent damage.
These two criteria are exemplified by persistent or-ganic pollutants (POPs) or by biosystem changes that remain stable over long periods. In a study prepared on behalf of the Council, the two criteria have been aggregated under the heading of ‘scope’ and ex-pressed in quantitative terms (Müller-Herold, 1998).
Here particular attention needs to be given to risks characterized simultaneously by high ubiquity, persistency and irreversibility (cases 6, 7 and 8 in Table C 4-1). The presence of these criteria is also an indication that it will be scarcely possible to compen-sate for damage. There are some risks that are only persistent but by no means irreversible (for instance, with a high energy input it would be possible to trans-form radioactive waste into isotopes with short half-lives), but most of the risks grouped in this class are characterized by high levels of both persistency and irreversibility. It is not so important whether the con-sequences arise after a time lag or not. The Council has named these risks after Pandora. The ancient Greeks explained many ills of their times with the myth of ‘Pandora’s Box’, a box that, although brought down to the Earth by the beautiful Pandora, created by Zeus, only contained evils. As long as the evils remained in the box, no damage was to be feared. If, however, the box was opened, all of the evils contained in it were released to plague the Earth irreversibly, persistently and ubiquitously.
Once released, the evils pose a persistent hazard to humankind. The consequences of these risks are of-ten still unknown or there are at best presumptions as to their possible adverse effects. The magnitude of damage does not approach the infinite, but is large enough to justify countering it with risk policies. This risk type is exemplified by persistent plant protectant residues and xenobiotics. It further includes many culturally conditioned risks insofar as they are taken Ideal type table for the Pythia risk class. Terms are explained in Box C 4-1.
Source: WBGU
Property
Criterion Low Tends to be low Tends to be high High Unknown
Probability of occurrence P Certainty of assessment of P Extent of damage E
Certainty of assessment of E
60 C Risk: Concepts and implications
universally, such as putting all hopes upon a small number of cereal crop varieties, pursuing globally uniform dietary habits and lifestyles, among others (Table C 4.4-1).
C 4.5 Cassandra
Case 9 in Table C 4-1 refers to a risk characterized by a relatively lengthy delay between the triggering event and the occurrence of damage. This case is nat-urally only of interest if both the probability and magnitude of damage are relatively high. If the time interval were shorter, the regulatory authorities would certainly intervene (the risk being in the pro-hibited range). The distance in time between trigger and consequence creates the fallacious impression of safety. Above all, the belief that a remedy will be found before the actual damage occurs can be taken as an excuse for inactivity. We can find examples in both the medical and the geophysical or climate are-nas. A typical example is gradual human-induced
cli-mate change, which can trigger severe damage in vul-nerable regions such as coastal and mountain areas.
The Council has called this class of risk ‘Cassandra’, because those who warn of such risks are rarely giv-en credgiv-ence. Many types of damage occur with high probability, but in such a remote future that for the time being no one is willing to acknowledge the threat. This was the problem of Cassandra, a seeress of the Trojans, who correctly predicted the danger of a Greek victory but was not taken seriously by her countrymen. The Cassandra class of risk thus harbors a paradox: both the probability of occurrence and the damage potential are known, but because the dam-age will not occur for a long period of time, there is little concern in the present. Cassandra-type risks of-ten also display relatively high levels of ubiquity and persistency. They also entail allotting an inequitable share of the risk to future generations, thus violating the principle of sustainability (Table C 4.5-1).
Property
Criterion Low Tends to be low Tends to be high High Unknown
Probability of occurrence P Certainty of assessment of P Extent of damage E
Certainty of assessment of E Delay effect
Table C 4.5-1
Ideal type table for the Cassandra risk class. Terms are explained in Box C 4-1.
Source: WBGU
Property
Criterion Low Tends to be low Tends to be high High Unknown
Probability of occurrence P Certainty of assessment of P Extent of damage E
Certainty of assessment of E Ubiquity
Persistency Irreversibility Table C 4.4-1
Ideal type table for the Pandora risk class. Terms are explained in Box C 4-1.
Source: WBGU
C 4.6 Medusa
Case 10 in Table C 4-1 refers to the potential for pub-lic mobilization.This criterion expresses the extent of individual aversion to risk and the political protest potential fueled by this aversion, both of which are triggered among the lay public when certain risks are taken. This type of risk is only of interest if there is a particularly large gap between lay risk perceptions and expert risk analysis findings. If the two assess-ments are congruent, then political and scientific pri-orities are set in parallel. If risks considered high by experts are rather underestimated by the lay public (as is the case for leisure-time accidents or for indul-gence in substances such as tobacco or alcohol), then
Case 10 in Table C 4-1 refers to the potential for pub-lic mobilization.This criterion expresses the extent of individual aversion to risk and the political protest potential fueled by this aversion, both of which are triggered among the lay public when certain risks are taken. This type of risk is only of interest if there is a particularly large gap between lay risk perceptions and expert risk analysis findings. If the two assess-ments are congruent, then political and scientific pri-orities are set in parallel. If risks considered high by experts are rather underestimated by the lay public (as is the case for leisure-time accidents or for indul-gence in substances such as tobacco or alcohol), then