Dunning-Kruger prevents non-experts from understanding whether they understand the problem or not.

Oh shit. It's a metaproblem.

That's true: Dunning-Kruger is a major contributor into making non-expert to skip "Understand the problem" step.

My frustration though is about the case when junior uses my help inappropriately:
Instead of carefully checking with me if his understanding of the problem is correct, junior is starting throwing crazy suggestions that do not make sense.
So time is wasted, problem is not solved and junior still has no understanding of the problem.

About that
http://www.psmag.com/health-and-behavior/confident-idiots-92793

Hmm... can you solve P=NP? Start with understanding the problem. Apply the best solution.

If I try to solve P=NP I would exactly start with understanding the problem.
And would not just to the step of suggesting solutions until I understand the problem.
So, since I did not complete step 1 and am NOT going to waste time on steps 2, 3, and 4.

Does it make sense?

How to understand whether you understand problem or not?

> How to understand whether you understand problem or not?

1) Talk with people who complain about the problem:
- Express problem in your own words.
- Ask other people if your explanation is correct.

2) Talk with experts in that space.
- Express problem in your own words.
- Ask experts if your explanation is correct.

3) Explain the problem to yourself.
Ask yourself if that explanation is correct.

There is no guarantee that your explanation is complete, of course.
If you cannot suggest possible solutions, then your understanding of problem is not sufficient.
So go to step 1.

If solutions that you suggest do not work - then your understanding is not complete, so you are back to step 1.

Does it make sense?

Well, if we talk about software bugs, your meta-algorithm is correct. Many people indeed refuse to follow it. On the other hand, refusing to follow it may be exactly their goal: in our company a very popular method of investigating bugs is "I turned off feature X and the bug disappeared. Therefore, it must be a problem with feature X, so I will simply assign it to the team that designed feature X, and now it is their problem.".

But I digress. For more complex problems, you may run into trouble even with step 1 (do I really understand the problem), not evening mentioning step 3, as testing a solution may actually create much bigger problem than the one you were trying to solve.

Right. Write a function f, in c, that takes an int x and returns f(x), such that f(f(x)) = -x.
Easy, eh.

> do I really understand the problem

Just answer these questions:
1) How properly working system looks like?
2) How incorrectly working system looks like?
3) What are the differences between properly working system and incorrectly working system?

That does not guarantee full understanding, but it allows you to start producing meaningful suggestions and evaluating them.

> testing a solution may actually create much bigger problem

These are special cases which need to be resolved individually (use different testing techniques). In any case, issues like that do not prevent us from using "understand problem -> generate idea -> test -> solution" approach.

Nature is famous for solving problems without understanding.
It tries intensively and selects the best

Good point. Nature is good at solving problems like that.

However "brainpower" of nature is many orders of magnitude bigger than brainpower of humans.
Our thinking algorithms must be much more efficient in order to successfully compete with nature.

5) if the "best solution" was wrong, return to the step 2)

Or return all the way back to step 1 (re-understanding the problem).

Edited at 2015-11-25 07:15 am (UTC)