There is a machine learning method called Bi-directional Long-short Term Memory or BiLSTM for short.
The machine goes back and forth over the data, remembering things from the past, trying them out of new cases, building up pattern, improving its accuracy of prediction.
Again. Let’s not stress about the method. Let’s think about what it might tell us about how we learn.
I am using as my guide on diagram. A complicated diagram. So let’s go through it step by step.
Across the top are two blue lines, one with a circle at each end (indicating a measure completed by the young person) and one with a cross at each end (measurement by family members.
The measure concerns family functioning. The family reckons things are getting worse between baseline and follow up, the young person that they are getting better.
Down the bottom, we find two red lines, this time measuring the young person’s mental health. This time, both young person and family members calculate that things get better between baseline and follow-up.
Also down the bottom we have a green (young person’s assessment) and black (family members assessment) lines measuring young person’s functioning from week to week. As is plain to see, progress is not linear. The lines go up and down. Second, the young person and family members perspectives don’t always accord. In fact, on some weeks they are diametrically opposite.
To me, the machine is able to analyse the world as it is.
It’s not as if these patterns have been unknown to era 2 scientists. It is just that we haven’t really known what to do with the data, how to bring it all together. So most scientists revert to the default position of estimating the before and after situation, usually relying on measures completed by young person.
The machine can accommodate the extra data. The BiLSTM method has it going back and forth looking for pattern not only in the outcomes summarised in the above diagram, but also taking into account some of the contextual variables emerging from the topic analysis.
To me, the machine is able to analyse the world as it is. Not a world with linear progression from bad to good on a single variable, but a world where some things are good, somethings are bad, and there is a lot of fluctuation from one moment to another.