Handy reference! Apart from the average rate, it seems also interesting to notice the variance in the table, spread over 4 orders of magnitude. This may point to something like 'global sanity' being an important existential risk factor.
I appreciated the clarity of the intro, and that it made it very clear what was in vs out of scope and how to situate this evidence in relation to other lines of evidence and the higher-level questions that are ultimately of interest
It seems quite weird - and probably misleading - to use the terms US right and US left for Republicans/Democrats who think political violence is sometimes justified?
"right" and "left" have widely understood meanings here. But what this is pointing to is just a specific subset of those things (well, a specific subset of Republicans/Democrats, which I guess is a separate & smaller label-category mismatch issue), and the subset presumably differs from the whole set in a way that's directly relevant to the question at hand?
I don't know what alternative short term I'd recommend, but one idea would just be to add anything to make it clear that this is specific jargon and not just the whole group. E.g., "US right PVSJ"
Thanks for this Michael, I think that's a good point. I've changed those labels to 'US radical right (see definition)' and 'US radical left (see definition)'. Not perfect, but less misleading.
How big are risks from non-state actors? Base rates for terrorist attacks — EA Forum
How big are risks from non-state actors? Base rates for terrorist attacks
This post presents the executive summary from Giving What We Can’s impact evaluation for 2025. At the end of this post we share links to more information, including the full report and...
I used AI to fix transcription errors, rerrarange the ideas, and suggest tweaks to the title and some sentences.
Three of the most exciting projects to come out of EA in recent years are, in a vague sense, CEA spinouts:
* Kairos is directly a spinout of CEA and now handles most support for university AI safety groups. Basically everyone I've found who knows them is really excited about what they do
* NEST is an opinionated ideas-fi...
Non-state actors deliberately using powerful new technologies to cause harm could be bad in a number of ways:[1]
By directly causing the deaths of millions of people
By creating instability and thus compromising humanity’s ability to combat other existential risks
In worst case scenarios, by directly posing an existential risk, via irreparable civilizational collapse or human extinction
But how plausible is it really that non-state actors would cause this much harm?
Crudely, this breaks down into two questions:
How many actors will be capable of committing harm in this way?
How many actors will be willing to commit harm in this way?
Question 1 is about future technologies: how cheap they will be, how easy it will be to access the relevant information, what human and natural resources will be necessary to create them, and how much harm they will be capable of. I don’t have the expertise to say sensible things about this.
Question 2 is about people. I think that to form an overall answer to the question there are a number of different approaches to consider:
Historical approaches: base rates for things like terrorist attacks, mass murders with omnicidal intent, genocides etc; qualitative histories of past omnicidal actors and past uses of dangerous technologies
Security approach: identifying and monitoring actors in the present who seem most worrying from this perspective
Psychological approach: base rates for things like dark triad traits, omnicidal wishes etc
Structural approach: factors which influence the rates of things like terrorist attacks, instability, political violence; and how these factors might change in future
This post doesn’t attempt an overall answer. Instead, it offers one concrete piece of the puzzle: some data on the base rates of terrorist attacks.
For a range of different groups, I’ve pulled together data on the number of attacks per million group members over roughly the last decade. This gives a sense of how common terrorist attacks are, and of how much the rates vary between groups.
In the main post, I discuss a summary table of the main findings. For each of the groups I look at, there’s also a more detailed appendix.
I think information like this helps us think about what rates of terrorism might be like in future, and contributes to getting a better sense of how big the risks from non-state actors are.
Summary table
The table below shows an annual rate of terrorist attacks for various different groups.
The data for terrorist attacks is an average of attacks from 2010-2019 from the Global Terrorism Database.
The data for groups comes from various years and sources. Follow the links for more detailed tables. Hover over the footnotes to see the data source for the group.
Countries included: the EU, UK, Norway, Iceland, Switzerland, US, Canada, Australia, New Zealand.
