Thread by @ejames_c

Totally obvious once you know it exists, but training methods for when you can extract tacit expertise (assuming you have access to experts, and the skill to do so) look VERY different from training methods without such an ability.

Example: tacit expertise is basically a pattern recognition process that generates 4 things: cues, expectancies, a prioritised and fluid list of goals, and an action script.

So, with this in mind, your training program ends up looking like a series of scenario simulations.

This is very different from the ‘pedagogical development and subskill identification’ view of teaching.

Here it’s “what series of varied scenarios may I design that allows students to gain the right set of cues, expectancies, goals, and actions that experts tacitly generate?”

In other words, you don’t need to distill everything into a framework if you have the exact set of cues, expectancies, etc that an expert has; you can just train the mental models directly, via simulation exercises.

For many years I’ve worked with better programmers, who are able to — through a combination of intuition and prototyping — pick out program structures that work. Whereas if I did them we’d have to redo things a few months down the road.

I wanted this skill for myself.

I thought that I would have to synthesise their tacit mental models into a framework.

I now see that’s mistaken. All I need to do is to extract the cues, expectancies, goals, and actions in their heads, and then design a set of simulation exercises that force me to mimic them.

This is a much lower bar than doing ‘proper’ pedagogical design or syllabus development.

I mean, looking back, this is so obviously the central thread that ties together most of Naturalistic Decision Making’s training programs: commoncog.com/blog/creating-…

(NDM is the field that specialises in techniques that can extract tacit mental models of expertise).