Hiring, Algorithms, and Criteria

A lot of businesses are pretty desperate to fill open positions.  Pay rates are being increased, and hiring bonuses are being offered.  Yet at the same time, the software that frequently intermediates between candidate and company is often structured and configured in a way that does nobody any favors. (except, perhaps, the vendors of such software)

Over-specification of job requirements is a  frequent problem…some hospitals, for example,  have scanned the resumes of nurses for ‘computer programming’ skills when what they really wanted was someone with data entry experience.  Years ago, the WSJ ran an article on overspecification, featuring the quote “Companies are looking for a five-pound butterfly. Not finding them doesn’t mean there is a shortage of butterflies.”  As an example, it mentioned a company that makes automobile bumper parts and was looking for a factory shift supervisor. They eliminated all candidates who didn’t have a BS degree, even though many had relevant experience, and also insisted on experience with the specific manufacturing software that was in use at the plant. It took six months to fill this job (during which time the position was being filled by someone who wouldn’t ultimately be chosen for it.) Another company, Wabtec, which makes components for railcars and buses, insisted on knowledge of a specific version of the computer-aided design system it uses, even though the differences between that version and the earlier version were not all that great.

It’s a basic reality of life that you can’t optimize everything at once. So, if you insist on a perfect fit for certain things, you are probably getting less of some other attributes–and these may be ones that matter more. I’d personally rather have a salesman who has demonstrated (for example) skill at managing the customer politics in a large and complex sale than one who has specific experience with the Snarkolator CRM system. It’s a lot easier to train for the second than for the first.

Similarly, if a newly-hired mechanical engineer doesn’t work out, the cause will probably not be his lack of experience with the latest version of a CAD system. More likely, it will be a lack of good design intuition…or poor interpersonal skills…or an inability to integrate mechanical design with electrical and electronics aspects of the same product…or fit with the cultural style of the organization. Maybe he comes from an environment where he was closely supervised, and the new environment is more open and requires more self-starting…or vice versa. These things are not easily represented in “checklist” form, as is knowledge of a specific software package and version, but they matter a lot.

But…in addition to the overspecification problem…I think it’s a fallacy to believe that an algorithm, at the current state of the art, can reliably look at someone’s resume to decide whether or not  they have a good shot at meeting the needs of the job. (Some programs are even looking at the applicant’s photograph, which opens the door to all kinds of problems.)  It’s hard enough for a sharp HR person (some of them do exist) to understand the needs of the hiring manager well enough to screen resumes appropriately, let alone expecting an algorithm to do so.

Someone I know recently applied for a job for which she seems to be an excellent fit.  The hiring software used by the organization sucked in her resume, formatted it to meet that software’s standards, and displayed the result.  It didn’t really communicate what she wanted communicated…she had published several articles that were very relevant to the potential employer, but whoever wrote the software had apparently never thought about ‘published articles’ as a potential hiring factor, and she had to wrestle with the software for some time to get them included.

Once a resume passes initial screening, there are also some malign trends related to the interview process itself.  In an seemingly-increasing number of cases, companies are requiring candidates to be interviewed by an unreasonable number of people…as many as 8 or 10, apparently.  Still worse, there are companies which will reject a candidate if any of the interviewers disapproves of them.  This is a guaranteed way to enshrine mediocrity and groupthink.

Employers do indeed have a real problem in dealing with the absolute flood of resumes that they get for just about any position–but need to be careful not to throw the babies out with the bathwater.  Hiring decisions are absolutely critical in any organization–‘personnel is policy’–and have huge impacts on productivity, innovation, and growth.

Thoughts from a Cosmonaut

Valentina Leonidovna Ponomaryova is a former Soviet cosmonaut: with a background in applied mathematics, she was selected in 1962 as a member of the first group of women cosmonauts.  Never got to fly a mission, though–she’d been scheduled to fly on Vostok 6, following Valentina Tereshkova’s scheduled flight on Vostok 5, but “Ponomaryova did not respond with standard Soviet cliches in interviews and her feminism made the Soviet leadership uneasy” and the crew assignments were altered. She later worked in orbital mechanics.

Interesting interview with her here.  Particularly interesting, IMO, are her thoughts about the respective roles of humans vs automated systems in spaceflight.

In the United States, spacecraft technology developed on the basis of aviation, and the respect for and trust in the pilot, characteristics of aviation, naturally transferred to spacecraft technology. In the Soviet Union, spacecraft technology was based on artillery and rocketry. Rocket scientists never dealt with “a human on board”; for them, the concept of automatic control was much easier to comprehend.

and

There is no doubt that, despite a large number of extraordinary and emergency situations, the Gemini and the Apollo programs were completed successfully because in the United States from the very beginning manned spacecraft were designed with orientation toward semi-automatic control systems in which the leading and decisive role was given to astronauts. The Gemini guidance system was already semi-automatic, and the Apollo guidance system was designed in such a way that one astronaut could perform all the operations necessary for the return from any point of the lunar orbit independently from information received from the Earth.

The opposites eventually met: semi-automatic systems constituted the “golden mean” that Soviet and American cosmonautics approached from two opposite directions: the Soviets coming from the automatic systems, and the Americans, one might say, from the manual ones.

