My hypothesis is that any mature, well developed system is bound to have a high percentage “Human Error”. As the system develops, and the reasons for technical failures are analysed, these are remedied. The number of technical failures will then reduce. On the human side, as the system develops, risks are better known, and procedures will be developed, bringing down the number of accidents due to "Human Error". However, humans will continue to have to make decisions based on incomplete information, while having to serve multiple, partly conflicting goals, in an imperfect designed and regulated environment. So it is much harder to bring down this number. The sum total is much less accidents, and as accidents attributes to technical failures reduce even more than accidents attributed to "Human Error" , the percentage "Human Error" will increase. A high percentage "Human Error" is an indication that the people at the sharp end are given tools that are 'fit for purpose'.

I like to just play with numbers to gain a better understanding of what the ramifications of statements are. Let's just do that with this statement: suppose that in a developing system there are 1000 incidents, 500 due to technical failures and 500 due to "Human Error". As the system matures, and the technical requirements are better known, adding a bit of steel here, fixing that connection, making that part a bit more resilient, brings down the number due to technical failure to 50. Better selection, training and procedures brings down "Human Error" to 150. So over the years safety increased dramatically. Accidents due to technical failures reduced from 500 to 50. Accidents due to "Human Error" reduced from 500 to 150. The total number of accidents reduced from 1000 to 200. In percentages the total number of incidents reduced by 80%, but "Human Error" has gone up from 50% to 75%(!) while technical failure reduced from 50% to 25%. Although I don't know of any research on the subject, I'm convinced that in the 17th century the percentage "Human Error" was much lower than presently, as many accidents happened because of technical failures, or uncontrollable circumstances.

By the way, although it can be assumed that mature, well developed systems have a high percentage of “Human Error”, this can not be turned around: a system with a high percentage "Human Error" does not necessarily have to be a mature, well developed system. There is always the possibility that a high percentage "Human Error" is due to insufficient selection, training, experience etc. However for a mature, regulated system it is unlikely that this would play a large part.
The message I want to bring across with this article is quite simple: Mature, well developed systems are bound to have a low percentage of technical failures. Consequently these systems will have a high percentage if Human Error. There is no need for a knee-jerk reaction to try to reduce this. Most likely it is a good sign and there is no need to reduce this percentage! We need other indicators to know what needs improvement and how that could be achieved.