Safety I vs Safety II

Note: This article purpose is to give the reader interested in safety a brief summary of the book “Safety-I and Safety-II: The Past and Future of Safety Management” written by Erik Hollnagel.
I obviously read the book before writing it. However, if you think I have missed important things or misinterpreted what the author said, please let me know in the comment section below.

What is Safety I?

In a Safety I context, safety could be defined as: “freedom from unacceptable risks/the number of accidents must be as low as possible”.

Safety I can be seen as a set of constraints and rules that the workers (also called the sharp end) must follow in order to achieve safety.

When accidents happen, we find out why it happened (the root cause), and we add constraints to the system so it does not happen again.

Why is it considered outdated in certain aspects?

Safety I was designed in the early XXth century industrial context. Concern for safety was then originating from machines. They tended to be fully tractable systems, thus the cause for an accident was relatively easy to track and understand. The working environment was stable and humans had simple tasks to do.

However, with time, the tendency to reproduce the same safety methods on humans as for machines has proven flawed. The reason is today’s work environment has changed drastically and is much more complex.

Therefore, non-compliance to work as imagined by the people that write the procedures (also called the blunt end), tends to happen more often and cannot be contained effectively. For instance, an employee could find a way to do his job that is more convenient for him, that may or may not be riskier than work as imagined by the blunt end. Workers need flexibility to continually adjust to the working environment. It is interesting to note that in this regard, safety understood as a set of rules and constraints sometimes competes with productivity, as it restrains the freedom of the worker.

There are also problems relying too heavily on investigations to increase safety. The most remarkable ones being:

• Finding the causes by analyzing the effect is not always possible.
• The impact of the environment (latent conditions) is not easily quantifiable.
• A simple root cause can steer eyes away from practical issues, personal biases and political priorities that may have influenced the investigation.

What is Safety II?

In Safety II context, safety could be defined as:” the number of non-events (no accident) must be as high as possible”.

A good outcome does not mean that the process to reach that outcome was identical each time. Not taking interest in that may result in loss of information in regards to how things can go right, through continual performance adjustments. Safety II gives adequate consideration to the work as done (as opposed to work as imagined). It is used as an input to develop best practices at work and dampen the poor ones. The understanding of how things go right habitually is a prerequisite to understanding how things go wrong: Unexpected combination of performance adjustments can lead to out of proportion consequences (resonance phenomenon).

A different culture of safety

The consequence of adopting either one or the other type of Safety has cultural consequences in the workplace. For example :

Something unusual has happened at work:
Safety I > I adapt myself to the situation to maintain productivity. I don’t report this event, because I didn’t comply with the work as imagined policy.
Safety II > I adapt myself to the situation to maintain productivity, and I report it, so others can learn from my unusual experience.

Deciding a budget for safety when there were no accidents:
Safety I> Defining safety by the number of incidents makes it difficult to argue for resources to be spent on safety when there are no incidents. The safer a system is, the less you have things to measure (in the paradigm of measuring safety through what goes wrong).
Safety II> We have identified several performance adjustments that boost both productivity and safety in the workplace. We have compelling arguments to invest in safety, as the number of non-events keeps increasing.

To conclude

Clearly, Safety I has shown its limits, it is nevertheless still needed. An ideal safety definition should mix concepts from both Safety I and Safety II. It behoves us to fight culturally accepted beliefs in the workplace that impede safety, to promote a better knowledge of real work as opposed to prescribed work and to leverage that information to get better safety standards.