Day Three
On the third day we started with a continuation of the Reason discussion from the previous day. One of the participants commented: “Models have their own life - they’re simple, they have utility. But we make them rules, and they’re not” (which I found a very nice variation on the famous George E. Box quote “All models are wrong, but some are useful”). JB stressed that safety people have the tendency to ‘glue’ a model on whatever data we have (triggering the WYLFIWYF effect). We should try to do it the other way around: see what model the data might give us.
The question ‘What is Human Factors’ was answered by a participant with “I am a Human Factor”. As it turns out definitions differ somewhat, for example between US, UK and Europe. Also it’s a dynamic/evolving field as we saw when Smokes took us through a ‘History of Human Factors’ starting way back in the 17th century. Through early health and safety legislation, like the 1802 Act for the Preservation of the Health and Morals of Apprentices, the 1833 Factory Act (delayed by a decade over a controversy of fencing machinery) and the 1854 Merchant Shipping Act (interestingly triggered by human errors vs. overloading ships - a clear conflict of interests problem) we ended up around 1900 with the rise of worker compensation schemes (first in Germany, later in UK and USA) and the rise of insurance schemes as the main means of risk management.
From 1915 on psychology popped up as an explanation for accidents. This gave rise to the search for “psychological causes” for accidents (remember Heinrich?), including (from 1926 on) attempts to find the key to ‘accident-proneness’. One solution was the attempt to engineer out the human by increasing automation and trying to reduce the probability of errors. Also Behaviourism started in this period. A positive notion from this period is the realisation that safety can be cost-effective.
World War 2 gave new impulses, among others through cockpit design and research into vigilance (e.g. radar monitoring), fatigue and stress. 1947, Fitts & Jones can be seen as the proper start of Human Factors in the USA (a paper from that period already mentions that increasing information and complexity in cockpits was going to have contra-productive effects). From this period we also have the (1951) division of ‘man is good at/machine is good at’ division that is still somewhat relevant today. Interestingly one of the participants compared this view with Kahneman’s System 1 (machine) and System 2 (man) thinking.
The growing automation and the above mentioned ‘division’ led Lisanne Bainbridge to comment on the Ironies of Automation: When automation fails the human has to step in, but how can they?
After this we went through the three main movements in Human Factors:
- Behaviourism (ca. 1910 - WW2)
- First Cognitive Revolution (1960s-1970s)
- Second Cognitive Revolution (after TMI)
Behaviourism
This movement treated human mind as a black box that responded on inputs (stimuli) and gave reactions/responses or outputs (the ‘right’ behaviour). Characteristics of behaviourism include: selection, training, supervision, observation schemes, incentives and actions to ‘increase motivation’.
Pavlov and Taylor are names that fit here. Interestingly, behaviourism sees different approaches in the USA and Europe. The latter has often a more naturalistic approach (talking to people instead of experimenting in a laboratory).
First Cognitive Revolution
After the Second World War the first cognitive revolution happened. This tried to open the black box and tried to look into what happened in human minds: information processing and/or cognitive processes. Cognitive models that came from this movement included Miller’s Information Processing Model, Chris Wickens’ workload model (which was basically a more complicated version of the information processing model) and the notion of Situational Awareness (Endsley, 1995).
Despite the fact that several scholars have challenged the notion of Situational Awareness and the fact that Endsley has argued that the original idea was never intended for (as a design tool) what it is commonly used for now (as an easy, counterfactual and normative tool explaining accidents), it has shown to be a rather dominant model with a strong ‘common sensical’ appeal (see examples here, here and here -and many more around, just use Google).
Other studies from the movement worked on subjects like memory and fatigue. The language used in this movement was often strongly influenced by computers which were used as a model for the human mind ‘black box’. Many of the models that came out of the first cognitive revolution (alternatively labelled Cognitive Psychology, Cognitive Engineering or the Information Processing Paradigm) were mainly controlled laboratory based studies. This also applies to the work of for example Kahneman who, as coming from economics, is not seen as part of the Human Factors field per se (rather as a stream parallel to the first cognitive revolution), but who is relevant with his work about decision making.
Second cognitive revolution
This movement started more or less after the Three Mile Islands incident (but not solely, also e.g. the Dryden crash was one of the cases that had influence on the move away from ‘human error’). Human error was something that was only in the eye of the beholder and also something that only ‘works’ in hindsight. Defining papers include Rasmussen’s “Human Error and the problem of Causality of Accidents” (1989) and Hollnagel’s “The Emperor’s New Clothes” and also David Woods wrote some groundbreaking stuff.
While the first cognitive revolution was characterized by thinking that “cognition is in the mind”, the second movement understood it as “cognition in the wild”. The ‘black boxes’ closed again and instead there was a focus on relations, interactions and complexity. Humans were now seen as goal driven creatures in real dynamic environments. Characteristics of the real world include constant change, dynamics, many interactions, variability, emergent properties, side effects and effects at a distance and the need for multiple perspective in order to grasp (some of) the complexity. How can you predict outcomes in complex systems? It’s nearly impossible!
Complexity is an important theme in the second cognitive movement. It’s something that characterizes a domain and/or of problem solving approaches/resources. It’s not a thing per se, but rather a situation to be investigated. It’s typified by the fact that an aggregation of the parts will not capture all relevant aspects of the whole (which is why a reductionist approach will not be the solution to a complex problem). Often complexity is transferred to the sharp end instead of offering solutions for the whole (Woods).
Work-as-done is hard to predict because people will find new and different ways of using tools and methods. This is one reason why training for specific situations (as often seen in behaviourist approaches) is often useless. Instead it’s often wise to train approaches of how to deal with situations. Cognitive models from the Second Cognitive Wave include joint cognition, coordination, control and ecological design.
Another idea that emerged was that one shouldn’t only look at what goes wrong, but very much at what goes right - the work-as-done. This obviously led to the Resilience movement which is all about success and sustained functioning. System performance is therefore often a measure.
Like Kahneman & co (who also mainly saw biases as a source of error) can be seen as a parallel track to the First Cognitive Wave, we find also a parallel track to the Second Cognitive Wave. This includes Herbert Simon with his bounded rationality, Gary Klein with naturalistic decision making and people like Gerd Gigerenzer. These scholars also tend to have more attention for success than for failure.
A great contribution from the group was when Paul mentioned an example of how a particular software firm treats problems where people have to be rung out of their beds. These are highest priority because this is seen as a highly unwanted situation. Using the canary in a coal mine metaphor: When the canary dies, you don’t get yourself a better canary, or hire more canaries (the typical business approaches to many problems, or rather symptoms) but you fix the problem!
>>> To Day 4