Discussion
There are some salient problems with these numbers:
The groups aren’t directly comparable to one another. Are Muslims in the UK who think suicide bombings are sometimes justified comparable to vegans in the UK? Or to environmentalists? Are Republicans who think that political violence is sometimes justified comparable to people in the UK who say they are right wing? I would caution against reading this table as a straightforward indication of relative rates of attacks. The tables in the appendices often contain additional categories for each group (e.g. UK vegans but also UK vegetarians, UK green voters etc). If you’re interested, you can play around with these numbers to see how they change the comparative rates.
The selection of groups isn’t random. I picked the groups that were most available to me, in terms of data and prior knowledge.
Often the data sources I’m using for group size aren’t great. Sometimes I couldn’t get data for the right years, and used the closest I could get (earliest 2007). Sometimes the sources are self-reports from a single survey. Sometimes I used a secondary source and didn’t bother to find a more authoritative one.
The Global Terrorism Database (GTD) doesn’t contain all terrorist attacks.[16]
Terrorism is a social rather than a natural category, and calling someone a terrorist is a political act. The activities involved in terrorism blur into other categories like mass murder, state sponsored violence, guerrilla warfare etc.
But the table does give a rough indication of the base rates for terrorist attacks.
Taking the worldwide average as a starting point, you’re looking at roughly one attack per 700,000 people.
However, it’s unlikely that access to the most dangerous emerging technologies will be uniformly distributed around the world. Exactly how it will be distributed is complicated, but if you think access will be concentrated in richer, more stable countries, then you might want to use the Western base rates instead: that’sroughly one attack per 3,000,000 people. So a bit lower.
I think looking at the rates by smaller groups is also interesting: it gives you a sense of how much the rates vary in subpopulations, which could be useful if you wanted to think about risks from specific future groups. To really understand these numbers, I think you need to have a deeper qualitative understanding of the contexts than I do, and I expect this means that some of the numbers are misleading in ways that I can’t see. That said, I think there are still some minimalist things to note
Among the groups I looked at and according to the numbers I found, the groups with the most attacks were self-reported Northern Irish nationalists, with rates of ~60 per million.[17] Then came Northern Irish unionists and radical Muslims in the UK and US,[18] with rates in the low 10s per million. UK Muslims, the US radical right, the UK right, UK and US vegans, and the UK left all had rates below 1 per million. (US Muslims were a bit higher at 1.7.) The US radical left, US pro-lifers and UK and US environmentalists[19] had rates below 0.1 per million. So the rates among the different groups I looked at vary by around two orders of magnitude.
Another thing worth noting is that terrorist attacks are reasonably rare in all of these groups, even where part of the group definition includes thinking that political/religious violence is sometimes justified. It’s cheaper to talk than to act, and we should expect many, many talkers to every act of terrorism (and millions of innocent bystanders whose ideas have been radicalised in ways they would never condone).
Even when it comes to the actual attacks, most of them are clearly not committed with omnicidal intent. Most terrorist attacks have no fatalities, and it’s plausible that terrorism in general isn’t really about causing lots of deaths.[20] Groups like al-Qaeda or the IRA have political goals which couldn’t be met if everyone died, and ideologies which genuinely support human extinction are mercifully rare. The rate of terrorist attacks with omnicidal intent must be far lower than the rate of terrorist attacks in general.
But given that there are 8 billion people in the world, the rough rates of terrorist attacks I’ve presented here aren’t exactlyreassuring when it comes to risks from non-state actors.
Appendices
There’s a more readable version of these tables here.
[a] The average responses to a few different questions over a few different years:
Do you think it is ever justified for citizens to take violent action against the government, or is it never justified?
True American patriots might have to resort to violence in order to save our country
"feel somewhat justified in using varying levels of violence to achieve their political goals" (my source isn't reasoning transparent about the questions that build this up)
[b] “Populist Right, who have less formal education than most other typology groups and are among the most likely to live in rural areas, are highly critical of both immigrants and major U.S. corporations.” 11% of the population. Plus [c].
