Thanks for posting this here as well as Jess’s excellent questions! This seems like a nice place to continue the conversation around the paper, so I’ll respond to what I take to be the most pertinent issues in the blog post here. As Jess notes, this is a relatively early attempt to formulate these ideas and the literature on longtermist institutional reform is extremely young, so the more conversation the better.
How will (short-term) vested interests try to capture these in-government research groups, and how will that be prevented? Why is this better done within the government rather than done in academia using grants from the government or philanthropists?
Most governments are swamped with expertise. It’s not that they have too little of it, but that they are overwhelmed with it, can’t absorb it, and don’t know who to turn to as a reliable source of information. Governments need one or a small body of epistemically reliable and nonpartisan research groups that they can turn to which fill the function of synthesizing extant research into consumable reports for government. These research groups in turn need to have strong working relationships and good lines of communication with government. If an academic or privately-funded research institute could play that role, that would be fine, but it’s harder to see how this would be possible, and in-government research groups and advisory boards have a good track record of playing this sort of role. (We use the OTA as one prominent example, but there are many others on smaller scale.) One additional benefit of research institutes that are set up by government is that when the government is perceived as legitimate, these institutes will also be seen as legitimate and reliable sources of information. It would be valuable for the described research institutes to have public legitimacy, so that if their publicly disseminated research were ignored by government this fact could precipitate public censure.
If public censure isn’t enough to command the attention of government to the research, then a research institute with government authority could also have the “put-it-in-their-face-power” we suggest in the paper, forcing reading and a response by government.
Short-term interest capture is an important worry, and we see this already in privately-funded research groups as well as in academia. One mechanism we propose in the paper for preventing capture by interest groups and industry is to have researchers selected by professional associations or by lot. If the research body is large enough and its key members and leadership are shuffled frequently enough, this should prevent a great deal of corruption. But of course, we are open to other ideas depending on the additional concerns that arise.
What will incentivize the citizen assembly to actually benefit future citizens? Merely because they are “explicitly tasked with the sole mandate”, with no enforcement or feedback?
The citizens’ assembly proposed doesn’t have a strong mechanism for amplifying the concern of assembly members for future people. It is assumed that they already have some interest in doing this, as roughly all people do. The role of the citizens’ assembly isn’t to amplify personal motivation, but rather to i) reduce election and funding incentives that disincentivize the electorate from focusing on the long-term, ii) reduce the deleterious effects of polarization on long-term deliberation, and iii) create designated agenda time for long-term issues. All of these sources of short-termism hamper governmental motivation to focus on the long-term, so we should expect the citizens’ assembly to be much more motivated to benefit future generations than existing government organs. The motivation comes from the citizens themselves, but it has far fewer obstacles to overcome than the motivation of the electorate.
That said, the literature on assemblies does suggest that participation in assemblies decreases citizen political apathy and increases empathy between deliberation participants, so there could be some salutary motivational effects of citizens’ assemblies that we haven’t considered here. Moreover, political decisions tend to operate with 2-5 year timelines, and the assembly members will in general live for much longer than this. Given that the citizens’ assembly will be deliberative and better-informed than the general public, it is possible that it will function more rationally, seeking to promote the diverse interests of the diverse group of people within the assembly across their lifespans, rather than over the next 2-5 years, and this would significantly decrease short-termism. But this is rather speculative, and the central purpose of the assembly is not to increase this kind of motivation.
Does thinking that the citizen assembly would be effective imply that most government assemblies should be selected by sortition (which, right or wrong, has deployed pretty rarely worldwide)? Or is there something about the future and/or soft-power that makes sortition particularly well suited for this body? (Personally, I like sortition as a governing mechanism in general, but if we can’t get hardly anyone to use it generally, why might they here?)
Sortition has perhaps been deployed less rarely than you think! There have been at least 120 citizens’ assemblies and citizen juries deployed worldwide, and sortition is regularly used for the selection of court juries. But it’s true that they’ve rarely been used for the selection of long-lasting government positions.
