[gmueckl]

I don't see how my argument is a strawman. I have learned a great deal of stuff in an autodidactic fashion myself. A big portion of my Ph.D. work is based on things I learned autodidactically. I know that I missed existing knowledge, didn't always look into the right places etc... I have been extremely lucky that none of the work I have published so far hasn't been covered in previous work that I missed.

Research, as much as trades, has its own oral history, informal institutional knowledge and so on. It's being told by the supervisor that it's not worth pursuing a certain course of action because there's a non-obvious roadblock that leads to unpublishable negative results. Or it's being told that you should try procedure X in this specific step in your experiment because $obscure_reason or "it just works better when you do it this way" and a shrug. There is tons of experience passed down from one generation to the next everyhere that is never put in proper writing. Getting access to that knowledge is just an enormous advantage.

[AmericanChopper]

Well both of us are really just putting forward our own anecdata relating to each method of learning. If we generalise that to apply criticism to larger groups of people, then I’d say that’s when we cross the line into creating a strawman. Each method certainly does have its pitfalls, but also it’s own ways of overcoming them. If we wanted to discuss the likelihood of falling into those traps (or more generally, the effectiveness of autodidactic learning vs formal education), then we’d need some data (I’m not aware if this has been studied, maybe it has?). We’d also need to put a fair amount of thought into defining the criteria for measuring that effectiveness. For example, I taught myself to change the oil in my car, if you assess my abilities as an oil changer, I think I’d be quite competent. If you assessed my abilities as a general mechanic, I’d be woefully incompetent.

> I have been extremely lucky that none of the work I have published so far hasn't been covered in previous work that I missed.

This reminds me of how Ramanujan is said to have independently remade numerous previously made discoveries in mathematics. Certainly a risk of autodidactic learning.

> There is tons of experience passed down from one generation to the next everyhere that is never put in proper writing.

Perhaps this depends on the field. For software development, huge amounts of this knowledge is available online. I set out to learn GraphQL recently, and after reading the spec, I gained most of my knowledge on the topic from a mixture of experimentation and reading online writing. But I guess that raises the question, did I teach myself these skills, or was I taught them by dozens of random stack overflow users and blog writers?