Links #10

Voting theory is approximately solved, voters are the problem.

1.

Voting systems are pretty important, not only for governance, but because most public goods funding schemes are essentially voting schemes in the large population limit. Unfortunately, there are a lot of impossibility theorems surrounding voting systems.

Maximal lotteries, get around these impossibility theorems using nondeterministic choices. They have a lot of nice properties and seem like they could work well in practice. Unfortunately, they have some weird attributes and are probably too complicated for voters to understand.

After bumping into this field a couple of times, I’ve converged on a hierarchy of voting systems depending on how sophisticated voters are.

For real-world voting, approval voting is by far the most common choice of reformers and voting theorists, using proportional representation where possible. It’s simple, easy to switch to, hard to manipulate, approximates utilitarianism, and has good Bayesian regret as well as voter-satisfaction efficiency.

For semi-sophisticated groups like representatives, Marcus Pivato’s Approximate implementation of Relative Utilitarianism via Groves-Clarke pivotal voting with virtual money could work. It has nice properties and is pretty easy to understand, though I think it would require repeated interactions of the group to work¹.

For repeated decision making amongst extremely sophisticated agents, the futarchy-like system posed here could work².

2.

Finding optimal voting systems is fun, but the real problem is voters. Voters are fundamentally unreasonable. We have a demand for misinformation and clear self-confirmation bias³. Worse, we have little incentive to become informed and are thus shockingly uninformed about political issues.

I don’t blame us, voting is intrinsically hard; I feel uninformed every time I go to the ballot box. Look at how much work Scott Alexander puts into the CA ballot propositions, and that’s probably not even 10% of the effort they deserve. Voting “correctly” is infeasible and wasteful in general.

So, how do we reconcile democratic governance with effective governance? We could try to educate voters more, but these interventions never scale. We could also enhance foot voting to give people direct choice over the policies they are subject to. But this does little to fix the problems of democracy within a jurisdiction.

The founders of the United States grappled with the same issues, and settled on representative democracy combined with clear limits on the domains that elected representatives had control over. If a social choice system is destined to be dysfunctional, we might as well limit it to the decisions where it’s really needed.

People have proposed ways to strengthen representative democracy through a parliamentary system, sortition, deliberation, epistocracy, less transparency, more legislator independence, or higher pay for public officials.

These ideas sound good in theory, but legislative procedures are hard to change. Are there technologies we can build today to improve democracy⁴?

3.

A tidbit from a new paper on fertility forecasts: Global births peaked in 2016. Since the average age of first childbearing is 30, that means that the 2040’s will be the most important decade for fertility decisions.

Looking at the population under age 1 in big countries (countries that will ever surpass 200 million people, which represent 54% of world population in 2100), most stay flat or peak later in the century, but some peaked remarkably early. India peaked in 2002, where the average age at first childbirth is 28, suggesting that their turning point could start in a little over 5 years.

China’s statistics look wildly different from everywhere else and it would be valuable to get better demographic data.

A related chart I made: U.S. population as a fraction of world population. Looking forward, this is projected to bottom out at 3.75% before rising again. Interested to hear people’s theories for how this explains everything.

Everything else

Algorithmic Bayesian Epistemology, Eric Neyman’s PhD thesis. Informal commentary here.

Balanced Donor Coordination and participatory budgeting are important facets of public goods funding that I didn’t cover in my review.

Singular Value Representation: A New Graph Perspective On Neural Networks

Aethero is making rad-hardened chips to do computation in space.

A Technical Review Of Nanosystems

Quanta piece on neutral atom quantum computers. I also find the atom arrays interesting for their ability to act as metamaterials and photomasks. EDIT: see also Optical tweezers throw and catch single atoms.

See Andy Mckenzie’s recent linkpost discussing a new result in connectomics. Andy says:

Ten years ago, there were debates about whether whole brain emulation could ever work because electron microscopy could only perform profile a subset of biomolecules, and ultrastructure alone might not be sufficient. With the advances in expansion microscopy, culminating in this study so far but sure to advance much further in the future, it’s hard to make that argument today.

Chaser: Eric Betzig: "Confessions of a Frustrated Optical Microscopist."

David Ho: “In Bergen, Norway, there are no garbage trucks and no trash piling up on streets because all over town, there are receptacles that connect to an UNDERGROUND PNEUMATIC WASTE TRANSPORT SYSTEM. The waste collects for a bit, and then WHOOSH...it's off to the recycling center”

Dean Kamen, inventor

How much have temperatures risen in countries across the world? I would love to see someone map this data at a finer resolution. Overall I’m surprised at how little some equatorial regions have warmed. These are the places that are already hot and poorly equipped to adapt to climate change, so that’s good to see.

Spray drying or spray evaporation might be a good desalination system. Penghua Guo seems to be a leading researcher in this area. See discussion here.

Tools for Thinking About Censorship

Pair the above with:

• The True Purpose of Propaganda

• Manufacturing Consent

• [Chinese] Censorship: How Does It Work?

A Classical Liberal Framework for AI Regulation

Anton Howes studies the industrial revolution:

The paper charts the emergence and spread of an improving mentality, tracing its transmission from person to person and across the country. The mentality was not a technique, skill, or special understanding, but a frame of mind: innovators saw room for improvement where others saw none. The mentality could be received by anyone, and it could be applied to any field – anything, after all, could be better.

But what led to innovation’s acceleration was not just that the mentality spread: over the course of the eighteenth century innovators became increasingly committed to spreading the mentality further – they became innovation’s evangelists. By creating new institutions and adopting social norms conducive to openness and active sharing, innovators ensured the continued dissemination of innovation, giving rise to modern economic growth in Britain and abroad.

Unflattering aspects of Effective Altruism

Confessions of a Cheeseburger Ethicist

The Best Tacit Knowledge Videos on Every Subject

1. I’m also not sure how well it fairs with agenda-setting, sybils, or coalitions but I think the latter problem can be fixed.

2. Though I expect AGI’s will be able to come up with their own system.

3. Don’t think that you or I are immune to such impulses!

4. Off the top of my head, decentralized and secure voting could be valuable. The internet and prediction markets seem like they increase political knowledge and consensus. Tools like Policy Engine and institutions like the OSTP or the CBO can help inform representatives. Language models might do a good job of facilitating consensus or representing diverse viewpoints.