Black Boxes

There is a great deal of hubbub in the legal tech community around a well-PRed artificial intelligence system called Ross. Aside from the cute trick of anthropomorphizing the inanimate object with a clever acronym, Ross is the sort of machine automation that many have been warning will herald an existential crisis in middle-class professions in the near future.

Automation is as funny thing. Knowing where the boundaries should lie to increase and maintain human effectiveness is rarely planned, and convention places huge biases on what we should or shouldn’t leave to machines.

My understanding of Ross, and a number of similar initiatives using AI and machine learning, is that they mechanize the sort of repetative tasks that today see unqualified paralegals and lawyers-in-training pouring over mounds of documentation trying to find things. It’s a classic case of industrialisation- specialisation and allocation of appropriate tasks to machines that need neither feeding or watering and can perform at scale.

But a question remains: to what extent do such repetitive, mundane tasks equate to deliberate practice? Do we run risk of automating activities that, whilst dull and repetative, are a crucial part of making a professional professional?

Let’s take two extremes.

I learn nothing about how to drive well these days from having an in-depth knowledge of the internal combustion engine. In fact my car barely allows me to check the oil level manually these days. Yet when my parents were learning to drive, knowing the basics of auto mechanics were a crucial skill because cars were so much less reliable.

On the other hand, I can get a computer to play musical scales faster and more precisely making a noise like just about any instrument imaginable. But doing so with make not one iota of difference to my own musical ability on the saxophone. Automation of playing scales would actually negatively impact my own performance.

As the application of AI and machine learning technologies starts to be explored do need to factor in the impact that they might have on human intelligence and learning. Some things can quite happily run as black boxes. However others are crucial building blocks to establishing a body of knowledge that can be applied effectively to the real world. Short term cost expediency might have significant unintended longer-term impact.

It’s also easy to dismiss drudgery as some sort of rite of passage, a desk-based tarring and feathering, that has no real purpose. But we should forget that most ceremonies are, on the face of it, pointless. The purpose they serve is to create and reinforce cultures. The impact of machines on professions is going to have many dimensions above merely time-effectiveness.