Note: NATO spending targets refer to *public* spending, where as CC mitigation refers to *total* spending. So if we only compared public spending, this chart would look even more extreme.

Putting spending targets into perspective serves as one of the most important reminders of this decade: climate action is possible, cheaper than usually framed as, and definitely cheaper than inaction.

Determined climate action turns out really cheap in comparison. Among other things, this makes the opportunity costs of conflict stupendously gigantic - not even talking about matters of justice here.

#philsky #econsky

Oh what an honour! Wishing them luck for the final push

Thanks Anna! You're exactly right - thats where it started. Indebted to many brilliant people at Cambridge HPS.

Lots of interesting problems to solve there!

This project was kicked-off by a great paper by @inkerikoskinen.bsky.social, which I recommend reading. While social epistemology talked about opacity & trust for years, debates on AIs got major attention for (seemingly) the same issues, but without much interaction to social epistemology.

True! But this still makes it consistent with a subset view in the sense that quant. evidence is special qual. evidence that by virtue of its special features has diff properties. Qualit. data already contains these relations, they are just not as cashed out. Quantification as a form of compression.

(e.g. word count (number), mentioned X (boolean), said Y is important (boolean), etc.)

Sorry, this got longer than expected!

For practical reasons, this context is often stripped away and you end up with data that looks and feels quantitative. But the other way works, too! An interview comes with a lot of context, but requires further processing if you wanted to squeeze some of its information in numeric data types

(„I went to the machine and read of this value from the display, interpreting it such and such, not sure if something was off though and I went and asked my friend“; „Procedure X spat out this Excel sheet“).

All of them are qualitative for two reasons: 1. your reason: even a simple quantitative boolean requires interpretation, just like a long-form text does. 2. if you add the context back to „raw“ quantitative data, you can turn every quantitative datum into data that also looks and feels qualitative.

lending it to certain types of analysis. For example, you can easily aggregate number types (sums, medians) but cant do that with qualitative interviews.

Numeric data types (or booleans, or timestamps, or geospatial types, etc.) are typically (and meaningfully) referred to as quantitative because they are different from (longer) text, audio or image data, in the sense that they require different ways of processing,