Safety science and models of accident causation

Safety science and models of accident causation

This course will start with a short review of different models within safety science. Let us first examine the history of thinking about the causes of accidents and how they happen.

“Act of God”

A long time ago people used to think that accidents were caused by gods and other spirits. Their strategy was to try and prevent accidents by performing rituals and sacrifices and pleasing these gods and spirits. This kind of thinking comes to life when two events take place very close together and people attribute causal power to the first event. The attribution of this power is based on classical or operant conditioning. People can break through this false attribution by conceptualization and rationalization.

Chain of events

Some people see accidents as the consequence of a chain of events. The simplest models have a single chain, one example being Herbert Heinrich’s Domino Theory. This theory contains five ‘dominos’, each labeled with accident causes. These causes have the following order:

• Social environment and ancestry. Undesirable personality traits can be passed along through inheritance or develop because of a person’s social environment.

• Fault of person. These traits can cause character flaws, such as ignorance and recklessness. Heinrich calls these ‘secondary personal defects’. They contribute to the next domino.

• Unsafe act and/or unsafe condition. The faults of a person can lead to unsafe acts and/or conditions. These are caused by careless people, poorly designed or badly maintained equipment. Think of starting machines without precisely following the safety instructions or checking if it is in a good state.

• Accident.

• Injury.

Even though this chain works in a straight line, it can be stopped by reinforcing or taking away certain dominos. An employer can try to eliminate unsafe acts by stronger control and regulation, or provide training to change the faults of the employees.

The Domino Theory and its criticism

First of all, there usually is not only one single cause to an accident. The reality contains many different little factors (or dominos), that all contribute (and fall at the same time), and eventually cause the accident to happen. By investigating the process you can see how different causes are woven together

Secondly, it may be so that an event or condition on its own cannot cause the accident, but only in combination with other events or conditions. None of the lines individually can be seen at the primary cause of the accident.

An alternative interpretation of the Domino Theory states that combinations may or may not become causes with a degree of probability. This is a more realistic interpretation, because it doesn’t state that A and B automatically cause C, but only that they might. Unfortunately this diminishes the causal power of any event or condition and the requirement for linearity.

Latent conditions and non-linear thinking

Real accidents have several causes and arise through a number of events and conditions. An example of such a model is called Tripod. This model has defensive barriers that are in between the hazard and the accident. These barriers can have holes though, and through them the hazard can cause an accident. These holes can be caused by all kinds of factors. This model is also known as the Swiss Cheese model.

Newtonian and Einsteinian universes

With a linear and deterministic model you can look at it from above and predict what will happen. This is also called a Newtonian universe. However, there might be forces at work that slightly influence parts of the models and change the outcomes. This is called a relativistic or Einsteinian universe. This has attractions and repulsions that turn this linear model into a three-dimensional one.

Non-linear and non-deterministic models

All these models choose to ignore that common effects of higher order causes on lower order barriers. Cultural factors have serious effects on many levels of organizations and as well on immediate defenses. The models should be expanded and allow for holes to be altered by common organizational factors. These factors are probabilities, which are in turn also being influenced by other higher-order factors.

Causal effects are non-linear and non-deterministic. The only conclusion to be drawn is that the relationships between causes are probabilistic and are themselves being influenced by higher order factors. Sometimes, small variations in the starting conditions can explain the accident.

Common mode failure

Common mode failure means that failure of one defense may increase failure of another defense.

Why keep using old models?

The Fundamental Attribution Error

During the attribution error people attribute the behavior of others to dispositional factors, whereas they attribute their own behavior to external factors. People tend to attribute failures in others to personal weaknesses, but their own failures are allegedly caused by the environment.

This also works the other way around. When a result is good, we tend to give ourselves credit for that. Other people might not think that you personally created that success, and think that you are profiting from situational forces.

Managers and people in charge like to use simple linear models, because they can attribute the causes of accidents to personal failings in certain people. At the same time, they can make it seem that they would not have made that mistake. The models are thus attractive because they make it seem the accident was predictable, even though really that is hindsight.

Hindsight

Hindsight is the tendency to exaggerate in hindsight what one knew in foresight. People tend to pretend that they knew all along what was going to happen. They also pretend others should have known that, and the fact that the accident happened, is being blamed on their incompetence to deal with that.

The attribution error and hindsight together make it more appealing for managers to choose a linear model of accident causation. The manager believes that the employee knew what was happening, he could do something about it, but was apparently incapable of doing the right thing. This way, the individual gets the blame and the manager goes free. The reality is of course far from this.