Differences

This shows you the differences between two versions of the page.

--- miscellaneous_institution_design_ideas [2022/12/17 23:52] – [Miscellaneous institution design project ideas] katjagrace
+++ miscellaneous_institution_design_ideas [2023/02/05 20:21] (current) – [Modes of activity] katjagrace
@@ Line 7: / Line 7: @@
   * **Virtual countries**: Try to make a real collective with really good institutions for all the main things you need to be a sort of agentic force (like good information spreading and processing, good decision making, ??) kind of like a non-land-based country and try to be functional enough to take over existing less functional countries (stretch goal)
-===== Modes of activity =====
-  * **Software projects**: make random things that seem fast and locally good (e.g. reciprocity, chathamroom.com/)
-  * **Blogging**: think about these things and write blog posts.
-  * **social signaling systems**: I've thought a lot about this, including looking at a bunch of econ papers and working on it in grad school a bit. I think people mostly think about it fairly badly, because it brings up statusy feelings and they don't kind of unemotionally consider the mechanisms involved very well. This seems like it could be improved upon by explicit thought. (e.g. often value is added by people being able to signal traits; adding extra mechanisms for them to signal traits is often great for them and people trying to sort them. This seems like a promising route to a lot of social justice stuff which is not naturally that interesting to social justice people. e.g. Manifold Markets contributes to gender equality by making it possible for quiet women to just numerically demonstrate that they are better at predicting stuff than Eliezer. I like this sort of thing a lot more than the ' try to shame everyone until they listen to you' vibe of current social justice attempts, and think it would be cool if such things could solve a lot of problems and also cool off unhappiness about such things and replace what I see as harmful SJ forces against e.g. honest and open discussion. But this is probably not the most important thing.)
-  * there is a kind of thing, '**officialness**', which is involved in a lot of institutions. I'd like to better understand what's up with it and what it is doing.
-  * **Status** is another interesting existing system for humans interacting that I want to understand better. It's kind of like a currency, but different in ways. What are the ways?
-  * **Systematizing vague social things** money is a pretty cool invention, and presumably replaced vague bartering that probably felt totally sufficient, and like really quantifying what was going on would be silly. Probability theory also seems great, and I think replaced vaguer thinking about likelihood that felt sufficient. I wonder if there are more things we currently do kind of vaguely that would be better if replaced by something intensely quantified? (status/reputation seems like a clear possibility there)
-  * **Making and displaying estimates** The Squiggle team seem focused on the first step of displaying. Nathan Young is interested in it. Ben Weinstein-Raun might be too.
-  * **Anonymous discussion**Ben WR made chathamroom.com/ and Katja is pretty interested in it. Seems great for having good discussions about AI safety that aren’t gummed up by people fearing that they seem stupid. Probably many other use cases
-  * Run **small scale experiments** of institutions that seem promising. Robin Hanson used to say this was a key bottleneck to innovation iirc—people are happier to theorise than to do stuff
-===== Experiments in **making reasoning really explicit and careful =====
-  * Paul Christiano and Katja’s ‘Structured Case Project’
-  * AI Impacts: arose out of relative failure of structured case project. Still involves some attempting at relatively watertight argument e.g. https://aiimpacts.org/likelihood-of-discontinuous-progress-around-the-development-of-agi/
-  * Generally intended as experimental epistemic object:
-  * collection of editable pages
-  * Each page is a modular node attempting to answer a question
-  * Each page has summary answer at the top, then the page supports that answer
-  * Lower level pages answer questions that are helpful for answering questions on higher level pages
-  * Turning into a wiki: wiki.aiimpacts.org/
-  * Someone made software that did the thing Katja is always asking for, and it somehow didn’t seem great, so that’s evidence against Katja’s intuitions on this
-  * Probably this?
-  * Attempt to analyze what is actually hard about making structured arguments
-  * Katja has done this some, still feels mysterious
-  * Improving people’s epistemics
-  * Pastcasting - forecasting past results so you get instant feedback
-  * https://www.quantifiedintuitions.org/pastcasting
-  * Calibration - see how justified your own confidence is
-  * https://www.quantifiedintuitions.org/calibration
-  * Open Phil one
-  * Critch (I think?) made an app
-  * Historical Base rates - much thinking requires knowing roughly how often things have happened in the past
-  * Seems like Our World In Data might do some work here
-  * There is at least one other effort
-  * Increasing summarisation - making it easier to get a basic understanding on EA/LessWrong topics
-  * Nonlinear offer a range of prizes for summaries of such works
-  * Nathan Young
-  * Fighting ‘Moloch’
-  * What is Moloch exactly?