These notes are based on two part blog post (part 1 and part 2), which themselves are based on Lisanne Bainbridge’s paper from 1983.
In 1983, she discussed counter intuitive effects of automation in industrial processes. She put her finger on what was happening, while it was being ignored for the benefit of automation.
Oddly (or not), there are a lot of parallels that can be drawn between her paper and the current AI automation wave. Important to note, she observes and discusses only human in the loop systems. Specifically, systems where human observes the work and can interfere when it goes wrong.
Observations summarized:
Experience effectiveness – operators, which have past experience, are much more efficient and effective than the inexperienced operators.
The unlearning dilemma – abilities, if not reinforced, deteriorate. Formerly experienced operators, who managed an automated process, may become inexperienced.
The recall dilemma – ability to retrieve knowledge from long term memory depends on how frequently it’s remembered. With the task automated, experienced operators use the knowledge less, hence forget.
Practice makes perfect – with the task automated, operators don’t actually do the task which they monitor. Less or no practice, no feedback, no learning, no skill reinforcement.
The next-generation dilemma – former manual operators, monitor new automated systems and hence ride on their skills, which later generations of operators cannot be expected to have. 1
Footnotes
-
As soon as I read this one, I finally felt like someone sees this, and sees this as an issue. This is something I observed first-hand. I felt like I can see a clear difference between those who learned programming in the pre-AI era vs those who learned it with AI. The former have qualities like patience and persistence in solving problems and searching online. Being able to dissect relevant information from irrelevant. Practice makes perfect, the former, who have done a task a few times have it imprinted and understood in their brain, while the latter has a vague feeling of them understanding what is being done. Another (under formed) point is juniors are an investment a company pays, to grow a more experienced developer. A medior with multiple agents is able to cover multiple juniors. Are companies gonna grow programmers? Will growth stagnate? I’m afraid for the next generation of programmers. ↩