This week I was watching a news report looking at how the Serious Fraud Office (SFO) were using Artificial Intelligence to analyse documents in complex cases. AI cuts down the huge amount of time it would normally take the SFO to analyse the vast amount of information collected for cases. What struck me was their demonstrations of the output. After the system had worked through the data (sometimes 10’s of millions of documents), the interface officers have to review the findings is an infographic. A stringed bubble comparison in fact, which is then used to identify interesting relationships and patterns in communication.
Obviously that in itself makes me happy, that the best way to analyse, make sense of, “see” the data is an infographic. That actually the work the AI is doing is essentially presenting the data in a different way so that we understand it better. Technology now means that infographics are no longer an end product, a presentation tool to show findings, but also the tool we use to actually analyse the data. We can manipulate the graphics to pull out key findings and use them to answer questions.
Going further, what was going to be a YAY INFORGRAPHICS! blog has developed into something even better…
It was obvious to me that the human input to the situation was essential. Initially telling the system what was interesting and needed to be highlighted and then to analyse the output, using the graphic to identify what was pertinent to the case. It was the officers knowledge and understanding of the subject which allowed them to make sense of all that information, however it was presented. You cannot get anyone or anything to answer a ‘so what’ question if they don’t have an understanding of the situation. Data Science is not just crunching numbers, its about knowing what evidence to layer up together, what extra information to have and trying to anticipate what questions will need answering.
I think collaboration is the key. Knowing who the specialists are, communicating and working with them to collect and paint as detailed pictures as possible.
Data is its own peculiar language and analysts, what ever their titles, are interpreters.