Human Intervention In Automated Systems
Creating automated websites is fun. Okay it’s not fun, but the dream has always been to develop sites that work with very little need for human intervention. Automating processes is a large part of the equation, but even while you program scripts and CRON jobs to do the work you realize there are areas that need decisions that the system just can’t make. Sometimes you can’t program in these deterministic moments.
For instance, a while ago I created a simple hyperlinking program to enable a developer to create multiple links for many pages in large documentation packages. The specs were only one link per paragraph, to link multiple grammatical versions of the same word and to link over the entire set of pages. Now the first thing I did was check for programs that already did this and there is a Perl module that does, but it was only a bare starter. So, I took the module from CPAN and included it in my program and started testing it. I soon realized that on occasion words can mean two different things depending on context. Like computer mouse and farmhouse mouse. Or the words: program and system. To program the script to analyze the content of the sentence seemed unrealistic, so the solution was to use a form that first allowed an option for automatic linking or human interaction. This showed the linker the paragraph and context of the word so that he could make a determination whether to link it or not. This automated system required human beings to make decisions.
Whether its emailing alerts from a website or viewing logs or examining packets from a firewall I have come to the conclusion that sometimes the best solution is a human being behind the keyboard watching the screen. I wonder if I wrote a program analyzing this thought process what it would decide?