The Big Bird Seed Experiment
An exploration of causal inference and ML.
I’ve recently been reading The Book of Why by Pearl and Mackenzie. This has made me want to design my own causal experiment, something where I can make an intervention A rarity for a rewilder, you might assume. and measure the impact. Serendipitously, work has just taken an order of AudioMoths, a bioacoustic device designed for recording bird and bat sound. The cogs started whirring. Bird presence is the output variable, but what can I do? I also (unrelated) just took an order of 10kg of bird seed. Sounds like as good of an input as any. Thus began the “Big Bird Seed Experiment” That's a big bird-seed experiment, not a big-bird seed experiment.
This project will be split over multiple posts, as I develop each part. The broad plan is:
- Detection: configuring bioacoustic sensors, hardware and software.
- Experimental Design: planning the experiment, cause and effect, causal inference, ethics.
- Collection: doing the thing, measuring feed and birds.
- Analysis: can we answer any of our questions from Part 2? Causal inference, Bayesian and frequentist statistics.
- Reflection: what now, what next and why.
I’ll be sharing as I go in a “digital garden” type approach. Follow at your own risk.