Controlling variables at the point of collection has consequences you don't always see until much later.
I've been working through a dataset documentation framework by DAIR, an independent AI research institute, for my research into date palm ash as an agricultural waste ash candidate for ceramic glazes. It's not just another data housekeeping exercise. With so many variables - from soil salinity and combustion temperature to frond position and irrigation source - decisions made in the field would compound by the time they reach the model.
I already knew I'd need to anonymise landowner data before I started. I realised a while ago that the initial model wasn't universal. It can't be. Filling the data gaps that would make it truly global would mean crossing multiple borders and coordinating stakeholders I don't have access to. So I'm keeping it localised, deliberately.
That decision made me think differently about what I'm actually building. Maybe it's less a model and more a framework. A methodology for characterising biomass ash candidates that others could apply elsewhere, with their own data, within their own boundaries.
The image is something I made while obsessing over ISO dates. A geometric tablet form on a red ground, overlaid with Python date code. The first documented data was recorded on clay tablets in Iraq. I'm half Iraqi. My father worked with clay tablets. The research occupying me now is rooted in the same geography. The act is the same across five thousand years: fix a moment so it can be trusted later.
Which brings me to the question I keep coming back to: how do you model the future of more sustainable ceramic glazes when some of the materials you want to use, beyond Europe and North America, haven't been properly characterised yet?