The role of the citizens’ assembly I mentioned above, I think, shows why sortition should be especially helpful here: it removes perverse election incentives to attend to the short-term, and it also reduces the effect of partisan forces, decreasing polarization. These seem especially important when considering long-term issues where our situation is epistemically precarious, but you’re right to point out that they are generally very important. I am personally quite open to the idea that a very large proportion of political leaders should be selected randomly. My own dissertation supervisor, Alex Guerrero, is writing an excellent book defending this idea at this very moment.
On why we might be able to get government to use it here: citizens’ assemblies have a relatively strong tradition of use for gathering information on the informed views of citizens, and have in the last decade become increasingly popular. As above, I would advocate for greater experimentation with sortition, but they have most popularly been used in citizens’ assemblies that are similar to that which we describe, and we expect it to continue to be popular in these institutions.
Will prosperity impact statements obviously improve the long-term future more than it will be used to block/delay projects for near-term reasons? Certainly, environmental impact statements suffer from this problem, and EIS have the advantage that at least there is often some way to objectively check whether they were right or wrong in a reasonable amount of time.
This is the issue raised in the blog post that I find trickiest. It’s certainly true that EIAs have frequently been used to block and delay projects on spurious grounds, and the point here that PIAs are less epistemically tractable is spot-on and important. One advantage of PIAs in the legislature is that many more resources can be put to ensuring that they are objective and accurate than can be put into, say, local jurisdictions, given the much greater resources of the federal government and the fewer number of items requiring assessment. An idea we considered but didn’t include here is that an independent, non-partisan body such as the in-government research institutions we defend could perform the impact assessments, taking them out of the hands of politicians who might use them for more obstructionist ends. But I remain quite uncertain on the best mechanism for ensuring that PIAs fulfill their information-gathering and soft censure functions rather than becoming used primarily to fuel partisan obstructionism, and I’d certainly be interested in other ideas.
Ah, it looks like I read your post to be a bit more committal than you meant it to be! Thanks for your reply! And sorry for the misnomer, I’ll correct that in the top-level comment.
I’m glad to see CRS take something of an interest in this topic and I’m particularly happy to see some meta-level discussion of representing the interests of future generations which has been sorely missing from the longtermism space.
We are in full agreement that most extant proposals to represent future generations involve very weak institutions and often rely on tenuous political commitments. In fact, it’s because political commitments are so tenuous that political institutions to represent future generations must at first be weak. Strong institutions for future generations have historically been repealed very rapidly, as Jones, O’Brien, and Ryan (2018) have argued from a couple case studies.
We are also in full agreement that there are problems of predicting the interests of future generations, and that getting more objective information about their interests is a key problem. This problem proliferates with increasingly longer timescales. This is why many of the solutions I am personally most favorable to are information interventions, such as creating research bodies like the now-defunct Office of Technology Assessment, which can distill and package extant expertise for legislative bodies, as well as posterity impact assessments, which can create strong incentives to gather more information about the future.
I find much less compelling the idea that “if there is the political will to seriously consider future generations, it’s unnecessary to set up additional institutions to do so,” and “if people do not care about the long-term future,” they would not agree to such measures. The main reason I find this uncompelling is just that it overgenerates in very implausible ways. Why should women have the vote? Why should discrimination be illegal?
The main long-term function that I see longtermist institutional reform, or any other kind of institutional reform playing is an institutional signalling role. There is compelling evidence that legal and political reform significantly shifts the norms and attitudes that people come to see as acceptable (Berkowitz and Walker 1967, Bilz and Nadler 2009, Flores and Barclay 2015, Tankard and Paluck 2016, 2017, Walker and Argyle 1964). Shifting laws and institutional norms credibly signals information about group attitudes to anyone who has access to information about those laws and norms. In this case, it signals that good, sensible, right-thinking people think that future generations are of great importance and that our political systems must be responsive to their interests. For this reason, there is a chicken and egg problem for institutional reform, but this chicken and egg problem is very friendly to supporters of institutional reform. Reforming institutions changes attitudes, which in turn creates the political will necessary to reform institutions further. Reformed institutions in turn create stable shelling points that prevent value drift away from core values.
