Derek

Karma: 863

Derek Jan 20, 2020, 3:12 PM
4 points
0 ∶ 0
on: Logarithmic Scales of Pleasure and Pain: Rating, Ranking, and Comparing Peak Experiences Suggest the Existence of Long Tails for Bliss and Suffering
Thanks for this—very interesting.
Do you think your claims would apply to broader measures of subjective wellbeing, e.g. questions like “Overall, how satisfied are you with your life?” and “Overall, how happy were you yesterday?” (often on a 0-10 scale)? Or even to more specific measures of valenced experience, like depression (e.g. PHQ-9)?
Because I’ve been wondering whether:
(a) the Weber-Fechner law is limited to perception of clear physical stimuli (weight, pain, spicyness, etc), as distinct from ‘internal’ states and cognitive evaluations (though the internal/external distinction may not make sense here).
(b) a log scale is less useful/accurate when considering long periods of time (a day, a year, a lifetime), over which the variance in average wellbeing in a population will be lower than the variance in the intensity of specific events.

Derek Jan 12, 2020, 7:39 PM
15 points
0 ∶ 0
on: Physical Exercise for EAs – Why and How
This is very good, but I think busy (or unmotivated) EAs without much exercise experience would benefit from even more specific recommendations, especially for resistance exercises (i.e. strength training).
I found the Start Bodyweight program useful when beginning resistance training at home with no equipment other than a pull-up bar. An EA recommended the book Overcoming Gravity for more detailed information on bodyweight exercises.
I now I prefer to use the gym. At a glance, the following (which I just found with a quick Google search) seem like sensible gym-based* options for beginners, but maybe you have better ideas.
https://www.muscleandfitness.com/workouts/workout-routines/complete-mf-beginners-training-guide-plan
https://stronglifts.com/5x5/ [I’d add some core exercises to this, like situps and planks]
https://www.shape.com/fitness/workouts/strength-training-beginners
When I’m too busy to do the full range of strength and cardio (or when I’m travelling), I sometimes do moderate/high-intensity interval classes at home using YouTube videos. The Body Coach is pretty good—he has a videos with a range of difficulty (beginner to advanced), duration (10 min+), and muscle focus (legs, upper body, abs, full-body, etc). There are also videos meeting specific needs, e.g. low-impact routines so you don’t disturb your neighbours or hurt your knees, and ones designed for small spaces. This kind of thing is perhaps the most efficient form of exercise: you can do it anywhere, it doesn’t require any equipment, it’s free, it covers both cardio and strength, and it doesn’t take much time.
When travelling, I also take a resistance band. If you choose the weight carefully, a single band (which folds up to the size of a cigarette packet) can arguably substitute for any dumbbell that you’d use in the gym, and some of the machines as well. (The main thing you’re lacking is the ability to do deadlifts, but there are ways around that too.)
I’ve heard some EAs recommend GymPass, especially if you travel a lot and don’t like to exercise alone.
Feel free to correct me on any of this – I don’t have any relevant expertise.
*They could obviously be done at home if you buy the equipment. The last one just needs dumbbells or resistance bands, which are pretty cheap.

Derek Jan 3, 2020, 1:44 PM
10 points
0 ∶ 0
on: 2019 AI Alignment Literature Review and Charity Comparison
Why isn’t there a GiveWell-style evaluator for longtermist (or specifically AI safety) orgs?

Derek Jan 2, 2020, 8:50 PM
2 points
0 ∶ 0
on: New research on moral weights
Section 4 on subjective wellbeing is interesting.
• Across poor respondents in Kenya and Ghana, the average life satisfaction ladder score is 2.8 (where 0 is the lowest and 10 is the highest score).
• Respondents with higher consumption have higher life satisfaction ladder scores; doubling consumption is associated with being 0.4 steps higher on the ladder.
• When describing different points on the ladder respondents most often referred to levels of money and material goods. In contrast, health states were mentioned much less often with regards to life satisfaction. Having a health condition was associated with being 0.3 steps lower on the ladder.
• Overall, taken alone, these findings suggest that consumption is of greater relative importance to wellbeing of respondents than their preferences (described in Section 1-3) indicate.
I notice they only measured life satisfaction. Can you tell me why they didn’t also include at least one measure of hedonic wellbeing, such as those used in the evaluations of GiveDirectly? It is really important to understand whether potential GiveWell top charity beneficiaries are actually unhappy (i.e. generally feel bad) or just dissatisfied with their material circumstances when someone with a clipboard asks them about it. (Life satisfaction is much more sensitive to relative wealth and status than is pleasure/misery.) For instance, this may be the critical factor when choosing between life-extending and life-improving interventions.

