Numbers are not a universal solution

By Oliver Keyes

I recently went to a programming conference which contained…a multitude of talks! Many great, funny, informative talks, and a couple that were essentially sales pitches. Whatever: sales pitches are a thing, and the conference overall was great.

Except. Except except except.

One of the sales pitches was from a HR analytics company. I didn’t know this was a thing that existed, but it sort of makes sense that with data infecting absolutely everything, it would eventually hit HR. I also don’t remember the name of the company because (for reasons that will become obvious) blinding rage seared out my memory of the first ten minutes of their talk.

The talk was essentially a pitch of: we do these cool things for insertBigCompanyNameAsClient, and you should come work for us, because {cool things,insertBigCompanyNameAsClient}.

…the case study they used, though, was (and I’m paraphrasing):

One example of where we worked is Google, where they started off a childcare programme. We were brought in to evaluate how it was working. So we did a load of analysis of birth rates around Google’s employees, and the company’s expected growth in staff numbers, and we concluded that the policy as it stood wasn’t sustainable, and they shut the programme down. Pretty cool, right?

I can’t speak for anyone else at the conference, but I found this…less-than-appealling, as a pitch for why I should go work for some people. I don’t demand that my employer be lawful good, although it’s nice if they are. I do, however, have the very low expectation of don’t be chaotic evil.

This story, is evil. It’s the denial of what are, from my end, pretty damn basic benefits, to a massive number of people, in an industry where the group that most benefits from this policy is also a group the industry, including Google, has a fucking terrible record of retaining or treating as first-class citizens. Either the company doing the work is evil, or the company doing the work doesn’t mind being evil, or both.

But it also says something more basic about how things tend to be evaluated. This is: one of the costs of this oversold ‘big data’ story is the idea that by simply getting enough data in a pile, it will eventually self-assemble into a useful thing, and if we apply quantitative methods to every problem, a solution will appear.

This isn’t being done by universities, but is instead being done by private organisations, which creates the wonderful concatenation of “lots of resourcing” and “no ERB”. And when the organisations are Big Data (TM) companies, they have a strong economic incentive to go into places pitching that data is the solution to every problem you’ve ever had. Or at least, every problem you’re willing to pay them to solve. And suddenly you’re killing childcare programmes because of a predictive model of your employees’ birth rates.

Some problems cannot be solved with numbers. Some policies cannot be justified economically. There are things that are moral or ethical imperatives, that you do because it is right to do these things, because it sends a wider message about how you value people and what your priorities are. And the idea of doing that kind of work, much less going into places and actively advertising it as a reason someone should come work for you, is appalling.