People are still bad at gauging their own interview performance

I've never interviewed for a job I didn't get. Possibly because I aim too low. But if I were to present myself honestly as a potential employee, I would enter the interview without greeting or making eye contact, then I would sit as far away from the interviewer as possible and doodle in a notebook for 15 minutes before raising my head and explaining everything I think is wrong about the company.

What a hatchet job.

Unfortunately we don't have the raw data available; but all we can infer from these heatmaps is that yes (and exactly as one would expect) there's a bit of variance between between the two sets of ratings (in both directions). Maybe a bit more than can be explained by the sample size, or from the simple fact people (on both sides of the process) are forced to discretize their assessments (so even if both parties were to objectively agree that the "real" performance was in the 2.4-2.7 range, which it is a good chunk of the time.... inevitably you'll find them be off by +/- 1 in their quantile assignments a good chunk of the time -- which is exactly what the data seem to show).

In particular: independent of how we explain the variance (whether due to sampling effects, or bona fide DK/IS) note that the vast majority of the variance is in the +/- 1 range. And on the alleged DK side (-2 variance) the measure is apparently quite miniscule.

However, that's not the original D-K effect. What D-K specifically predicted (and were able to show with much better data and methodology -- in their very specific sample population, at least) is that, among the general population there will be a striking degree of -2 variance, specifically in the bottom quartile. Going by the oft-quoted abstract from their original paper:

Across 4 studies, the authors found that participants scoring in the bottom quartile on tests of humor, grammar, and logic grossly overestimated their test performance and ability. Although their test scores put them in the 12th percentile, they estimated themselves to be in the 62nd.

In other words: "Performers in the bottom quartile estimated themselves to be performing, on average, at mean performance in the third quartile."

So if were to look for something analogous to a D-K effect, we'd expect to find a "hot spot" (or at least the strongest amount of red ink, for that quartile) in the upper left of the first chart.

Instead we find most performers in the bottom quartile (correctly) identifying themselves as in the bottom quartile; with a smaller portion bleeding up into the second quarter; and an an almost negligible portion placing themselves in the D-K zone -- namely, the third quarter.

If anything what the data show is that (even assuming there's nothing to contend with in how the actual data were taken), while inevitably there's some variance (both ways) in the two sets of rankings, and probably some of due to distorted self-perception -- on the whole there's actually much less of a D-K effect among engineers than among the general population (or rather, "Cornell undergraduates taking psychology classes" -- which, lest we forget, were the actual subject class used in the original D-K study).

So basically the opposite of what the author of the original article is claiming to infer from this data.

Based on your comments in this thread, it sounds like you may need a new job.

A company's bad interview process might have nothing at all to do with the rest of the company.

At this point, isn't it generally accepted that all tech interviewing sucks, and companies would do better throwing darts at resumes?

If all interviewing is bad, and all companies are not bad, there's no benefit to judging a company based on that process.

Yep. They had a bad experience during the interview. When they ask themselves "how do I feel about this company?", their main data point is negative.