I’m having trouble interpreting the first graph. It looks like 600 people put poverty as the top cause, which you state is 41% of respondents, and that 500 people put cause prioritisation, which you state is 19% of respondents.
The article in general seems to put quite a bit of emphasis on the fact that poverty came out as the most favoured cause. Yet while 600 people said it was the top cause, according to the graph around 800 people said that long run future was the top cause (AI + non-AI far future). It seems plausible to disaggregate AI and non-AI long run future, but at least as plausible to aggregate them (given the aggregation of health / education / economic interventions in poverty), and conclude that most EAs think the top cause is improving the long-run future. Although you might have been allowing people to pick multiple answers, and found that most people who picked poverty picked only that, and most who picked AI / non-AI FF picked both?
The following statement appears to me rather loaded: “For years, the San Francisco Bay area has been known anecdotally as a hotbed of support for artificial intelligence as a cause area. Interesting to note would be the concentration of EA-aligned organizations in the area, and the potential ramifications of these organizations being located in a locale heavily favoring a cause area outlier.” The term ‘outlier’ seems false according to the stats you cite (over 40% of respondents outside the Bay thinking AI is a top or near top cause), and particularly misleading given the differences made here by choices of aggregation. (Ie. that you could frame it as ’most EAs in general think that long-run future causes are most important; this effect is a bit stronger in the Bay)
I’m having trouble interpreting the first graph. It looks like 600 people put poverty as the top cause, which you state is 41% of respondents, and that 500 people put cause prioritisation, which you state is 19% of respondents.
I can understand why you’re having trouble interpreting the first graph, because it is wrong. It looks like in my haste to correct the truncated margin problem, I accidentally put a graph for “near top priority” instead of “top priority”. I will get this fixed as soon as possible. Sorry. :(
We will have to re-explore the aggregation and disaggregation with an updated graph. With 237 people saying AI is the top priority and 150 people saying non-AI far future is the top priority versus 601 saying global poverty is the top priority, global poverty still wins. Sorry again for the confusion.
-
The term ‘outlier’ seems false according to the stats you cite
The term “outlier” here is meant in the sense of a statistically significant outlier, as in it is statistically significantly more in favor of AI than all other areas. 62% of people in the Bay think AI is the top priority or near the top priorities compared to 44% of people elsewhere (p < 0.00001), so it is a difference of a majority versus non-majority as well. I think this framing makes more sense when the above graph issue is corrected—sorry.
Looking at it another way, The Bay contains 3.7% of all EAs in this survey, but 9.6% of all EAs in the survey who think AI is the top priority.
The claim you’re defending is that the Bay is an outlier in terms of the percentage of people who think AI is the top priority. But what the paragraph I quoted says is ‘favoring a cause area outlier’ - so ‘outlier’ is picking out AI amongst causes people think are important. Saying that the Bay favours AI which is an outlier amongst causes people favour is a stronger claim than saying that the Bay is an outlier in how much it favours AI. The data seems to support the latter but not the former.
Hey Michelle, I authored that particular part and I think what you’ve said is a fair point. As you said, the point was to identify the Bay as an outlier in terms of the amount of support for AI, not declare AI as an outlier as a cause area.
The article in general seems to put quite a bit of emphasis on the fact that poverty came out as the most favoured cause.
I don’t know that this is necessarily true beyond reporting what is actually there. When poverty is favored by more than double the number of people who favor the next most popular cause area (graph #1), favored by more people than a handful of other causes combined, and disliked the least, those facts need to be put into perspective.
If anything, I’d say we put a fair amount of emphasis on how EAs are coming around on AI, and how resistance toward putting resources toward AI has dropped significantly.
We could speculate about how future-oriented certain cause areas may be, and how to aggregate or disaggregate them in future surveys. We’ve made a note to consider that for 2018.
I don’t know that this is necessarily true beyond reporting what is actually there. When poverty is favored by more than double the number of people who favor the next most popular cause area (graph #1), favored by more people than a handful of other causes combined, and disliked the least, those facts need to be put into perspective.