Derek Jan 1, 2020, 6:23 PM
5 points
0 ∶ 0
in reply to: cole_haus’s comment on: Uncertainty and sensitivity analyses of GiveWell’s cost-effectiveness analyses
Did you ever get round to running the analysis with your best guess inputs?
If that revealed substantial decision uncertainty (and especially if you were very uncertain about your inputs), I’d also like to see it run with GiveWell’s inputs. They could be aggregated distributions from multiple staff members, elicited using standard methods, or in some cases perhaps ‘official’ GiveWell consensus distributions. I’m kind of surprised this doesn’t seem to have been done already, given obvious issues with using point estimates in non-linear models. Or do you have reason to believe the ranking and cost-effectiveness ratios would not be sensitive to methodological changes like this?

Derek Dec 30, 2019, 9:02 PM
13 points
0 ∶ 0
on: Is mindfulness good for you?
This is very useful – thanks for writing it up.
This heterogeneity across intervention types means that we should be cautious about broad claims about the efficacy of mindfulness for depression and anxiety.
True, but that applies equally to claims of null or small effect sizes, e.g. some forms of mindfulness could be very effective even if ‘on average’ it’s not. Did any of the meta-analyses contain useful subgroup analyses?
(For what it’s worth, a few years ago I used the Headspace app ~5x/week for 3 months and found it to be actively detrimental to my mood. Anecdotally, this seem fairly common: https://www.theguardian.com/lifeandstyle/2016/jan/23/is-mindfulness-making-us-ill)

Derek Dec 28, 2019, 8:09 PM
1 point
0 ∶ 0
in reply to: Derek’s comment on: Learning to ask action-relevant questions
I guess “action-relevant” has a better noun form, which could be a non-trivial advantage.

Derek Dec 28, 2019, 8:06 PM
1 point
0 ∶ 0
on: Learning to ask action-relevant questions
Ask yourself: “If I imagine a world in which I have answered this question, what would look different?”
This sounds like the “importance” part of the ITN framework. From EA Concepts:
If all problems in the area could be solved, how much better would the world be?

Derek Dec 28, 2019, 8:02 PM
1 point
0 ∶ 0
on: Learning to ask action-relevant questions
I’m not sure if “action-relevant” is accepted terminology
About a year ago I heard the term “action guiding”, which I guess is the same?

Derek Dec 17, 2019, 11:27 PM
9 points
0 ∶ 0
in reply to: MichaelStJules’s comment on: We’re Rethink Priorities. AMA.
I’ve asked several academics with domain expertise to review draft posts, or sections of posts, or advise on specific issues. Some have been very useful, but they understandably do not have time to engage fully (if at all). As a consequence, I often worry that I’m making dumb mistakes, or just reinventing the wheel, and there are often substantial delays while waiting for expert input. I think the lack of access to academic networks and infrastructure is perhaps the biggest weakness of RP as a research organisation, and it is related to the youth and inexperience of EA as a whole.
I’m not sure it can be fully solved – some fields only have half a dozen people in the world working on them, so it may be impossible to find someone with enough free time to help out. But I suspect a lot of progress could be made, e.g. I bet there are a lot of statisticians and economists who would be willing and able to help if only they knew we needed it. At the mid- and late career professionals’ meetup at EAG San Fransisco last June, it was suggested that retired academics, professional groups, and LinkedIn might be good sources of mentors/advisors. Someone mentioned https://taprootfoundation.org/ as well – perhaps not for academic advice, but for support in other areas where EA orgs tend to be lacking, such as management. I’d be interested to see an effort to systematically connect experts with EA projects, perhaps through the EA Hub or 80,000 Hours.

Derek Dec 15, 2019, 12:37 PM
24 points
0 ∶ 0
in reply to: MichaelStJules’s comment on: We’re Rethink Priorities. AMA.
The following is a tidy, oversimplified version of what happened.
I learned about Bentham and Mill in A-level history class (aged 17) and I think read a Peter Singer book. I was very left-wing at the time but I remember being really frustrated that all the other altruistically-minded kids in my class supported standard leftist policies for ideological reasons even when they harmed disadvantaged people. This influenced me to study philosophy at undergrad level, where I defended utilitarianism.
Unfortunately EA hadn’t been invented at the time so I spent the first year after graduation working in warehouses and call centers, followed by about nine years of direct development work in low-income countries. I got frustrated by the inefficiency of most development orgs and decided to switch fields into either law (‘earning to give’ before I’d heard of the concept) or public health (to do direct work with more quantifiable impacts).
Around the same time I was searching online for information about charity evaluation and came across GiveWell, then the Singer TED Talk and the wider EA community. This may have influenced me to choose public health, though there were other factors (e.g. the 2008 financial crash made it even harder than usual to pursue a lucrative law career). I spent 18 months in Australia doing whatever work I could find – mostly farm labouring – to pay for my master’s course.
During the course I became more involved in EA, and got interested in health economics, especially methods for cost-effectiveness analysis. But I couldn’t get a job or PhD in health economics with a general public health background, so to save up for a second master’s I spent two more years doing mostly sub-minimum wage temp jobs, or saving dole money when I couldn’t find work (though I also got a bit of contract work with GiveWell towards the end of this period). Halfway through that course I ran out of money and had some health issues, so I took a leave of absence, during which time I worked on the 2019 Global Happiness Policy Report (Chapter 3), then got the Rethink job.
My reasons for continuing to work in EA are some mixture of those given by my colleagues.

