Doo wop, doo wop

Richard Lewis, Bill Horton, Earl Beal, Raymond Edwards, and John Wilson—the Silhouettes—were a doo wop/R&B group whose single Get A Job was a number 1 hit on the Billboard R&B singles chart and pop singles chart in 1958. Even back then it sold over one million records, and was later used in ads and movies.

Today I want to talk about hiring faculty, as we are getting near the end of the usual job hiring cycle.

From the view of the many PhD candidates who are looking for jobs, this year must seem pretty bright. Companies, company labs, and universities all seem to be hiring. We are seeing a large number of very qualified people on the market—I glad I have a job and do not have to compete with them.

At Tech we do the job search as an potential employer in the old-fashioned way. We look at applications, ask some to visit and give a formal presentation, and have them talk to our faculty and students. Then we vote on making offers, make them, and try to convince the fortunate recipients to accept.

## Dollars to Donuts

This method has been used forever it seems, and it works reasonably well. However, a question is: can we use our own methods to make the recruiting and hiring process better? In computer science theory we have many results about making decisions under uncertainty, yet when we do hiring of faculty, we use completely ad hoc methods.

This year at our first faculty meeting to discuss hiring I brought donuts from Krispy Kreme for all to enjoy. The initial presentation on a candidate by one of our faculty had a slide that quoted a letter writer:

While you are sitting around eating donuts and evaluating candidates remember that ${\dots}$

Somehow the writer of this recommendation letter ‘knew’ we would be eating donuts—I cannot decide if it was funny or scary.

## Let’s hire Alice or Carol

I wonder if we can use theory methods to rethink the hiring process. Perhaps we will always do it the old way, and always eat donuts and chat. But perhaps too there is some way to use mathematical methods. In any event, I thought I would share some simple observations about this with you.

Imagine that Alice and Carol are on the job market. Assume that both are “above the bar” and would be solid additions to our school of Computer Science. Suppose also that we have one offer we can make—and if it is declined then we cannot make another offer. So our choice is simple:

• Make an offer to Alice;
• Make an offer to Carol.

How do we chose?

A model is that each candidate has a secret probability of whether they will select our offer: ${p_{a}}$ is the probability for Alice and ${p_{c}}$ is the probability for Carol. Let’s assume that

$\displaystyle p_{a} + p_{c} = q>0.$

Here is the key issue. If we make an offer with a deterministic old-style method, we could pick a candidate that is very unlikely to come. This is what we wish to avoid.

A simple strategy is to flip an unbiased coin. If it’s heads make an offer to Alice, if it’s tails make the offer to Carol. Note, this trivial strategy yields an expected number of accepts of ${q/2}$. And it does pretty well. If ${p_{a}}$ is much larger than ${p_{c}}$, for example, we get Alice with probability within a factor of one-half. If on the other hand ${p_{a}}$ and ${p_{c}}$ are near each other we also do pretty well.

What is wrong with this strategy? Is it better than chatting and eating donuts?

## It’s More Complicated

Of course in real life the situation is much more complex:

• We may wish to make several offers.
• We may be able to make offers after we get rejections—so the problem becomes a sequential decision model.
• We may have other constraints: candidates are from various areas and so we make wish to get more people from one area than another.

And so on.

Can we make a reasonable model and find a decision strategy that still works well in a real world situation?

## Open Problems

So is it donuts forever, or can we use some decision methods in hiring?

Any way good luck to all trying to: “Get A Job.”

Yip yip yip yip yip yip yip yip
Sha na na na, sha na na na na
Sha na na na, sha na na na na
Sha na na na, sha na na na na
Sha na na na, sha na na na na
Yip yip yip yip yip yip yip yip
Mum mum mum mum mum mum
Get a job, sha na na na, sha na na na na

1. April 27, 2016 7:09 pm

While it is all very fine to focus on the game theory of the decisions, a bigger issue that I believe occurs in practice is gaming whom to invite. Is the right strategy to invite the best candidates in an area who have applied?

At a lower ranked school with a limited interview budget do you spend money to invite people whom you guess you have little chance of attracting?

I have heard of instances where schools in the middle of the country won’t bother to interview people from strong coastal places who have applied because the schools have been unable to compete for people from the coasts in the past.

This may make sense but it is a problem if the institutional perception is very far from reality. I think that this general issue has sometimes worked against good people from getting interviews: they are far down the list for the top places to hire but apparently seem too out of reach for the next level of places to interview.

2. April 28, 2016 4:14 am

I would love to receive a reject saying that I seemed better, but the coin came up tails, and anyhow they assumed that probably I wouldn’t accept the offer if they made one…

April 30, 2016 2:28 pm

Interesting. In this rare event with *only* females on the job market I think the random coin makes a perfect voting protocol.