But this is helpful. Makes me think I should write up this argument more formally 🤔

Sure, but also not an issue space. Point is: Low-dimensional behavior doesn’t necessitate a low-dimensional issue space. There is also a simple argument to show that if any low-dimensional issue space warps in nonlinear ways under perturbation.

Party != issue space In any case, most large parties are big coalitions with contradictory positions.

Exactly. Low-dimensional outcomes are not indicative of an issue space. Interactions also break symmetries.

This is obviously a major problem for the dominant frameworks, where we imagine voters are twisting around on solid ground as they change preferences. But the world is changing too! So this picture is misleading. The ground is flat only (!) if you allow for the dimensions of your space to change.

There's no fixed issue space. In fact, the axes of this low-dimensional issue space are being warped in nonlinear ways! Not only are the set of topics changing, but the ground is warping, melting, and folding all around you.

Issue space is a low-dimensional mapping of a much larger space of objects, i.e., beliefs, interests, culture, etc. The full set of things are being constantly regrouped and reweighted into new "issues"--think Republican party today versus 10 years ago.

Why is political science so adamant that the "issue space" is fixed??? It doesn't have to be, and it's not. This is a central assumption of the spatial voting theory frameworks, and people still cling onto it today. See Martin & Quinn and Poole & Rosenthal. 🧪🦋

Delighted to be here today!

This is overlooked even in current theory. Thus, in a strange sense, we are seeing a consequence of the myopia of neoclassical economics.

Makes it clear that that the current US administration’s focus on diminishing institutions while “promoting” means of production is actually imposing substantial costs in terms of uncertainty.

Reading Douglas North and the importance of transaction costs in economic activity, which has played no role in neoclassical theory. Yet, they are prominent in an exchange-based economy.

Statistical mechanics ~= statistical learning The former aims for minimal complexity in representation, the latter maximal. Thus, one can formulate a statistical mechanics of learning, but not vice versa.

'This presents a fundamental challenge in strategic decision-making for both biological and artificial agents across diverse competitive and cooperative landscapes [1–3], of which war [4,5] and conflicts [6–8] are poignant examples.'