Derek Dec 15, 2019, 11:41 AM
9 points
0 ∶ 0
in reply to: Kirsten’s comment on: We’re Rethink Priorities. AMA.
Most likely academic research related to the use of subjective wellbeing in prioritisation systems (healthcare, central government, maybe EA orgs, etc). Might have applied for researcher positions in other EA orgs.

Derek Dec 13, 2019, 11:15 PM
8 points
0 ∶ 0
in reply to: MichaelStJules’s comment on: We’re Rethink Priorities. AMA.
I’ve become a bit more longtermist in outlook and more uncertain of the sign/effect size of most interventions/projects, mostly due to issues around indirect effects/cluelessness.

Derek Dec 13, 2019, 11:08 PM
16 points
0 ∶ 0
in reply to: MichaelStJules’s comment on: We’re Rethink Priorities. AMA.
I’m not a philosopher, but to the extent I have opinions on such things they are about the same as Moss’s, i.e. classical hedonistic utilitarianism with quite a lot of moral uncertainty. I have somewhat suffering-focused intuitions but (a) I’ve never seen a remotely convincing argument for a suffering-focused ethic, and (b) I think my intuitions – and, I suspect, those of many people who identify as suffering-focused – can be explained by other factors. In particular, I think there are problems with the scales people use to measure valence/wellbeing/value of lives, both in reality and in thought experiments, e.g. it seems common for philosophers to assume a symmetrical scale like −10 to +10, whereas it seems pretty obvious to me that the worst lives – or even, say, the 5th percentile of lives – are many times more bad then the best lives are good. So if the best few percent of lives are ¹⁰⁄₁₀ and 0 is equivalent to being dead, the bottom few percent of any large population are probably somewhere between −100 and −100,000. (It is not widely appreciated just how awful things are for so many people.) If true, classical utilitarianism may have policy implications similar to prioritarianism and related theories, e.g. more resources for the worst off (assuming tractability). But I haven’t seen much literature on these scale issues so I’m not confident this is correct. If you know of any relevant research, preferably peer-reviewed, I’d be very interested.

Derek Dec 4, 2019, 5:18 PM
20 points
0 ∶ 0
on: How do cash transfers impact the people who don’t receive them?
[EDIT: I no longer endorse all of this comment. After looking more closely at the papers, I’m more confident that the spillover effects of the latest version of the program are neutral to positive (at least on humans – growth in meat consumption is an important caveat).]
Thanks for posting this.
Though not reported here, I was pleased to see that non-market effects were also recorded in the study, and that these were neutral or positive for both recipients (‘treated households’) and non-recipients.
For treated households, we find positive and significant effects for four of the six indices: psychological well-being, food security, education and security [i.e. crime rates]. Estimated effects are close to zero and not significant for the health index and female empowerment index. When looking at total effects including spillovers for the treated, we find a similar pattern for all but the security index. For untreated households, we find no significant effects of local cash transfers except for the education index, which is higher by 0.1 SD (p < 0.10). Importantly, we do not find evidence of adverse spillover effects for untreated households on any of the indices, with point estimates positive for all but the security index, which is indistinguishable from zero (-0.02 SD, SE 0.07).
I’m particularly interested in the “psychological wellbeing” index, which Appendix C1 says comprises a “weighted, standardized average of depression (10 question CES-D scale), happiness, life satisfaction, and perceived stress (PSS-4)”. I would like to know: (a) what measures were used for “happiness” and “life satisfaction”; (b) how the components of the index were weighted; and most of all (c) a breakdown of scores for each measure. I can’t find this information in the paper.
I’m asking because there is a fair amount of research suggesting that one person’s income increase causes wellbeing declines among other members of the community (i.e. people feel worse when their neighbour gets richer), at least for some accounts of wellbeing. For instance, Haushofer, Reisinger, & Shapiro (2019) found that neighbours of GiveDirectly cash recipients experienced a decline in psychological wellbeing (seemingly measured by a similar index to the one used in the most recent study) about half as great as the psychological wellbeing benefit to the recipient. Depending on how many neighbours are affected by each transfer, this would seem to indicate that GiveDirectly may have a net negative effect on aggregate wellbeing. However, this effect was driven entirely by life satisfaction, an ‘evaluative’ or ‘cognitive’ measure; there were no negative spillovers on measures of ‘hedonic’ wellbeing, namely “happiness”, “stress”, and “depression”. As the authors note:
This result is intuitive: the wealth of one’s neighbors may plausibly affect one’s overall assessment of life, but have little effect on how many positive emotional experiences one encounters in everyday life. This result complements existing distinctions between these different facets of well-being, e.g. the finding that hedonic well-being has a “satiation point” in income, whereas evaluative well-being may not (Kahneman and Deaton, 2010).
Without seeing the disaggregated scores from the new study, it seems possible that there were non-trivial and statistically significant harms (or benefits) according to some components of the index. This matters to those with a preferred moral theory or conception of wellbeing, e.g. a classical utilitarian probably cares more about hedonic states than life evaluations, and a prioritarian more about severe states like depression than positive ones like happiness.

