Laws and Algorithms (from the Vault)

Topics: HRExaminer, by Heather Bussing

example of algorithm on HRExaminer.com article by Heather Bussing 2016-03-24

Algorithms follow the rules every time. That is what they are – a set of instructions and rules. This distinction is really important in evaluating HR software.

Laws are a set of rules. Algorithms are a set of rules. They ought to work the same? If laws can run the world, then so can algorithms.

Yeah, not so much.

Here’s how law actually works.

First, the legislature makes a new law. Theoretically, laws should be designed to be objective, in the public good, and as clear as possible about what is covered and how the law works.

There used to be revisers’ offices in all the legislatures that drafted new laws, made sure they were consistent with the existing laws, and tracked amendments and court decisions that involve that law. Now, lobbyists draft most proposed laws. So more inconsistencies between statutes arise. And that whole thing about being clear and in the public interest is getting a little muddy too.

Legislatures generally work to solve an existing problem or concern. They never work too far into the future because it’s difficult to design rules for every situation, especially those that have never happened. This is the fundamental problem in drafting HR and company policies too. Law and policies are never one size fits all.

That’s where the courts come in.

Courts apply the laws (both statutes and prior court decisions) to individual circumstances. The courts are always backwards looking. They decide cases about things that have already happened. (Sometimes, they will issue an injunction that prevents something in the future, but you have to be able to show it’s probably going to happen and really bad stuff will occur when it does.)

Courts make decisions about what is fair and fits the rules in specific circumstances, on a case by case basis. Judges also get help from people in the community. Jurors can figure out who is telling the truth and what the right result should be, often better than either the lawyers or the judge. That’s because judges and lawyers spend way too much time dealing with crazy people, drama, and people who have strong incentives to lie.

So we have a set of laws or rules, that get tested and refined in many individual cases over a long time. The cases are all based on things that have already happened. Law rarely deals with what might happen.

Algorithms work very differently.

Algorithms are a set of rules that give instructions for something to happen in the future. They are based on pure logic. They can be extremely complex and nuanced, creating instructions for many different things that could happen and what comes next if it does.  For example, the instructions might be that if the player summons a blue troll to the meadow, then the troll farts purple clouds. But if it’s a orange dragon, then its farts explode into tiny red tree frogs that hop into a basket.

When you have enough data about what people do, algorithms can also figure out what is likely to happen next. That’s what the auto fill is does on your phone, and why you get such weird options when you enter text into a search on Google. The computer has instructions to take the most often used searches that match the words and letters you typed. It’s a straight-up, literal matching of letters that may or may not have anything to do with what you are thinking or what you mean.

Algorithms also get tested and refined, and some even shift to adapt to what happens over time, because the algorithms have instructions on how and when to do that. But fundamentally, the testing is not to determine better rules or even how the rules should be applied. With algorithms, the test simply tells you whether the set of instructions does what it was designed to do.

So while algorithms and laws are both sets of rules, their purpose, design and applications are completely different.

The big difference though, is that algorithms follow the rules every time. That is what they are – a set of instructions and rules.

This distinction is really important in evaluating HR software, creating policies, and deciding whether robots will take over the world.

When you have something that you want to happen the same way every time, then algorithms are wonderful things, and software and computers are the tools you need.

When you need fairness, consideration of unique circumstances and context, and when exceptions to the rules should apply, then you want something that works more like law or policy and that is administered by humans.

Humans don’t follow the rules every time. (That’s why we have courts.)

Humans also don’t like to be told what to do. And when they are tested, they may or may not even try to respond.

Humans are messy, illogical, inconsistent, annoying, and weird. They are also brilliant, kind, creative, funny, brave, strong, silly, and wonderful.

So next time someone tells you that software can solve your problem, think about whether you need the rules followed every time under every circumstance. Or not.