A literature review on training finds that only 17 traits and techniques are robustly associated with later real world performance.

Popular things that *do not* have good evidence: Active Learning, Extroversion, Desire to Learn, Technological Support, and Accountability.

One teaching techniques that stood out in the literature review was “error-based examples”:

“[An] instructional strategy that has been studied to promote transfer is the use of error-based examples, or sharing with trainees what can go wrong if they do not use the trained skills back on the job.”

Multiple studies involving airline pilots and firefighters who watched videos of mishaps or reviewed detailed case studies where mistakes led to disaster show that this technique is highly effective. So don’t just teach someone a new skill. Always teach them first how lacking that skill has led others to ruin.

Another surprisingly strong way to improve training is with “Learning Goals”:

Giving trainees real world performance goals that require them to apply your training actually works. And it works even when these goals are unilaterally imposed on trainees. Instructors can simply state expectations of what constitutes “acceptable performance” or “desired performance”.

Having goals be relevant and attainable helps, but the most important thing seems to be that the trainee believes the person assigning the goals (1) cares about the goals, and (2) will check later to see if they are achieved.