For this reason, longtermist institutional reform is quite beneficial for information-gathering purposes. Representing future generations creates greater political and cultural will to gather objective information about the interests of future generations. It’s an exercise in movement-building.
I don’t know if you meant to narrow in on only those reforms I mention which attempt to create literal representation of future generations or if you meant to bring into focus all attempts to ameliorate political short-termism. In the latter case, it’s worth noting that there are a large variety of likely causes of short-termism. Some of them are epistemic (we don’t know what to do) and motivational (we lack the political will), but others are merely institutional. In these latter cases, the problem is not that we don’t have enough information or will, but rather that the right information is not getting to the right people or that institutional mechanisms are preventing appropriately-motivated and informed actors from acting for the long term. These sorts of problems sometimes require different fixes, and they can sometimes be fixed simply by creating designated stakeholders who create relevant coordination points in government and have time allocated explicitly to considering the long-term. Political problems are often a problem of institutional incentives rather than of political will, and there are currently very strong incentives to focus on the short-term. I canvass many of the various causes of political short-termism in my (now rather lengthy) review on longtermist institutional design and policy.
As a classical utilitarian, I’m also not particularly bothered by the philosophical problems you set out above, but some of these problems are the subject of my dissertation and I hope that I have some solutions for you soon.
In short, I think there is reason for more optimism about longtermist institutional reform than you express here, but I am happy to have some further discussion of the problem and to see a call to consider more seriously the epistemic problems that plague such reform along with some possible solutions.
Thanks for clarifying all of this! Given that most questions are optional I no longer have this concern, and I’m glad that you’ve clarified this on the application.
Much looking forward to seeing you there as well!
Thanks so much to those involved in organizing! I wanted to share that I found the registration process (with its 40 or so questions, many requiring detailed information) quite onerous and I can imagine that it might deter some people from submitting completed applications. While this might sometimes be useful for a physical conference, to ration spots in part on the basis of the amount of effort put in, I can’t as easily see how it would be useful for a virtual conference. But I may simply be insufficiently creative!
See also my 2018 EAG talk on shaping the long-term future through antispeciest legislative initiatives. Most of the relevant discussion starts at 8:40.
While I at the time thought the dominant beneficial effect would be through AGI alignment, I now think that we should think of these interventions as improving the value alignment of humanity and our descendents in general.
And cf. my and Jeff Sebo’s paper on the indirect effects of eating meat and farming animals on human moral psychology and its importance for consequentialists:
In general, I’m with Michael in thinking that we should expect the dominant beneficial effects of vegetarianism and abolitionist efforts against animal agriculture to be their effects on human morality, which can positively shape the long-term future by better aligning the values of our descendents (and therefore their behavior) with our own values.
Thanks! I appreciate your wariness of overemphasizing precise numbers and I agree that it is important to hedge your estimates in this way.
However, none of the claims in the bullet you cite give us any indication of the expected value of each intervention. For two interventions A and B, all of the following is consistent with the expected value of A being astronomically higher than the expected value of B:
B is better than A in most of the most plausible scenarios
On most models the difference in cost-effectiveness is small (within 1 or 2 orders of magnitude)
One could reasonably believe that B is better than A or that B is better than A
Extremely little information is communicated about the relative expected value of A and B by the above points, and what information is communicated misleadingly suggests that both interventions are quite close in expected value. Because EAs are concerned with the expected value of interventions, I think you ought to communicate more about the relative expected value of the interventions and frame your summary of the interventions in a way that is less likely to mislead people about the relative expected value of each intervention.