Derek Nov 17, 2019, 7:08 PM
1 point
0 ∶ 0
in reply to: toonalfrink’s comment on: Does climate change deserve more attention within EA?
It’s an interesting idea. Often the resource fungibility won’t be huge so it may not make much difference, but in some cases it might.
It also seems to assume that it will use fewer total resources than working on both problems less intensively for a longer period. I would guess that it would usually be more efficient to divide resources and work on problems simultaneously, in part due to diminishing returns to investment. E.g. shifting all AI researchers to climate change would greatly hinder AI research but perhaps not contribute much to climate change mitigation, even assuming good personal fit of researchers, since there are already lots of talented people working on the issue.
But I’ve thought about this for less than 5 minutes so it might deserve a deeper dive. I’m not likely to do it, though.

Derek Nov 17, 2019, 6:01 PM
2 points
0 ∶ 0
on: Concrete next steps for ageing-based welfare measures
This seems like a promising approach to comparing hedonic wellbeing across species of nonhuman animals. Comparing human with nonhuman wellbeing is also important for prioritising interventions, and there doesn’t seem to be a good way of doing this at the moment. Do you think this could be part of a solution?

Derek Oct 13, 2019, 11:39 AM
2 points
0 ∶ 0
in reply to: toonalfrink’s comment on: Does climate change deserve more attention within EA?
“[AI safety] is currently taking up a large amount of attention from competent altruistic people. If the issue were to be solved or its urgency reduced, some of those resources might flow into [climate change mitigation]”
So hurry up, Toon ;)

Derek Aug 26, 2019, 4:57 PM
1 point
0 ∶ 0
in reply to: Derek’s comment on: List of ways in which cost-effectiveness estimates can be misleading
I wish I’d spent more time reviewing this before publication as I failed to mention some key points. I’ll add some of them as comments.

Derek Aug 26, 2019, 4:53 PM
21 points
0 ∶ 0
on: List of ways in which cost-effectiveness estimates can be misleading
Most cost-effectiveness analyses by EA orgs (and other charities) use a ratio of costs to effects, or effects to costs, as the main—or only—outcome metric, e.g. dollars per life saved, or lives affected per dollar. This is a good start, but it can be misleading as it is not usually the most decision-relevant factor.
If the purpose is to inform a decision of whether to carry out a project, it is generally better to present:
(a) The probability that the intervention is cost-effective at a range of thresholds (e.g. there is a 30% chance that it will avert a death for less than my willingness-to-pay of $2,000, 50% at $4,000, 70% at $10,000...). In health economics, this is shown using a cost-effectiveness acceptability curve (CEAC).
(b) The probability that the most cost-effective option has the highest net benefit (a term that is roughly equivalent to ‘net present value’), which can be shown with a cost-effectiveness acceptability frontier (CEAF). It’s a bit hard to get one’s head around, but sometimes the most cost-effective intervention has lower expected value than an alternative, because the distribution of benefits is skewed.
(c) A value of information analysis to assess how much value would be generated by a study to reduce uncertainty. As we found in our evaluation of Donational, sometimes interventions that have a poor cost-effectiveness ratio and a low probability of being cost-effective nevertheless warrant further research; and the same can be true of interventions that look very strong on those metrics.
See Briggs et al. (2012) for a general overview of uncertainty analysis in health economics, Barton et al. (2008) for CEACs, CEAFs and expected value of perfect information, and Wilson (2014) for a practical guide to VOI analyses (including the value of imperfect information gathered from studies).
Of course, these require probabilistic analyses that tend to be more time-consuming and perhaps less transparent than deterministic ones, so simpler models that give a basic cost-effectiveness ratio may sometimes be warranted. But it should always be borne in mind that they will often mislead users as to the best course of action.