Re the current discussion my attention was drawn to the NASA report
’The Analysis of the Contribution of Human Factors to the In-flight Loss of Control Accidents’ *.
An initial reading heightened my inherent bias, the word ‘error’ (what is it, a cause, action, or consequence) and the use of HFACS (categorising, boxed-in thinking, the risk of relating human variability to numbers); then there were Bayesian Belief Networks … !!
However, having struggled to the end, there was acknowledgment of the assumptions and restrictions in the data base, and the limitations of the research in that this was only a model or a process of modelling.
An alleviating concluding statement triggered a re-reading the report.
“The analysis of the historical data showed that the deficiencies at the airlines’ organizational levels are often the underlying cause of flight and maintenance crew related accidents. Consequently, the authors developed a high-level airline organizational hierarchy to trace and identify the deficiency propagation”.
Back to Fig 1. which shows the error paths (vectors?) and combinations of contributors, noting that one is a direct vector by-passing the human. Also, that the data relates to the number of accidents, and not the previously discussed number of fatalities.
Idling some numbers, it is interesting that the total percentages in the HE – LOC path totals 77.8%, which is adjacent to the oft misquoted 80% human contribution in all accidents. Is this a model of the real world, or just a model of our perception of the real world?
Considering an ‘automatons’ view, then the “80%” HE would have to be blocked, yet 5.5% of that origionated from ground personnel, and not to forget the previous system related direct path (SC-LOC).
A personal experience of LOC (amongst others, intended and not so) involved the non-existent direct path ENV-LOC; the aircraft was restored to stable flight (LOC-HE-ENV). This represents the reverse, recovery path - the successes of human involvement; in a logic diagram this could involve negative HE - the argument for automation, or alternatively positive human behaviour (a negative vector) – cf concepts that error and success have the same cognitive root.
Automation would have to consider all of the reverse paths in Fig 1, yet there is little or no data which identifies the mechanism of the potentially large number of ‘hidden’ (unreported) successes.
Without identifying this mechanism how might we be sure that automation can replace the human.
Also, because this line of argument is based on a model (computation / automation) it is unlikely that we can ever provide any assurance of sucess; yet the pro automation argument is based on similar (the same model), and that any implemented solution will also use similar computation / automation technology.
Pulling hard on boot laces?
*
The direct link my not work:- try NASA Technical Reports Server (NTRS) - The Analysis of the Contribution of Human Factors to the In-Flight Loss of Control Accidents
A retired sceptical and biased pilot who spent 25 yrs developing and testing automation, then 10yrs investigating accidents to understand weaknesses in the auto – human interface, and the 40 yrs of accumulated personal ‘error’ in aviation.