Causation

A significant correlation coefficient can be taken as an indication of association between two variables, but it is important to realize that this does not automatically imply causation. 

This type of error is always liable to occur when our data is existing plant records winch we observe. It is less likely to occur if the data is obtained by actual experiment. Thus in an experiment if we raise the temperature Ti to a chosen value, and the running of the plant as measured by the dependent variable y (which may be yield, quality, etc.) improves, then for practical purposes it may be sufficient to say run the plant with T, at this value. However, this still may not be real evidence of the causative effect of T i upon y, for the plant may be such that in raising Ti to the new value we have at the same time affected the real causative variable Tj. Thus, for example, in a counter-current process, raising the temperature at one point will probably affect the temperature at other points. 

The considerations outlined in the last three paragraphs do not in any way invalidate the desirability of the testing for significance of any apparent relation. They merely imply that having shown a correlation to be significant, caution is necessary in assuming that this association is evidence of a causative effect of one variable upon the other. 

Where it is suspected that the relation between two variables is approximately linear, there are three statistics which between them summarise the most important properties of the available data — 

In-using a regression equation it must be always remembered that it is only valid over the range of the independent variable which occurred in the data used in calculating it. Extrapolation is most unwise, except when there is a very sound theoretical basis. It is good practice to quote the range used of the independent variable in order to discourage extrapolation. ( ) 

---

Halogenated compounds, found in high concentrations in seaweeds consumed in Japan and Hawaii, have been suspected of being carcinogenic, largely based on epidemiological extrapolation (high incidences of hepatic carcinoma). However, direct human causation has not been estabUshed (107). 

Diseased groups No extrapolations Susceptible groups Long-term, low-level effects Many covariates Minimal dose-response data Association vs. causation... 

Vernon, H. M. (1948). An investigation of the factors concerned in the causation of indnstriai accidents. In Baste Principles of Ventilating and Heating. T. Bedford, Lewis Co., p. 346,... 

The book begins with a discussion of the theories of error causation and then goes on to describe the various ways in which data can be collected, analyzed, and used to reduce the potential for error. Case studies are used to teach the methodology of error reduction in specific industry operations. Finally, the book concludes with a plan for a plant error reduction program and a discussion of how human factors principles impact on the process safety management system. 

This model of accident causation is described further in Figure 1.3. This represents the defenses against accidents as a series of shutters (engineered safety systems, safety procedures, emergency training, etc.) When the gaps in these shutters come into coincidence then the results of earlier hardware or human failures will not be recovered and the consequences will occur. Inap-... 

FIGURE 1.3 The Dynamics of Incident Causation (adapted from Reason, 1990). 

From the organizational view of accident causation presented in the previous section, it will be apparent that the traditional approach to human error, which assumes that errors are primarily the result of inadequate knowledge or motivation, is inadequate to represent the various levels of causation involved. These contrasting views of error and accident causation have major implications for the way in which human error is assessed and the preventative measures that are adopted. 

The system-induced error approach can be restated in an alternative form as an accident causation model (see Figure 1.4). This shows how error-inducing conditions in the form of inadequate PIFs interact with error tendencies to... 

In the shorter case studies, only the immediate causes of the errors are described. However, the more extended examples in the latter part of the appendix illustrate two important points about accident causation. First, the precondihons for errors are often created by incorrect policies in areas such as training, procedures, systems of work, communications, or design. These "root causes" underlie many of the direct causes of errors which are described in this section. Second, the more comprehensive examples illustrate the fact that incidents almost always involve more than one cause. These issues will... 

The traditional safety engineering approach to accident causation focuses on the individual rather than the system causes of error. Errors are primarily seen as being due to causes such as lack of motivation to behave safely, lack of discipline or lack of knowledge of what constitutes safe behavior. These are assumed to give rise to "unsafe acts." These unsafe acts, in combination with "unsafe situations" (e.g., imguarded plant, toxic substances) are seen as the major causes of accidents. 

