Early in my working career I found that I would always get a job offer if I was at least asked in for an interview. This was true until I had a series of interviews with Lockheed-Martin. :-(

The job was referred to as a "data analysis" position, though they claimed they wanted a Ph.D. I noticed during the interview process that it appeared that I was being classified as a "modeler". Unfortunately, I didn't redirect their impression, and I later received a call from their HR department saying that they were wanting someone that was more of a "data analyst".

At this point there was nothing I could think of that an HR person would have understood in terms of trying to get myself out of this "cubby hole" in which they had filed me.

In my mind, of course I'm a "modeler". Isn't all of science about "models"? What is a theory but a "model" that has withstood all tests thus far? Isn't "data analysis" about "modeling" the data, applying the appropriate methods for which the data "model" apply?

Sure, if all they wanted was someone to apply "black box" algorithms to their data, I wasn't their man. I wouldn't even want such a job. But if what they wanted was someone that could come up with new data analyses, based upon new "models", then I was their man.*

So, yes, interviews can certainly mislead.

Unfortunately, what's a good alternative? For academia there is past academic performance (jobs do have somewhat similar data, provided the candidate has had similar positions), and standardized test scores (what standardized tests would one suggest for jobs?).

Perhaps it's not a fair comparison to lump academic admissions interviews with job interviews, even though they both have very similar reliability characteristics.

David

* The data they wanted "analyzed" was synthetic-aperture radar data. They wanted to be able to create maps and surfaces from the data. I had already thought of some novel approaches to such issues, but I wasn't going to tell them what I was thinking until I got the job. (Maybe that was a mistake, maybe not. I wouldn't be surprised if the person they eventually hired never thought of what I was thinking, since it was way outside of any standard "data analysis" methodology, and relied upon a "model" of what the data was of, and what information should be available within synthetic-aperture radar data, so long as the data wasn't "preprocessed" to the point that such information was lost.)