“Firstly, let us define the objective function. This is the function that we wish to maximise. It seems reasonable that we define the objective function as the delta in profits that the company makes in, say, 5 years of the employee being in the position, with all other things being equal. “
Dubya: Profit would only be one metric. By itself it would likely be a poor choice. Profit over what period? You are getting into short term vs the medium and long term outcomes e.g. a company can increase its annual profits by reducing capital expenditure, stripping down operations to what will deliver a profit in the next reportable period, but at the expense of its longer term position. UK companies are guilty of this far too often. You are trying to explain Goodhart’s law . I get it. I even agree with it to some extent. This is still an error in implementation not policy. Get the metrics right, which probably means broader or wider spectrum of metrics making gaming the system becomes less of an issue. The right metrics may even need to be dynamic and change as circumstances change.
“That is to say, the best employee is the one who will make us the most money. (Note: this is still a subjective test; maybe performance is better, as is the case for a striker in football. The point is that the objective test is future performance of some flavour.)”
Dubya: If we are going to engage in a thought experiment, we should stick to the model I presented to start with. Which is the football team. This helps to limit the chaos you want to insert into the model. The goals the striker scores or not could be profits, or they could be something else.
“Unfortunately, we do not have access to a time machine, and do not know all of the market changes between now and 5 years from now! This means that the form of the objective function is unknowable due to the laws of the universe, and we must instead guess at what it might be.”
Dubya: even if we are talking about other organisations or businesses, I am very sceptical that a competent manager does not understand the trends within their expertise over a medium-term period of circa five years. The risk of chaos wrong footing well run businesses is reasonably low, unless the business is poorly managed.
Further, it is worth noting that the objective function must necessarily depend not just upon traits and qualities that the employee has, but also the traits and qualities that our current employees have, the other new employees we hire within that time will have, etc.
Dubya: Meritocracy is an ongoing process. People who fail to meet expectations should not prevail in their roles. If that becomes apparent over the long term, take action. You make it sound like this is one way traffic. It goes both ways.
From here, we can conclude that we cannot identify the true best candidate 100% accurately each time we go through a hiring process. We might accidentally do it, but we would never know and could never prove that we were being truly meritocratic, just that we were possibly trying to approximate it.
Dubya: Why would anyone expect 100% accuracy in any process regarding the interactions of people? I think the 80/20 rule applies here.
The question now is, “What is the best method of approximating the objective function?” To me, the answer seems obvious: the hiring manager’s reason and intuition. The reason being that any attempt to codify or denote a formula for hiring is bound to:
- miss things that might come up in an interview but cannot be patched into the recruitment process; and,
- create performative responses to questioning, which makes the test useless, formulaic, and obsolete (Especially as best practice for these interviews/reviews of job applications, which gets shared and becomes increasingly well-known).
Dubya: I am not advocating for automated recruitment software. I would agree with you that this stuff leads to very mediocre recruitment. This stuff is totally dehumanising and fails. You will end up with cookie cutter employees. The CV should still be first port of call.
Relying instead on one person, or perhaps a small panel, to use their understanding of the company, market, and role to use fuzzy logic to extrapolate how well each candidate will perform will allow them to avoid the pitfalls of the modern recruitment system, however, they could never prove that they truly did that and were not swayed by nepotism.
Duyba: I am in full agreement. That is by far the best way.
My answer is that that does not matter. It is for companies to hire and discern the people who have high levels of honour, who would not abuse any power they were given. And that, again, is best done by people talking to people, not by arbitrary tests.
Duyba: Totally true.
“tl:dr What makes a person meritorious is subjective; the true merit of a person is unknowable within the confines of the universe; the best we can do is make subjective guesses at the form of the merit function; the best way to do that is through a system which is by necessity open for abuse; it is my opinion that the potential for abuse is less of an issue than the current system of hiring.”
Dubya: thanks for your time. I am pleased to see, other than in some semantics, we largely share the same outlook. This ironically was part of the point of my original post.
