split this into two blogs, against take homes and other approaches

Research: Read 5 articles on types of technical interviews

Case Interview (case study)

I’ve never seen this, but it sounds like you propose an engineeer with a business scenario and ask the interviewee to study the problem, perform an analysis and discuss the problem in person

Seems good for those who are able to break down business problems as well as communicate

Coding interview

Assessing the candidate’s coding skills.

white board, pair programming, etc

tech quizzes which tell you very little about coding skills

https://www.holloway.com/g/technical-recruiting-hiring/sections/technical-interview-formats

Effort and skill enumerated for qualification / phone screen / remote conversation / remote coding

onsite whiteboard

onsite task / project

take home coding project

portfolio or prior work

Has a great set of descriptions on interview formats

phone screen / video call take home / assignment in-person meeting

Best interview format is contraversial

Read triplebyte Read google on fermi questions

Engineers find whiteboarding intimidating, stressful and not predictive of job performance

Dropbox couldn’t contain cheating on take homes

it’s not fair because people are cheating - talk about my own experience

https://www.shecancode.io/blog/its-time-to-end-whiteboard-interviews-for-software-engineers

Boot camps are an entry point to software engineering to underrepresented groups

Lots of engineers talk negatively about whiteboarding interviews

Whiteboard interviews are intimidating, especially for those with imposter syndrome

Remote can compound this factor

https://artsy.github.io/blog/2019/01/23/artsy-engineering-hiring/

artsy on interviewing by open communication

Pair programming (still kind of whiteboarding)

Talking through past work / code

Leaning into references?

2023 - 08 - 24

What is this discussion

https://www.wired.com/2015/04/hire-like-google/

First five minutes of an interview matter the most

thin slices

early judgements are useless

confirmation bias, move from assessing to finding evidence to support our initial impression

how many golf balls fit inside of a 747

makes the interviewwer feel clever

SVP people operations at Google

Unstructured job interviews are bad at prediccting how someone would perform once hired

reference checks 7% work sample 29%

general cognitive ability tests discriminate against non-white, non-male test takers, despite being a strong indicator for work performance

structured interviews are predictive even for unstructured jobs (interesting)

People who score high on conscientiousness “work to completion”

The goal of the interview process is to predict how a candidate will do once they join the team

Do what the science says, behavioral and situational structured interviews with cognitive ability, consientiousness and leadership.

Measure on a consistent rubrik

Don’t let only the boss interview

Get a disinterested assessment

https://rework.withgoogle.com/guides/hiring-use-structured-interviewing/steps/avoid-brainteasers/

Google interviewing practices

https://home.ubalt.edu/tmitch/645/session%204/Schmidt%20&%20Oh%20validity%20and%20util%20100%20yrs%20of%20research%20Wk%20PPR%202016.pdf

1998, Frank Schmidt and John Hunter

Individual performance assements are poor

general mental ability coupled with something else (integrity, structured interview, etc) general cogniative ability coupled with a structured interview

This essay is changing drastically

General Cognitive ability seems to be the best predictor by far Take home tests seem to do little

Keep in mind this depends how on the job performance is measured

The case against take-home tests

variability in hires matters a lot

Might be worth discussing take home assignments

Why not take home assignments, references, past work

Why programming exercises

intro

As a software engineer with some experience under my belt, I’ve seen a lot of interviews from both sides. I’ve seen the good and the bad from both sides of the line, and after a somewhat uncomfortable recent interview I wanted to flush some of my stronger thoughts out to disk. At first I wanted to bring to light some criticisms on take-home projects, but after exploring my thoughts around this practice I found that there thoughts outside of it which I also wanted to share. With that in mind, I’ve set out to talk about a variety of practices to screen someone’s technical abilities. This doesn’t take into account ways to make interivews approachable, or screening for non-technical traits that provide generous value.

programming exercises

Data Structures and Algorithms provide a shared foundation that can be learned in a relatively short period of time. Absent this shared foundation, the overlap of skills between the interviewer and interviewee could be nil, and no signal could be gained.

The go to screening process

Process that most organizations include is some number of programming exercises, and a system design portion

The cost for a mismatch? find citation

The take-home project, signal on stilts

Lots of overhead, both for the company and the person taking the project home

I don’t think many decent projects come back in two hours.

quote the artsy thing quote the dropbox thing (stopped doing it because of lots of cheating) tell story about my experience asapp 2 weeks, python, javascript, sql schema files, bash scripts to get it running mets 2 hours - debugging project setup, debugging csv library, reading the documentation on chartjs (which I’ve used but not in conjunction with react) ended with the create-react-app boiler plate, a text input, a couple data transformation functions, and a graph which showed nothing

The project which I worked on for asapp was much better than the one I worked on for the mets, but the engineer who would be working today is a much better engineer than the one who was working 6 years ago

What does it bias towards? People without worklife balance. Good for startups. To a degree. I feel like me working at 40 hours a week, is going to develop much better software than the me who would work 60 hours a week. herein lies the crux,

The technical quiz, an aside on a slowly fading practice on interviewing

Rote, doesn’t actually gage skill in applying People can game for this by searching “top 20 sql interview questions” etc. Requires little skill to deliver the right answer

The past-projects interview

The “day in the life” Interview, a platonic ideal I’ve never experienced

talk about circleci technically illegal if not done correctly if done correctly, requires a lot of overhead in filling out tax documents etc assumes your systems are easy enough to run / test / deploy. biases for toolchains

Probing alternatives