- Humans are hardwired to desire and compete for social status. Status games are not designed to maximise wellbeing and often have bad outcomes. This creates and exacerbates myriad problems from conspicuous consumption to genocide.
- This problem is extremely neglected. No one seems to be working on it directly.
- Relatively speaking, systemically tackling this problem is not very tractable. However, there is a lot of uncertainty due to neglect.
We are constantly competing for respect, recognition, acceptance and prestige.  From an evolutionary fitness standpoint, this is great. From a moral standpoint, this is less great. Status hierarchies and games were not designed with concern for morality. The aims and behaviours they encourage often have bad consequences. These can take a number of forms, including interpersonal violence, group conflict, social-comparison-based misery, and meaningless consumption and effort.
The following are some examples of these bad outcomes:
(I am not claiming that status is the only thing at play here, just that it is crucial to the existence/severity of these outcomes)
- Nuclear war
- The Precipice opens with a recounting of a nuclear incident at the height of the Cuban Missile Crisis, in which a Russian submarine captain says: “Maybe the war has already started up there, while we are doing somersaults here. We’re going to blast them now! We will die, but we will sink them all—we will not disgrace our Navy!”
- Putin recently made nuclear threats. His invasion of Ukraine might be primarily about status.
- Genocide seems to be driven by perceived status threats to a high-status group and/or attempts on behalf of low-status groups to rise in status. This might seem preposterous; see Appendix B for a relevant quote.
- The Easterlin Paradox
- The “Keeping Up With The Joneses” dynamic is a leading explanation for this.
- Meat consumption
- Meat can be a status symbol for those with low socioeconomic status. My intuition is that as the world grows richer, more people become able to afford this symbol. That’s not great for animal welfare.
- AI capabilities research
- AI capabilities research seems to pay very well. The non-EA AI safety people I’ve met tend to think AI safety is something of a joke. Neither of these trends awards AI safety very many status points. I don’t know how much AI safety pays, but those are reasons to hope it chumps the salaries of capabilities researchers.
- Malevolent actors
- Driving trucks
- Perhaps the second-highest GHG emitting status symbol on this list.
This list is far from exhaustive, but should be enough to convey that unaligned status games come in all shapes and sizes. To create some order, I’ve classified them as follows. High/low stakes refers to how drastic the potential for status change is. 
|High stakes + Group|
|High stakes + Individual|
|Low stakes + group/individual|
The importance of this issue is difficult to quantify because of its enormity. My BOTEC suggests it’s big enough that even scratching its surface could yield 1000x returns (see Appendix A).
Focusing on the sub-issue of status-based violence, I estimate the direct cost of violent status plays to be roughly 11 million DALYs per year, such that spending 11 million dollars to yield a 1% reduction in this problem would be a 1000x return (see Appendix A).
Of course, this is not counting the cost of electing demagogues, directly escalating x-risks, misery from social comparison, etc.
Most of my uncertainty lies in this area. This issue might be too large for real progress to be made. I may have misconceptualised it, and it might be better broken down into smaller components or different categories. That said, because there is no organised effort to directly tackle this issue (to the best of my knowledge), there may be some undiscovered low hanging fruit waiting to be plucked by a prodigious researcher  or Focused Research Organisation.
I see two approaches to tackling this issue: steering and stabilising.
|Steering||Changing the aims of a status game. Wherever a status game revolves around a particularly harmful idol, swap it for something else Indiana Jones Style. Most applicable in low stakes cases.|
Making AI safety jobs more prestigious than capabilities research jobs so that the Paul Christianos of the capabilities research world flock to safety research.
Making alternative meat the culinary equivalent of a Gucci bag. (This would also make its price less of an issue.)
Selectively inflating salaries.
Mimetic theory research: mimetic theory suggests that we don’t actually know what to want, so we imitate others’ desires. A better understanding of how true this is and if/how we can influence which desires are mimicked would be useful.
|Stabilising||Making sure status hierarchies don’t become too unequal and individuals don’t subsist at extremely low status levels. These seem to lead to resentment and violence. Most applicable in high stakes cases.|
Having militaries avoid actions that could humiliate civilians and enemies when intervening overseas.
Cognitive Behavioural Therapy (CBT) seems promising for reducing small-scale violent events. Useful for reducing feelings of envy from social comparison.
Gratitude practice is also effective in reducing social comparison and the associated misery.
Speculative: research and social programs to spot and aid dangerously low status people and/or on reducing status inequality across society.
As far as I can tell, no one is working on it directly as a systemic issue.
Luckily, there are efforts to tackle some of its instantiations:
- Positive psychology research
- Research on humiliation, social status, and violence
- Status-boosting social programs for veterans
- Mental health services
- Drug companies
- This ad
Some manifestations of unaligned status games are not at all neglected. For example, many are working indirectly on status stabilising for individuals. About 570,000 mental health professionals work in the US. Extrapolating globally, I’d estimate the total number of people working on this issue across the globe to be on the order of 10^6. The depression treatment industry is valued at around 16 billion dollars. However, depression is not always related to status, so this connection is tenuous. All in all, this problem still seems very neglected.
