Our collaborators on this work at the Newcastle University Open Lab were engineers, experts in machine learning. These words ‘machine learning’ will become ubiquitous in the next era of learning, as will ‘artificial intelligence’. (Machine learning describes the ability of a computer to learn from experience, artificial intelligence is the application of that learning to real world problems).

The machine is a computer. It learns. It can learn with or without human supervision, what they engineers refer to as ‘unsupervised’ and ‘supervised’ learning.

We started off with a couple of short exercises aimed at intimating how the machine thinks. The first is all to do with housing in San Francisco and New York.

We drew several conclusions from the example.

1. The machine is interested in pattern. The green and blue dots that appear on the first page are the machine’s stock in trade. It is looking for differences in pattern. Researchers in the group reflected on how this is not unlike how they have approach qualitative data, going over and over again to see if a pattern emerges, and then checking the data again to see if what we think we are seeing is real. The machine is doing the same. But it can remember more than a human, it can keep a lot more in ‘its head’ than a human.

2. The machine doesn’t stop learning. As long as nobody pulls the plug and there is more data to analyse it will keep cutting and splicing the information it has to see if it can learn more. The second example was just a game, a little battle between human and machine, a battle that machine will inevitably win, because it doesn’t stop learning.

3. The machine is asking different questions from a human. We might ask things like ‘are homes in New York City smaller than in San Francisco?’. Or ‘are homes in New York City more expensive than in San Francisco’ The machine thinks: ‘what are the distinctive patterns in this set of data’?

4. But therein lies an opportunity for the human. The machine will always find pattern. That is what it is trained to do. But sometimes that pattern will be meaningless. The human -what machine learning engineers call the ‘content expert’ has to decide if there is meaning. So machine learning demands collaboration. The machine will see things we don’t see. We will see things the machine doesn’t see. Ergo. Let’s make friends with a machine.

As I stressed in the meeting, I am not advocating for machine learning. The downsides are manifold and possibly better explained by Charlie Booker in his TV series Black Mirror than by me. But nor is the much point resisting machine learning. It is upon us, just as the World Wide Web was upon us at the turn of the century. We can work to make it a force for good. But we cannot make it go away.

In the inputs that follow, I will give a few examples of machine learning from the analysis of the Care4 Functional Family Therapy data. I am not so much making the case for machine learning, as drawing out the implications for the next era of learning.