One of the origins of this view of error and accident causation is the theory of accident proneness, which tried to show that a small number of individuals were responsible for the majority of accidents. Despite a number of studies that have shown that there is little statistical evidence for this idea (see, e.g., Shaw and Sichel, 1971) the belief remains, particularly in traditional industries, that a relatively small number of individuals accoimt for the majority of accidents. Another element in the emphasis on individual responsibility has been the legal dimension in many major accident investigations, which has often been concerned with attributing blame to individuals from the point of view of determining compensation, rather than in identifying the possible system causes of error. 

FIGURE 2.9. Sequential Model of Error Causation Chain (based on Rasmussen, 1982). 

In the previous chapter, a comprehensive description was provided, from four complementary perspectives, of the process of how human errors arise during the tasks typically carried out in the chemical process industry (CPI). In other words, the primary concern was with the process of error causation. In this chapter the emphasis will be on the why of error causation. In terms of the system-induced error model presented in Chapter 1, errors can be seen as arising from the conjunction of an error inducing environment, the intrinsic error tendencies of the human and some initiating event which triggers the error sequence from this imstable situation (see Figure 1.5, Chapter 1). This error sequence may then go on to lead to an accident if no barrier or recovery process intervenes. Chapter 2 describes in detail the characteristics of the basic human error tendencies. Chapter 3 describes factors which combine with these tendencies to create the error-likely situation. These factors are called performance-influencing factors or PIFs. 

The type of data collected on human error and the ways in which these data are used for accident prevention will vary depending upon the model of error and accident causation held by the management of an organization. This model will also influence the culture in the plant and the willingness of personnel to participate in data collection activities. In Chapters 1 and 2 a number of alternative viewpoints or models of human error were described. These models will now be briefly reviewed and their implications for the treatment of human error in the process industry will be discussed. 

The TSE model of causation that accidents are primarily due to individually controllable unsafe acts is unlikely to be modified over time. This is because very little evidence on the background and conditions which led up to an accident will be collected. The data collection strategy is therefore likely to remain static, since the data collected wiU, by definition, not contradict the underlying assumptions. 

FIGURE 6.2. Accident Causation Model (From Chapter 2). 

Because of the emphasis on modeling accident causation, data collection systems based on the system-induced error approach are likely to modify their data collection strategies over time. Thus, as evidence accumulates that the existing causal categories are inadequate to accoimt for the accidents and near misses that are reported, the data collection philosophy will be modified, and a new accident causation model developed. This, in turn, will be modified on the basis of subsequent evidence. 

Such examples illustrate the fundamental need to provide guarantees of anonymity and freedom from sanctions in any data collection system which relies on volimtary reporting. Such guarantees will not be forthcoming in organizations which hold a traditional view of accident causation. 

The data collected on site can easily be downloaded to a central data base, thus ensuring that any significant trends in error causation could be rapidly identified and remedied. 

This is accomplished using the ECFC and the Critical Human Achon Profile (CHAP), a fask analysis-based method used to identify the most critical actions necessary for the performance of the task. Change Analysis is a technique for investigating the role of change in accident causation. It will be described in Section 6.8.6. 

Evaluation of effects of management influences and policy factors on error causation... 

Management and Policy Influences on Error and Accident Causation As has been emphasized in Chapters 1,2, and 3, the system-induced error view states that it is insufficient to consider only the direct causes of errors. The underlying organizational influences also need to be taken into accoimt. However, most of the available techniques stop when an immediate cause has been identified, such as less than adequate procedures or poor equipment design. The questions of why the procedures were poor, or why the equipment was badly designed, are rarely addressed at the level of policy. Kletz (1994a)... 

Vernon, H. M. (1918). "An Investigation of the Factors Concerned in the Causation of Industrial Accidents." Memo Rl, Health of Munitions Workers Committee London. 

---