A history of ITRI, Taiwan's national electronics R&D institute. It was established in 1973, when Taiwan's income was less than Pakistan's income today. Yet it was single-handedly responsible for the rise of Taiwan's electronics industry, spinning out UMC, MediaTek and most notably TSMC. To give you a sense of how insane this is, imagine that Bangladesh announced today that they were going to start doing frontier AI R&D, and in 2045 they were the leaders in AI. ITRI is arguably the most successful development initiative in history, but I've never seen it brought up in either the metascience/progress community or the global dev community.

I'm considering writing about "RCTs in NGOs: When (and when not) to implement them"

The post would explore:

• Why many new NGOs feel pressured to conduct RCTs primarily due to funder / EA community requirements.
• The hidden costs and limitations of RCTs: high expenses, 80% statistical power meaning 20% chance of missing real effects, wide confidence intervals
• Why RCTs might not be the best tool for early-stage organizations focused on iterative learning
• How academic incentives in RCT design/implementation don't always align with NGO needs
• Alternative evidence-gathering approaches that might be more appropriate for different organizational stages
• Suggestions for both funders and NGOs on how to think about evidence generation

This comes from my conversations with several NGO founders. I believe the EA community could benefit from a more nuanced discussion about evidence hierarchies and when different types of evaluation make sense.

How people who write on the EA Forum and on LessWrong can have non-obvious significant positive impact by influencing organizations (like mine) - both through culture and the merit of their reasoning.

"EA for hippies" - I managed to explain effective altruism to a group of rich hippies that were in the process of starting up a foundation, getting them on-board with donating some of the revenue to global health charities. 

The post would detail how I explained EA to people who are far from the standard target audience.

I have a hastily written draft from a while back called "Cause neutrality doesn't mean all EA causes matter equally". It's a corrective to people sometimes using "cause neutrality" as a justification for not doing cause prioritisation/ treating current EA cause areas as canonical/ equally deserving of funding or effort. I didn't finish it because I ran out of steam/ was concerned I might be making up a guy to get mad at. 
I'll consider posting it for Draft Amnesty, especially if anyone is interested in seeing this take written up.

Some things you might want to do if you are making a weighted factor model

Weighted factor models are commonly used within EA (e.g. by Charity Entrepreneurship/AIM and 80,000 Hours). Even the formalised Scale, Solvability, Neglectedness framework can, itself, be considered a form of weighted factor model.

However, despite their wide use, weighted factor models often neglect to use important methodological techniques which could test and improve their robustness,  which may threaten their validity and usefulness. RP's Surveys and Data Analysis  team previously consulted for a project who were using a WFM, and helped them understand certain things that were confusing them about the behaviour of their model using these techniques, but we've never had time to write up a detailed post about these methods. Such a post would discuss such topics as:

• Problems with ordinal measures
• When (not) to rank scores
• When and how (not) to normalise
• How to make interpretable rating scales
• Identifying the factors that drive your outcomes
• Quantifying and interpreting disagreement / uncertainty

How to interpret the EA Survey and Open Phil EA/LT Survey.

I think these surveys are complementary and each have different strengths and weaknesses relevant for different purposes.^[1] However, I think what the strengths and weaknesses are and how to interpret the surveys in light of them is not immediately obvious. And I know that in at least some cases, decision-makers have had straightforwardly mistaken factual beliefs about the surveys which has mislead them about how to interpret them. This is a problem if people mistakenly rely on the results of only one of the surveys, or assign the wrong weights to each survey, when answering different questions.

A post about this would outline the key strengths and weaknesses of the different surveys for different purposes, touching on questions such as:

• How much our confidence should change when we have a small sample size from a small population.
• How concerned we should be about biases in the samples for each survey and what population we should be targeting.
• How much the different questions in each survey allows us to check and verify the answers within each survey.
• How much the results of each survey can be verified and cross-referenced with each other (e.g. by identifying specific highly engaged LTists within the EAS).

