Hacker News Reader

Hacker News

Comment by tptacek | Hacker News Reader

Parent Full thread

tptacek

Obviously, yes.

Problems like this are endemic to zero-sum schemes like stack ranking.

View on HN

How?

Edit: The parent post was edited.

Are other employee evaluation schemes free of appearance of sexism (or actual sexism), or is it harder to see?

The edit wasn't necessary to support the first sentence. There are thousands of tech companies, large and small, and most of them monitor employee performance. Meanwhile: allegations of sexist bias in review processes are infrequent enough to be newsworthy. QED.

A quick google immediately shows pages like this one [] which frets about sexism without including the words "stack" or "ranking." It has the usual citing of questionable concepts like implicit bias, or the Bloomberg article's "unconscious bias."

What I would like to know is, suppose you need to make a system in a large organization in an attempt to lay off your 10% worst employees. It's pretty reasonable to expect sexist rankings, even made by non-sexists, and to also expect beliefs by individuals, wrongly held, that their personal low rank was the result of sexism. How do you determine whether an organization's ranking system is sexist, in which direction, and by how much?

[] https://blog.impraise.com/360-feedback/how-performance-revie...

Edit: By non-sexists, I mean, like, non-garbage people.

You asked a simple, straightforward question with an easily obtainable answer, which I provided. I'm not interested in the new tangent you're trying to start here, sorry.

Yeah but your answer was so weak, talking about lawsuits that have made the news. I have also seen, personally for example, allegations of sexism regarding employee ratings that didn't turn into a news article about a lawsuit. That means you can't use notoriety of lawsuits to precede the use of "QED."

Translation: “I don’t see sexism so it isn’t a problem.”

On a serious note, tech companies, large and small, have varying amounts of lawsuits and news coverage. I agree that stack ranking and other rating methods make illegal discrimination more likely. I do not agree with your assertion that unreported discrimination based off sex doesn’t occur.

That is not the assertion I am making. You've also shifted the goalposts: you're talking about sexism in the workplace, which I agree is rampant in our industry, and I'm talking about performance review systems that are structurally or endemically biased against women, which obviously also exist, but pretty clearly don't exist everywhere.

Sorry, I missed the distinction.

You claim the correlation between companies with allegations of sexism in employee review (covered by the news), and companies that do stack ranking are sufficiently strong to declare a causal relationship between the two - in fact, strong enough that the former is universally a consequence of the later?

If you rate employees using objective metrics, then you can avoid sexism. The problem with stack ranking is, as the article said, it forces managers to use subjective measurements to rank employees because they're forced to put the employees in a particular order (instead of saying two employees performed the same). Once you start using subjective measurements, it's almost impossible to avoid having biases creep in.

The problem is you can't rate employees by objective metrics (like, line of code count) without them getting gamed, unless you can tie them directly to company success.

But you can rate them by metrics such as are they completing all their assigned work in a reasonable amount of time, and are they producing code that's up to the quality standards of the team (which could be measured by tracking whether they're introducing new bugs).

You can do that, and it sounds like a good idea as a backstop for subjective evaluations, but if this metric becomes important to the process, then people fight over estimates and game the bug tracker.

But you can rate them by metrics such as are they completing all their assigned work in a reasonable amount of time

No you can't, unless you have a method for dividing a project up into blocks of exactly equal difficulty. The person who takes more time to do genuinely difficult things is probably closing fewer JIRAs than the one quickly cranking through easy ones, which would score more highly on that metric, but which is the more valuable to your organisation?

The metric isn't "number of completed tasks". That would be a terrible metric. The metric is "amount of valuable work completed", which is a combination of number of tasks and how much work (and value) each task is. The manager who assigns tasks should have an understanding of the expected amount of work for each task and the expected difficulty.

Except you can't even do this unless you're managing a team of contractors who never communicate with each other or the rest of the company. Otherwise how do you propose to rate the person who is often behind on their own tasks but boosts the productivity of the rest of the team?

> The problem is you can't rate employees by objective metrics (like, line of code count) without them getting gamed, unless you can tie them directly to company success.

Why would you ever want to evaluate employees by any metric not directly tied to their contribution to company success?

You wouldn't. That's why people use, or include, subjective evaluation.

Subjective measures are, inherently, “not directly tied to their contribution to company success”.

"Makes people on the team uncomfortable by being argumentative", "is condescending during code reviews", "strongly supports team members via mentorship", and "defuses technical disagreements effectively and without bad feelings" are all highly positive and negative contributions to company success, but are inherently subjective since they require evaluations of soft interpersonal skills, not technical mastery. I can't imagine that you'd want to ignore such factors (employees will quit if you let toxic people stay around unchecked), but absent extreme situations where HR complaints are filed, they are also incredibly difficult to measure other than by listening to peer opinions.

Usually from the point of view of the problematic person, everyone else is the asshole. At some point the evaluation has to boil down to a judgement call, rather than an objective measurement. This process is always fraught, for instance when the bosses are the assholes themselves or support them (ahem, Travis), but I don't love the idea of just not considering interpersonal factors when evaluating performance. Bad behavior shouldn't have to rise to the level of an HR complaints before consequences are faced, and good behavior should be rewarded.

Would it surprise you to know that many managers and business leaders don't only care about company success?

Why would you ever want to evaluate employees by any metric not directly tied to their contribution to company success?

What metric would you recommend?

I would recommend not evaluating employees if you can't be bothered to develop and validate a meaningful metric, since doing so with a bad metric is actively harmful (even aside from potential liability issues if it adversely impacts a protected class without demonstrable validity.)

This isn't, by the way, an original concept; the dangers of the inappropriate use of micro-level metrics that aren't tied to business outcomes in decision-making (and the individual employee level is about as micro as you can get) was one of the key problems W. Edwards Deming identified.

"Problems like this are endemic to zero-sum schemes like stack ranking."

Stack ranking is nothing more than a more aggressive form of regular hiring promotion and advancement.

Moreover, it's just as likely that female managers prefer females.

There's nothing inherently sexist about 'stack ranking' - rather, possible sexism is derived during performance reviews - which are commonplace.

Reply on news.ycombinator.com