I think the ideally informative way to both communicate the relative expected value of the interventions and hedge on your model uncertainty in the summary is to (1) provide your expected value estimate, (2) explain that you have high model uncertainty and one could arrive at a different expected value estimate with different assumptions, and (3) invite participants to adjust the Guesstimate and generate their own predictions.
Thanks for doing this! Though it seems like you kinda buried the lede. Why isn’t this in the top level summary?
In expectation, THL is >100x better than AMF
In the median scenario, THL is about 2-4x more cost-effective than AMF
A 71% chance that THL is more cost-effective than AMF
On the topic of the outlier age group:
“If it really is the case that the 55 to 64 year old age group is an outlier as the more present-day-centric group, it suggests that a simple “rational” explanation (“why care about the future when I’ll be dead soon anyway”) might not be the best explanation. Other socio-cultural factors may be at play.”
I can see two decent explanations for why the 55 to 64 age group would have less longtermist values than either adjacent age cohort.
The first is cohort effects. As the Pew Research Center points out, there is no simple relationship between age and political ideology. While voters tend to become more conservative as they age (along with other effects on time preference, etc), their ideological identity is also greatly affected by the administration under which they come of age politically.
The second is the findings from Ahlfeldt et al. and others that 1) while voters become more conservative as they age, they become rapidly more conservative around retirement age, and 2) the very oldest people seem to experience some “end of life altruism” because they have very weak self-interested reasons (due to so little time remaining) and so their self-interested reasons are dominated by ego-transcending values such as altruism. (See especially the provocative graphs on p. 15 of Ahlfeldt et al.)
If either of these explanations is true, then it could be that the “rational” explanation is empirically adequate, but there are other effects in play as well.
More on the question of what best explains these trends:
Ahlfeldt et al. analyze 305 Swiss referenda and argue that aging effects swing free from cohort effects and status quo habituation effects. “The evidence, instead, suggests that voters make deliberate choices that maximize their expected utility conditional on their stage in the lifecycle.”
I think these trends are not better-explained by the hypothesis that older people are more conservative.
1. In the study, older voters were more likely to support health spending on risks to elderly health and less likely to support health care cost cuts, and less likely to support education spending, public transportation and infrastructure spending, and job creation. They were also neutral on the creation of sports facilities.
While I unfortunately haven’t been able to look at the 82 referenda to examine their specific content, on its face this looks less like a division on party lines and more like a division on lines of generational self-interest.
2. The authors report that “[W]e find that controlling for party affiliation (conservatives and greens) and region (Baden vs. Württemberg) reduces the age effect by about one-third (Table 5, columns 3 and 4).”
3. The fact that older people are more conservative itself requires explanation. Part of the explanation is plausibly that conservative ideology and political parties cater to the self-interest of older people. How much can be explained this way I cannot say.
This is helpful indeed. Thanks for the reply!
1. Good point on clarifying the timescale for the sake of the report. I think the timescale you define for the UK is about right for narrowing the scope of the institutions considered by the report. Then the “effectiveness” evaluation criterion can do the work of identifying which institutions are best by longtermist lights, ranking institutions cardinally as a function of, among other things, their temporal reach.
2. You did previously share your list with me and I’m glad you’ve reshared it here. Ideas you mention here did not end up on the list I shared to the EA Forum for one of a few reasons: either there exists a similar proposal in the document already or the suggested change is in my list of smaller, incremental changes or I excluded it because I wanted to prioritize concrete, particular proposals over abstract, general ideas. Some of them simply involve ideas that are still on my to-read list. All of your suggestions are included in a more complete list off-site.
3. Max Stauffer also recommended adding a criterion based on strength of evidence. I think this is a good idea. I also like your suggestion to broaden my “political feasibility” criterion to “overall implementability.” As you imply, there are considerations beyond political feasibility that are relevant to a design’s implementability. I’m incompletely convinced that symbolism should be ignored completely in the context of this report, but I have been convinced by your point that symbolic value depends on contextual interaction with a lot of things, and an otherwise uninspiring change can function as a symbol with the right packaging.
