Sunday, March 21, 2010

Experts respected for organisational knowledge more than technical knowledge

This was my first experience as a "knowledge engineer" trying to build an "expert system" for a chemical processing plant in the mid 1980s. That experience still shapes a lot of my thinking. Firstly the context was the archetype expert system one, the expert operator (and long term foreman), 35 years + experience, retiring in 12 months ... his name was Maurie. He had the total respect of the rest of the operating crew (who I might add only averaged 20 years on the job).

After doing my knowledge engineering thing and extracting a few hundred "expert" rules, I began testing them with the "junior" operators. Anything special or insightful? No not really... the common answer was yeah you could do it that way. Would you change your action if this was recommended? Maybe ... not sure if it matters. Even Maurie was a bit ambivalent and supported them in saying yeah that could work too. This was pre- TQM days and shortly afterward the standard operating procedure (SOP) was born, so there was a lot more support for standardisation ... not so much from what might work or what might not, but a view that if we standardised actions we would at least have a measurement environment that operating performance drifts could be more easily identified.

When we implemented the system I would have to honestly say that the value the operators gained was not so much in the "insightful" recommendations the system made, but the "evidence" in terms of signals tracked and displayed to justify the recommendations that were most valued.

I continually experienced this in my Expert Systems days. A case based reasoning system for a consumer call centre was of most use to novices. More experienced staff would want to make their own decisions but appreciated the support information. Expert Systems in my experience worked best in the "complicated" domain (viz Cynefin)...where the effort of logically breaking down a decision process was both viable and valued.

As for Maurie ... why was he so respected as THE expert when the knowledge base we built from his so-called tacit knowledge was not seen as anything special? Well I learnt that respect and expertise can be different things. Perhaps Maurie's technical expertise was not necessarily superior any more to the 20 year "juniors". His people and organisational skills in working with the other operators was superior ... hence the respect that he was given. As one operator quipped ... Maurie knows where everything is .... you want a shovel or a broom....Maurie knows where it is!

I've recently interviewed some chief engineers that will retire soon. I found the same thing...its not their technical "tacit" knowledge that is valued as much as their "organisational" knowledge...especially the "how do you get stuff dome around here" tacit knowledge.

1 comment:

  1. Found this blog today.

    Interesting to see somebody tackling some of these issues. I don't yet grok what you're doing here but interesting.

    Some observations:
    It's hard to find real expertise (sometimes it would be better to ditch the project than to carry on);
    Digging out the expertise refines it especially if the KE has domain insights (i.e. expertise sometimes gets "corrected");
    After something is put down the next expert finds it easier;
    Mixing science with expertise can be very powerful;
    Boolean/Logic/Rule-Based representations are a very bad fit in some technical areas;
    If it's really good, it might be best to hive it off as management of an operational concern likely is unsuited for managing a good product;