Using AI to build evidence can easily produce a house of cards. Give it some data and it can quickly return code, categories, charts, networks and a confident conclusion, with each layer resting on choices we may never have examined. That is a good reason to look underneath—but also an invitation to play. AI gives us a fast way to try methods, change assumptions and see what happens when we pull out a card.

At heart, building evidence involves two deceptively simple questions. How do our data represent the thing we want to study? And what are those data telling us, in those terms? I can show you some fundamental methods for counting, comparing, plotting distributions and finding relationships. Networks are a particularly vivid example: choose what counts as a node and a connection, draw the result, then explore centrality, clustering and communities. Change the representation and the evidence may change with it.

I have spent much of my career using data, mathematical models and network analysis to study biological and political systems. I can show you how to use AI and Python to investigate small political datasets, learning fundamental methods for counting, comparing, visualising and analysing networks as we go. The aim is to build enough practical understanding to produce a simple analysis, explain what its data represent and recognise when its conclusions outrun the evidence.

Core reading

• David Spiegelhalter, The Art of Statistics: Learning from Data (Pelican, 2019).
• Duncan J. Watts, Six Degrees: The Science of a Connected Age (W. W. Norton, 2003).