I agree—my comment was in the context of the false graph; given the true one, the emphasis on poverty seems warranted.
I’m having trouble interpreting the first graph. It looks like 600 people put poverty as the top cause, which you state is 41% of respondents, and that 500 people put cause prioritisation, which you state is 19% of respondents.
The article in general seems to put quite a bit of emphasis on the fact that poverty came out as the most favoured cause. Yet while 600 people said it was the top cause, according to the graph around 800 people said that long run future was the top cause (AI + non-AI far future). It seems plausible to disaggregate AI and non-AI long run future, but at least as plausible to aggregate them (given the aggregation of health / education / economic interventions in poverty), and conclude that most EAs think the top cause is improving the long-run future. Although you might have been allowing people to pick multiple answers, and found that most people who picked poverty picked only that, and most who picked AI / non-AI FF picked both?
The following statement appears to me rather loaded: “For years, the San Francisco Bay area has been known anecdotally as a hotbed of support for artificial intelligence as a cause area. Interesting to note would be the concentration of EA-aligned organizations in the area, and the potential ramifications of these organizations being located in a locale heavily favoring a cause area outlier.” The term ‘outlier’ seems false according to the stats you cite (over 40% of respondents outside the Bay thinking AI is a top or near top cause), and particularly misleading given the differences made here by choices of aggregation. (Ie. that you could frame it as ’most EAs in general think that long-run future causes are most important; this effect is a bit stronger in the Bay)
Writing on my own behalf, not my employer’s.
09/05/17 Update: Graph 1 (top priority) has been updated again
I can understand why you’re having trouble interpreting the first graph, because it is wrong. It looks like in my haste to correct the truncated margin problem, I accidentally put a graph for “near top priority” instead of “top priority”. I will get this fixed as soon as possible. Sorry. :(
We will have to re-explore the aggregation and disaggregation with an updated graph. With 237 people saying AI is the top priority and 150 people saying non-AI far future is the top priority versus 601 saying global poverty is the top priority, global poverty still wins. Sorry again for the confusion.
-
The term “outlier” here is meant in the sense of a statistically significant outlier, as in it is statistically significantly more in favor of AI than all other areas. 62% of people in the Bay think AI is the top priority or near the top priorities compared to 44% of people elsewhere (p < 0.00001), so it is a difference of a majority versus non-majority as well. I think this framing makes more sense when the above graph issue is corrected—sorry.
Looking at it another way, The Bay contains 3.7% of all EAs in this survey, but 9.6% of all EAs in the survey who think AI is the top priority.
Thanks for clarifying.
The claim you’re defending is that the Bay is an outlier in terms of the percentage of people who think AI is the top priority. But what the paragraph I quoted says is ‘favoring a cause area outlier’ - so ‘outlier’ is picking out AI amongst causes people think are important. Saying that the Bay favours AI which is an outlier amongst causes people favour is a stronger claim than saying that the Bay is an outlier in how much it favours AI. The data seems to support the latter but not the former.
I’ve also updated the relevant passage to reflect the Bay Area as an outlier in terms of support for AI, not AI an outlier as a cause area
Hey Michelle, I authored that particular part and I think what you’ve said is a fair point. As you said, the point was to identify the Bay as an outlier in terms of the amount of support for AI, not declare AI as an outlier as a cause area.
I don’t know that this is necessarily true beyond reporting what is actually there. When poverty is favored by more than double the number of people who favor the next most popular cause area (graph #1), favored by more people than a handful of other causes combined, and disliked the least, those facts need to be put into perspective.
If anything, I’d say we put a fair amount of emphasis on how EAs are coming around on AI, and how resistance toward putting resources toward AI has dropped significantly.
We could speculate about how future-oriented certain cause areas may be, and how to aggregate or disaggregate them in future surveys. We’ve made a note to consider that for 2018.
Thanks Tee.
I agree—my comment was in the context of the false graph; given the true one, the emphasis on poverty seems warranted.