1. ^{^}

   Reassuringly, they also seem to generate very similar results, when we directly compare them, adjusting for differences in composition, i.e. only looking at highly engaged longtermists within the EA Survey.

I'm thinking of writing a longer/ more nuanced collaborative piece discussing global vs local EA community building that I touched on in a previous post.

At some point I'd love to post something on ‘How to think about impact as a journalist’. I've accumulated a few notes and sources on the subject, and it's a question I often come back to, being directly concerned. But it's a big one and I haven't yet decided how to tackle it :)

I'm going to post about a great paper I read about the National Woman's Party, and 20th century feminism that I think has relevance to the EA communtiy :)

Reputation Hardening

Prompted largely by the fall in EA credibility in recent years. And also being unsatisfied with GiveWell's lack of independent verification of the charities they recommend.

Here is a lightly edited AI generated slop version:

Reputation Hardening: Should GiveWell Verify Charity Data Independently?

"Reputation hardening" involves creating more resilient reputations.

Recent events have shown how reputation damage to one EA entity can affect the entire movement's credibility and therefore funding and influence. While GiveWell's evaluation process is thorough, it largely relies on charity-provided data. I propose they consider implementing independent verification methods.

Applying to GiveWell/GHD

• Mystery evaluators conducting unannounced site visits
• Independent surveyors (potentially hiring former charity employees) collecting impact data
• Creating funding opportunities for third-party monitoring organizations GiveWell have already been involved in market shaping

These measures could help detect potential issues early and strengthen confidence in effectiveness estimates.

This is a preliminary idea to start discussion. What other verification methods or implementation challenges should we consider?

I’d like to write: 

A post about making difficult career decisions with examples of how I made my own decisions and some tools I used to make them, and how they worked out. I have it roughly written but would definitely need feedback from you Toby before I post :))

A post about mental health: why I’m focusing on it this year, why I think more people in EA should focus on it and what exactly I’m doing, what’s working etc. Haven’t written it yet, but a lot of people are asking about it so I do think there is potential value. 

I would write how there's a collective action problem regarding reading EA forum posts. People want to read interesting, informative, and impactful posts and karma is a signifier of this. So often people will not read posts, especially on topics they are not familiar, unless it has already achieved some karma threshold. Given how time-sensitive front page availability is without karma accumulation and unlikely relatively low karma posts are too be read once off the front page, it is likely that good posts could be entirely ignored. On the other hand, some early traction could result in OK posts getting very high karma because a higher volume of people have been motivated to check the post out. 

I think this could be partially addressed by having volunteers, or even paying people, to commit to read posts within a certain time frame and upvote (or not, or downvote) if appropriate. It might be a better use of funds than myriad cosmetic changes. 

Below is a post I wrote that I think might be such a post that was good (or at least worthy of discussion) but people probably wanted to freeride on others' early evaluation. It discusses how jobs in which the performance metrics actually used are orthogonal to many ways in which good can be done may be opportunities for significant impact. 

https://forum.effectivealtruism.org/posts/78pevHteaRxekaRGk/orthogonal-impact-finding-high-leverage-good-in-unlikely

My previous attempt at predicting what I was going to write got 1/4, which ain't great.

This is partly planning fallacy, partly real life being a lot busier than expected and Forum writing being one of the first things to drop, and partly increasingly feeling gloom and disillusionment with EA and so not having the same motivation to write or contribute to the Forum as I did previously.

For the things that I am still thinking of writing I'll add comments to this post separately to votes and comments can be attributed to each idea individually.

A BOTEC of base rates of moderate-to-severe narcissistic traits (ie, clinical but not necessarily diagnosed) in founders and their estimated costs to the ecosystem. My initial research suggests unusually high concentrations in AI safety relative to other cause areas and the general population.

My ideas for draft amnesty week are replied to this message so they can be voted on separately:

I'll post my ideas as replies to this, so they can be voted on separately.