Model 1
Longtermism summary
In the course of running a series of studies investigating public attitudes towards longtermism and how longtermism might be best promoted for an org a couple of years ago, RP developed a number of different measures of attitudes related to longtermism and gathered some pilot data with them.
This document describes the results of that pilot. We would be interested in running additional studies which build on and extend these results using these measures in the future. However, we’d be interested in getting feedback on the pilot results to get a sense of what would be useful.
The summary is deliberately non-technical, though we are happy to provide more technical details on request.
Measures
After starting with a broader pool of items which we thought would measure different attitudes relevant to longtermism, we used a combination of exploratory and confirmatory factor analysis to identify two measures of two distinct constructs related to longtermism.
The first factor, presentism appears to measure something like the inverse of longtermism (the view that people who are alive now, rather than those in the future).1
The second factor, pessimistic ability, concerns a pessimistic view of our ability to influence the far future. You might think of this latter factor as concerning a belief that there are practical barriers to helping the long-term future.
It’s worth noting that both these measures are intended to capture general underlying attitudes related to longtermism, rather than capture agreement or disagreement with an explicitly stated theoretical characterisation of longtermism (for that see below).
Scales
Presentism
1: We should focus on helping people now rather than those who are not born yet
3: People who actually exist are more important than people who might come to exist in the future
4: We should care more about people who are alive today than people who might be born in the future
14: We should care more about people who are suffering now than people who might suffer hundreds of years from now
24: The government should focus on helping people alive now and their children, not people in the distant future
Pessimistic ability
6: There is nothing much we can do to influence the distant future, so we should not try
7: It would be a waste of resources to try to influence things thousands of years away
15: No-one knows what will happen hundreds of years in the future so it doesn’t make sense to try to influence it
16: If we try to influence what happens thousands of years in the future we will likely just make things worse
Explicit arguments
We also presented respondents with some longer more explicit statements of different theoretical positions related to longtermism, one ‘presentist argument’ and one ‘longtermist’ argument.2 While longer, these are of course still short, non-comprehensive descriptions. Future work could explore attitudes towards different statements of these positions to observe how they vary.
Presentist argument
Some people argue that we have a special obligation to help people who are alive and in need today, as opposed to people who have not been born yet but may come to live in the future. They argue that people are suffering terrible pain and injustices today and that it should be our number one priority to help them. They argue that it would be heartless to turn a blind eye to people who need our help today in order to try to make things better for people who might live in the distant future, but don’t actually exist yet.
Longtermist argument
Some people argue that humanity could continue to exist for many thousands or millions of years. If so, then the number of people alive in the distant future many years from now will likely be much larger than the number people alive today. Moreover, the distant future might be much worse or much better than the world today. As a result, they argue that acting to influence how well the future thousands or millions of years from now goes is much more important than anything else.
On average, respondents supported the explicit presentist argument, while responses towards the explicit longtermist argument centred around the midpoint. We suspect this pattern (rather than respondents’ agreement with the presentist argument being matched by disagreement with the longtermist argument) may reflect respondents being unsure what to make of the explicit longtermist argument. Given this, we focus on the explicit presentist argument below.
Presentist argument, presentism and pessimistic ability
We find the explicit presentist argument is strongly associated with the presentist scale, while having little association with the pessimistic ability scale, which offers evidence of convergent and divergent validity for these measures.
Presentism, Pessimistic Ability and Presentist Argument
Longtermist preference
We also asked respondents about their preferences regarding a series of practical policy choices, intended to represent more or less longtermist preferences.3 These measures are somewhat noisier than the attitude measures (especially the first item4), though collectively seem to measure a longtermist preference passably well, with responses to these being strongly associated with responses to the attitudes scales.
Please indicate which of the following options you most prefer:
Preference 1: (1) The government does not raise taxes - (7) The government raise taxes to spend more on reducing threats that may destroy humanity
Preference 2: (1) The government spends more on helping people today (e.g. in schools, hospitals) - (7) The government spends more on reducing threats that may destroy humanity
Preference 3: (1) The government spends more on research to help people alive today - (7) The government spends more on research that may improve the world hundreds of years from now
Preference 4: (1) Working to reduce present day injustices - (7) Working to make the long term future better
Longtermist preference model
Policy preferences
Longtermist preference, presentism and pessimistic ability
Our longtermist preference, presentism and pessimistic ability measures are all strongly associated with each other.
Longtermist preference, Presentism, Pessimistic ability
Longtermist preference predicted by presentism and pessimism
Both presentism and a belief that our ability to influence the future is limited seem important for predicting longtermist preference.
Demographics
We also examined associations between the measures and basic demographics. Due to the low sample size of our initial survey, many of these estimates are quite imprecise. We may examine these, in addition to associations with other measures (e.g. attitudes towards risk, numeracy, field of employemtn etc.) in subsequent larger surveys.
Gender
Male respondents gave significantly higher ratings of the explicitly longtermist argument and lower ratings of the presentist arguments. For the other measures, differences were non-significant, though in the same direction.
Education
The education measure corresponds to (1) No GCSE, O-level of equivalent, (2) GCSE, O-Level or equivalent, (3) A-level or equivalent, (4) BA degree or equivalent, (5) MA or equivalent, (6) PhD or equivalent. There were few respondents outside of levels 3-5, so the other estimates are very imprecise.
There were few consistent effects, though pessimistic ability appeared to decrease with higher education levels.
Political orientation (left to right)
There were largely no clear linear associations with the left-right political orientation scale. However, as noted, the estimates are quite imprecise, due to low sample size (particularly in the most right-wing categories), so we would want to explore these further in a larger sample.
Age
Presentism and pessimistic ability appear to weakly increase with age, while longtermist preference weakly declines with age. Support for the longtermist argument is highest among the youngest respondent, while support for the presentist argument appears highest among older respondents.
Footnotes
Of course, this is likely somewhat related to person-affecting views. However, we did not develop a distinct measure which would allow us to disentangle person-affecting motivations from other modes of presentism. (We did look into developing a measure of person-affecting views specifically, in an earlier project but with little success. People seemed to find it very difficult to understand what was being asked).↩︎
Responses were measured on a (1) Strongly disagree - (7) Strongly agree response scale.↩︎
One might wonder why these are somewhat abstractly stated choices, rather than more specific policies that longtermists might be interested in e.g. a pause on AI development. This is because this measure was designed to capture a general longtermist preference. In future work, it would be interesting to see to what extent a general longtermist preference, alongside the other measures, predicts support for particular policies (which one might expect would additionally be influenced by other factors particular to the policy in question).↩︎
This is likely due to this first item tracking attitudes pro or con raising taxes, specifically, to a greater extent than the other items.↩︎