[bllguo]

if you really need your data scientist to know these things then just invest some time in training. So many skills can be quickly picked up on the job, at least to a "good enough" level, yet people insist on looking for unicorns that appear to tick all the unrealistic checkboxes

[smallnamespace]

This is one reason I ask brain teasers in interviews.

I don't just care about what you know now, I need to know your willingness and ability to think on your feet when confronted with a seemingly random puzzle and actually persevere towards an answer if necessary.

Not everything will fit neatly into the box of tools that were previously learned.

[yodon]

Be careful about focusing on brain teasers during interview sessions. In the early days at Microsoft they focused heavily on brain teaser interviews, and hired what was quickly seen to be an incredibly smart bunch of devs BUT much later seen to be a highly concentrated monoculture that over indexed on one set of skills at the expense of bringing in team members with other strengths. It took a lot of work to break away from that monoculture driving interview style and get the full set of skills and capabilities the company actually needed.

Not all smart people play chess, or know every line of Dr Who dialog, or are Makers, or like brain teasers.

[smallnamespace]

Fully agree with that, brain teasers should definitely not be the dominant component of any interview process.

That said, I've run into interviewees who have simply refused to fully engage with a brain teaser -- like they just shut down and gave up, even after I provided hints. For me, that seems to be a signal that they're not likely to deal well with unexpected puzzles that appear in the normal course of engineering work.

[yodon]

It may also be they dislike the artificiality of the question style, particularly if they look like “find the twist hidden deep in the description because this isn’t a normal real world problem” sort of questions.