Applicable to systems of many kinds in addition to spacecraft, I think, and many American organizations seem to be departing from the ‘golden mean’ in the direction of too much dependence on automated systems which are insufficiently understood and supervised.

Pwosesis Ayiti A

No reward for resistance; no assistance, no applause.

— Neil Peart, “Lock and Key

 

For none of us lives to himself, and none of us dies to himself.

— Paul of Tarsus, Epistle to the Romans

La merde a frappé le ventilateur; my earlier post became abruptly more topical on Wednesday the 7th, when we woke to the news of the assassination of Haitian President Jovenel Moïse. This follow-up will consider the implications of developments since late June and will specifically respond to commenters on Dilèm Aksyon Kolektif nan Matisan. Most of the structure of this post will follow the Deming process-workbench model, because history is, to a great extent, a series of contingent events, and because I am a giant process nerd.

Follow along, kids, as I summon the shade of W. Edwards Deming (1900-1993) to analyze the biggest mess I’ve ever been in!

Read more

Tiananmen OSINT

“Either write something worth reading or do something worth writing.” — Benjamin Franklin

[Readers are directed to the end of this post for an explanation of my timing and motivation.

UPDATE 6/5, 11 AM CDT: videos embedded!]

I. Anniversary Reconnoiter

At around nine in the morning local time on the thirtieth anniversary of the “June Fourth Incident,” I began a reconnoiter of Tiananmen Square in central Beijing to observe security measures and, if possible, witness any attempt at commemorating the massacre. I accompanied Dr. Andrew R. Cline, professor of media, journalism, and film at Missouri State University in Springfield. We were part of group of eleven people—four students, two faculty, and five others including me—comprising a “Study Away” program from MSU which had spent the previous twelve days in China, flying into Beijing and taking high-speed trains to Xi’an and Xining, then on via the Qinghai–Tibet railway to Lhasa before flying back to Beijing. Of all days, Tuesday 4 June 2019 was designated a free day for the group: no itinerary—and no guide. The remaining nine group members, as it turned out, had other ideas about what to do that day.

Andy’s motivation was broadly journalistic, garnished with a specific interest in whether any actual Marxists would show up. I went along out of a feeling that I had something of a reputation to uphold, and quickly decided during our approach that I would evaluate the security measures and write up a more quantitative report, although I will also pass along some thoughts about the organizational behaviors involved.

Read more

“Root Causes”

The Biden administration wants to find and solve the ‘root causes’ driving the flood of refugees to the US from the south, and has assigned that task to VP Kamela Harris.  More generally, liberals and ‘progressives’ like to talk about ‘root causes’ for all kinds of things: crime, for example: instead of arresting criminals, just solve the Root Causes of crime!

Someone needs to explain to these people the concept of ask why five times, and how that concept is properly implemented.  Example:

PROBLEM: There is oil on factory floor.  Why?

Looks like it’s coming from that machine over there.

ACTION: Clean up the oil. But then ask…
WHY is there oil leaking from that machine.

The machine has a bad gasket.

ACTION: Replace the gasket. But then ask..
WHY was the gasket bad?

Check out the condition of the gaskets on some other machines.
Looks like we’ve been buying inferior gaskets.

ACTION: Change the specifications so we don’t get any more of these. But also ask..
WHY did we decide to buy the gaskets that we did?

Uhh…they were cheap? Turns out the purchasing policy for supplies like this says “always buy the low bid.”

ACTION: Change the policy to give more weight to quality as well as price. But also ask…
WHY did the head of Purchasing ever approve a policy like this in the first place?

Maybe because his *incentive program* includes a big component for year-over-year reductions in supplies cost, with no measurement for downtime impact of bad items?

ACTION: Change the incentive program.
WHY did a one-sided incentive program like this get created and approved?

And so on. (There is nothing magic about the number Five)

But importantly, you don’t wait until you run all the way up and down the chain of causation before you clean up the oil on the floor before someone slips on it and hurts himself. You don’t go through analysis of why inferior gaskets are being purchased before replacing the gaskets before the machine loses oil again and shuts down or destroys itself.

Democrat politicians often act like they don’t understand these points, even informally and intuitively. Many of them really don’t, I think…but also, many of them just don’t care; accumulation of political power for themselves and their faction is all that matters.  Among their voters/supporters, though, there may be some who can be brought to understand the fallacies of root-causes-only thinking.

And, very importantly, if you pursue the chain of causation upward to enough levels, you are likely to find causes which are either beyond your ability to influence, or for which such influence has a very long time constant.  In the manufacturing example, for instance, you may be a factory manager in a large company with very little influence on the incentive policies that drive Purchasing to acquire inferior gaskets.  That still doesn’t mean you don’t need to clean up the oil and replace the failed gaskets, anyway.  In the Biden/Harris policy case, serious thought would show that the ability of American leaders to influence the policies, economic systems, and cultures of our southern neighbors is strictly limited, and what influence we can exert is likely to have a very long time constant. That doesn’t mean we don’t need to do anything about the border crisis.