It’s refreshing to see mention of the DK effect linked to a more recent paper, one where the original authors are amending (contradicting) their original work. But I have never ever seen the DK effect mentioned in a context where it actually works as a reasonable explanation of human behavior, this post included. For example, if we take these papers at face value, one of the few things they actually do show is that there is a positive correlation between confidence and what they call “performance”; the more confident someone is in themselves, the more likely they are to be right about it. The most common (mis)conception of DK is the opposite of that. Being confident and then having reality set in is not what DK measured. Really, the only use DK has is a way for the speaker/author to position themselves as smart & authoritative by citing their awareness of it, often as if knowing of it helps you avoid it.
The original paper just did not measure anything even remotely close to startup success by founders, and thus it’s conclusions simply do not carry into this context. The paper did not measure any kind of job performance by any professionals. It didn’t measure complex tasks either. The tasks were basic and academic, e.g., a little grammar, ability to get a joke (seriously!), and the primary statistics they gathered were based on people ranking themselves against others whose performance they didn’t know, not primarily on isolated or objective self-evaluation. The sample of people was a tiny(!) set of Cornell undergrads(!) volunteering(!) for extra credit. There are just so many things wrong with assuming this work represents real human behavior, and the paper was misleading and is so completely misunderstood that I wish references to it would just stop: they’re never correct and never useful.