Thanks again for reaching out here and via email. I’ll be in touch about collaboration in just a moment.
Edit: Upon revisiting I realized that I had already read this paper. It’s one of the more useful things I’ve read in this area, so good nod.
Thanks! I’ve spoken to the APPG and seen some of their policy statements but I had not seen this particular paper. Super helpful.
It’s worth noting that one important assumption here is that experts are pretty good at determining the counterfactual value of past policy decisions. I think this is right, but if we gave it up then no system like this one would be effective, since the feedback from future generations would be near-random. On the other hand, if the assumption is correct then there should be some feasible system that provides useful intergenerational feedback of the kind described here, though it may need to include a mechanism for increasing the influence of experts in the decision process.
On (1), I’m not currently considering any existing institutions, other than existing variants of the proposals mentioned. You’re right that it would be useful to know which institutions we should preserve, and there also might be other things to learn from analyzing these institutions, such as what has worked well about them and what has kept them from working better. I’ll have to consider adding these sorts of institutions.
On (2), that’s definitely of concern to me in light of the fact that so many recently-adopted future-focused institutions have not been able to survive even one election cycle. I’ve been including this (the permanence of the institution) under effectiveness, but maybe it’s worth graining the categories a bit more finely.
I agree it will probably not change voter epistemic behavior. The thought was that it would change the epistemic behavior of the parties catering to voters and the representatives acting on behalf of the voters, since the voting rule will select for parties and representatives which are less short-termist. This of course can’t be guaranteed—if parties are not motivationally longtermist but are merely trying to appease voters to hold power, for example, it won’t change their epistemic incentives very much unless competing actors (parties, media) can demonstrate to young people that their plans are bad. But even in this case this is plausible.
Thanks, I’ve looked at some of the inclusive wealth and natural capital accounting stuff a little bit and will continue to do so. Do you currently have any sense how useful this sort of accounting will be for general future generations issues (incl. catastrophic risks, positive moral & economic trajectories) beyond concerns related to environmental degradation?
I am extremely interested in the question of how religions transmit ideas and values across many generations, but at the current moment I have no idea how they do this so successfully. If anyone has ideas or empirical sources on this I’d be quite keen to get more info on this.
Surprising (and confusing!) as it may be, there is some evidence that voters would vote differently with their Demeny vote than with their first vote.
I’ve asked Ben Grodeck (who clued me into Demeny voting) to weigh in with more data, but for now see this study from Japanese economist Reiko Aoki, who found (Table 8 and Figure 7) that the voting preferences of surveyed participants who are permitted to cast one vote on behalf of themselves and one vote for their child sometimes vote differently on their second vote. The effect isn’t drastic, but it is certainly non-trivial.
The study authors further find that policy preferences on behalf of oneself and on behalf of one’s children diverge to a greater degree, and the authors hypothesize that we would see more divergence between the multiple votes of Demeny voters if they had different political options that better reflected the divergence between these sets of preferences. Thus, they think that instituting Demeny voting would cause party platforms to change to try to cater to the policy preferences of parents voting on behalf of their children.
Excellent. This is a much better idea than the “allow the 2119 people to decide whether to sentence the grandchildren of the 2019 political leaders to the tribunal of death” feedback mechanism that, disturbingly, came to me more readily.
It would be interesting to think about whether there are other feasible ways to see to it that the decisions of future people provide an incentive for the actions for present people.
Two concerns I have with this general kind of scheme is that it requires citizens to have lots of faith that the relevant institutions and the policy will persevere 100 years into the future (the 100 year bond stuff is relevant to this) and that they might not play well with high rates of immigration (since fluidity in polity membership could undermine the efficacy of long-term feedback mechanisms for members of that polity). But these might just be details to be ironed out rather than insolvable problems with the design.