In cultures placing strong emphasis on honour, status disparities mediate murder rates such that murder decreases by 0.41 for every 1 unit decrease in status inequality (measured by a Gini Index). Hence I assume that status concerns are a crux in 40% of instances of violence (that is, if the perpetrators never felt the need to defend their position in a social hierarchy, 4/10 cases of violence would be avoided).
Interpersonal violence costs about 27 million DALYs per year.
Terrorism costs about 1.5 million DALYs per year.
(Terrorism and conflict together, as defined by Our World In Data, cost 6.2 million DALYs per year, killing 0.81 per 100,000 globally. 24,000 of the associated deaths are from terrorism, which evens out to about 1.5 million DALYs.)
0.41*(26.9+1.56) = 11.6 million DALYs
Open Phil’s valuation of a DALY: $100,000, so averting 1 DALY lost for 100 dollars is a 1000x return.
If you could reduce this DALY cost by 1% at $100/DALY, it would cost 11 million dollars. So with a bar of 1000x you should be willing to spend 11 million dollars to make a 1% dent in this issue.
“Genocides can happen when a high-status group,
‘experiences a decline in or threat to its status’ or a low-
status group ‘rises or attempts to rise in status’. It’s the
reduction in rank between them that helps generate much
of the horrible madness. Toxic morality is deeply implicated
in these episodes: ‘genocide is highly moralistic’. Genocides
are dominance-virtue games, carried out in the name of
justice and fairness and the restoration of the correct order.
They’re not about the mere killing or ‘cleansing’ of foes,
they’re about healing the perpetrators’ wounded grandiosity
with grotesque, therapeutic performances of dominance and
humiliation. They’re often notable for the killers’ ‘outright zeal in
humiliating’ their victims: during the Armenian genocide,
Turkish gendarmes played a game of tossing people from
horses onto swords sticking upwards from the ground;
during the Gujarat genocide, Hindus pulled the beards of
Muslims, defecated on the Koran, paraded them naked and
with their fingers cut off, forced them to shout ‘Praise Lord
Ram’ and played cricket with their decapitated heads;
during the Rwandan genocide, the high-status Tutsi were
symbolically ‘cut down to size’ by having their Achilles
tendons severed, forcing them to crawl before being killed.
Hutu rapists told their victims: ‘You Tutsi women think you
are too good for us’; ‘You Tutsi girls are too proud’;
‘Remember the past months when you were proud of
yourselves and didn’t look at us because you felt we were
lower than you? Now that will never happen again.”
—Storr, The Status Game, p.235-236.
Note: I wrote this in a very short timespan. The topic is very broad, and this overview is much too shallow to do it justice. The issue’s scope also makes it difficult to quantify neglectedness and tractability, and those sections are severely underdeveloped. I figured it was worth publishing nevertheless. If anything, I hope it made for an interesting read.
Many thanks to Jennifer Zhu and Navya Sheth for their very helpful feedback.
Storr, The Status Game.
Storr, The Status Game, p.233-237
Wikipedia definition: “The paradox states that at a point in time happiness varies directly with income both among and within nations, but over time happiness does not trend upward as income continues to grow”
Frank, Choosing the Right Pond
This anecdote is (1) based on only a small number of experiences and (2) an anecdote, so take it with a large chunk of salt.
I’m not saying AI safety salaries should necessarily dwarf capabilities salaries, just that status should be a major consideration. Perhaps other things like optics outweigh this benefit.
Playing status games poorly could be especially costly for EA in politics. Sam Bankman-Fried said he’d be willing to spend over $100 million to prevent Trump from winning again in 2024. He already spent $10 million on Carrick Flynn, who lost in part by failing to capture the Latino vote. If Flynn had pulled off something similar to the Trump dynamic discussed above, I wonder if things would have been different.
Srinivasan, The Right to Sex, p.84
Storr, The Status Game, p.78
Storr, The Status Game, Ch. 8. Malevolent actors also pose a significant s-risk threat.
The distinctions I chose as the basis of this classification are somewhat arbitrary. You could easily divide the problems along other relevant lines. The lines I drew are also not clear-cut; for example, you could argue that truck driving is very fundamental to our sense of status, but this importance is obscured by the relative tameness of its consequences.
Originally, the idea of trying to steer and stabilise status hierarchies reminded me in its intractability of trying to solve the tragedy of the commons. But we’ve had traction on the problem: Ostrom identified a bunch of principles for successfully governing a common resource without government intervention. A similar set of principles for avoiding desperate status plays could lead to at least a “1% dent.”
An underlying assumption here is that players are not playing for the sake of the specific idol in their game. Girard’s theory of mimetic desire supports this assumption.
This assumes something like the 80/20 rule for AI capabilities research. My uninformed impression is that it holds for AI safety research.
Lindner. Making Enemies: Humiliation and International Conflict. Also: https://www.nytimes.com/2020/09/08/opinion/biden-trump-humiliation.html.
This is also accounting for (1) people working in the depression treatment industry and (2) anyone who might be working on marketing good-for-society products with status appeals.
https://www.emergenresearch.com/industry-report/depression-treatment-market. Note that here I’m assuming that depression is associated with dangerously low status (where the afflicted person could, for example, commit an act of violence).
I’m not an expert in the field and had to do a bunch of googling to decipher what was being said, so it’s very possible I’m reading the source wrong.