[c] “Faith and Flag Conservatives are intensely conservative in all realms; they are far more likely than all other typology groups to say government policies should support religious values and that compromise in politics is just “selling out on what you believe in.”” 10% of the population.
[d] “Devoted Conservatives (6 percent of the population) are deeply engaged with politics and hold strident, uncompromising views. They feel that America is embattled, and they perceive themselves as the last defenders of traditional values that are under threat.”
[a] Voters for the following parties: Conservative, Democratic Unionist Party, Ulster Unionist Party, Ulster Conservatives and Unionists - New Force, Traditional Unionist Voice, Christian Peoples Alliance, UKIP, The Brexit Party, Independence from Europe, English Independence Party, Democratic Nationalist Party, Christian Party, English Democrats, Libertarian, National Front, BNP
[b] Voters for the following parties: UKIP, The Brexit Party, Independence from Europe, English Independence Party, Democratic Nationalist Party, Christian Party, English Democrats, Libertarian, National Front, BNP
[c] Voters for the following parties: English Independence Party, Democratic Nationalist Party, Christian Party, English Democrats, Libertarian, National Front, BNP
[d] I included attacks by: Anti-Muslim extremists (and suspected), Anti-Semitic extremists, English Defense League (EDL), Neo-Nazi extremists, Right-wing extremists, White supremacists/nationalists
Appendix 8: US left
Category
Number
Year
Number per million on the left*
Sources
Democrats who think political violence is sometimes justified [a]
[a] The average responses to a few different questions over a few different years:
Do you think it is ever justified for citizens to take violent action against the government, or is it never justified?
True American patriots might have to resort to violence in order to save our country
"feel somewhat justified in using varying levels of violence to achieve their political goals" (my source isn't reasoning transparent about the questions that build this up)
[b] “Outsider Left, the youngest typology group, voted overwhelmingly for Joe Biden a year ago and are very liberal in most of their views, but they are deeply frustrated with the political system – including the Democratic Party and its leaders.” 10% of the population. Plus below.
[c] “Progressive Left, the only majority White, non-Hispanic group of Democrats, have very liberal views on virtually every issue and support far-reaching changes to address racial injustice and expand the social safety net.” 6% of the population.
[d] “Progressive Activists (8 percent of the population) are deeply concerned with issues concerning equity, fairness, and America's direction today. They tend to be more secular, cosmopolitan, and highly engaged with social media.”
[e] I included attacks by the following groups: Anarchists, Anti-Fascist Activists, Anti-Police extremists, Anti-Republican extremists, Anti-Trump extremists (and suspected), Left-wing extremists, Students For Insurrection.
Though it probably does contain most of them for this time period. See Our World in Data: “ A valuable resource which also provides impressive accounts of terrorist incidents across the world is the many detailed entries in Wikipedia by year, by region or by country. Using this as a cross-reference with the GTD, we have high confidence in the completeness of global data in recent years.”
Maybe this is higher than it should be: you could argue that the correct denominator should actually be Irish nationalists, not Northern Irish nationalists.
The numbers for radical US and UK Muslims may also be higher than they should be. As Islamist terrorism is so politicised, it seems likely to me that moderate Muslims are less likely to say that suicide bombing is sometimes justified, than moderate Democrats to say that political violence is sometimes justified, for example. So maybe I’m selecting for a more extreme denominator, which inflates the rates.
The rates for environmentalists are drawing on terrorist attacks by animal rights groups as well as environmentalist groups, so arguably the real rates should be even lower. Then again, if I could get numbers for environmentalists who think violence is sometimes justified, that might increase the rates quite a lot.
See theseposts by Gwern. I’m told Bruce Hoffman’s Inside Terrorism also has good discussion of the motivations and aims of terrorist groups, though I haven’t read it myself.
Handy reference! Apart from the average rate, it seems also interesting to notice the variance in the table, spread over 4 orders of magnitude. This may point to something like 'global sanity' being an important existential risk factor.