Two thirds of the world lives in an autocracy (OWD). Most of them don't live under brutally coercive military dictatorships, rather under regimes where the ruling group faces little restraint, but still maintains a certain democratic facade (often labeled electoral autocracies/hybrid regimes/anocracies). In such regimes minorities who are not part of the ruling coalition often face harsh discrimination. Not only is their right to self-govern severely limited, but they may face state-enabled barriers to employment, business opportunities, schooling, parental and other personal choices. For the billions whose rights are violated and are discriminated upon by their government, defeating autocracy is an obvious priority. And unlike in the realm of poverty reduction or public health, there has been no clear progress on this front since the '90s.

There is a strong utilitarian case for valuing democratic societies both for its intrinsic and instrumental value (See Sen, 1999). Having more rights and opportunities for human flourishing is a good in itself, but causal studies of democratization also found that it boosts growth and education (See Acemoglu et al., 2019, Gallego, 2010). The new technologies championed by Effective Altruists (AI, life extension, geoengineering, etc.) are only going to live up to their positive potential if they exist within a broadly democratic and just system (an assumed condition typically unmentioned or overlooked by EAs). Political institutions also shape the outcomes of poverty relief programs, including those favoured by effective altruists. A cash donation program to the poor in low income countries can only be effective as long as the host government doesn't expropriate the benefits it has brought about. An educational intervention which proved effective when run at small scale by an NGO, ultimately ran into political difficulties when scaled up in Kenya (See Bold et al. 2017). A (causal) study evaluating the effects of international campaigns eradicating various diseases in developing countries also offers a cautionary tale: it found that while such epidemiological programs extended life expectancy and led to population growth, they didn't boost incomes per capita in the longer term (See Acemoglu & Johnson, 2007). The material benefits brought about from wiping out tuberculosis can easily be lost to poor governance.

Despite the self-evident harm from autocracies, it seems to be a topic of lesser interest by members of the effective altruist community. I see few posts on the topic on EA & LessWrong forums (more relevant/discussed 1,2,3,4), the topic isn’t featured among charities evaluated by Givewell, I see only marginally related efforts by OpenPhil or priorities by 8000 hours, which are either centered on US politics or focused on long-termist global governance challenges. Such considerations also seem to be largely absent in the evaluations of existing EA programs implemented in autocratic countries. In contrast, there are many other ‘regular’ non-profits conducting various governance & human rights interventions around the world. But such organizations tend to justify their work using deontological reasoning, focusing on the moral clarity of their intent and the transparency of the processes they follow. Defeating autocracy is often presented just as a distant aspiration, rather than the outcome against which they measure their success. So why is there a dearth of EA approved interventions? And why do governance interventions seldom have justifications  based on "cold utilitarian calculus" favored by EA?

One potential (but unsatisfying) answer to the above questions is that we don’t know how to defeat autocracies. True. We have good tools (e.g. RCTs) and steadily growing quality evidence on the impact of health and nutrition interventions, but there are no reliable approaches to quantify, say the overall impact of Nobel-peace prize awards or that of Radio Liberty broadcasts. Nor can we clearly decompose the factors that contributed to the downfall of any particular dictator. Stil, assuming that my diagnosis is right, with regards to the strength of the  evidence demonstrating that good governance matters, then it would follow that we should then invest more in gathering evidence and learning on how to achieve (lasting) institutional change. 

I have conceded that measuring the impact of political interventions is hard. I cannot present bullet-proof evidence of interventions that work everywhere. Nor am I able to summarize the related scholarly literature and efforts. Instead, I offer a list of practical considerations that I think (with further refinements) could be incorporated into the utilitarian calculus of political interventions. I am hoping these can help EA advocates think more strategically about toppling autocrats around the world.

No marginal revolutions - A recurring problem with interventions in various domains is that they often don't scale very well. Local context varies and village scale intervention may be run smoothly by a nimble NGO, they often run into bureaucratic hoops when implemented nationally (e.g. teacher reforms in Kenya). A different, additional threat faces interventions geared at bringing about political change. While we may observe successful small interventions increasing accountability in various autocratic contexts, we should worry they may become victims of their own success. For example, local environmental campaigns were increasingly successful in China for years before facing a crackdown.  An autocratic regime may strike down on it through various coercive methods as soon as it sees it as a threat (See Chaudry, 2022). Similarly, campaigns mobilizing people for a cause may  help win over hearts and minds (See Madestam et al. 2013). But there again lies the risk of sudden reversal in attitudes,  when the groups opposed mount their own campaign by fueling people's resistance to rapid change. Such pendulum swings in public opinion are everywhere, but autocracies can leverage their superior media access to fight back.Or consider the impact of donating weapons to freedom fighters in their uprising against an oppressor. These weapons could either help bring about victory or they may just prolong defeat and bring about additional suffering. The road to any political change is highly non-linear, making any contribution’s marginal impact harder to predict. 

Tricky moral calculus - While the harms done by autocracies are evident, quantifying the expected benefit from (an increased probability of) achieving this aim is incredibly hard. Bringing down a regime will have mixed impacts: some people will visibly incur major losses, whereas hopefully many more will gain from it. These diffuse gains may also take more time to fully materialize.  How should one assign weights to conflicting impacts? A related problem is that self proclaimed do-gooders may be inclined to stay away from ethically contentious interventions. Take information campaigns, as an example, where we have various effective interventions (as demonstrated by numerous high quality econometric studies).One could start systematizing the evidence on whether a more bland, factual message is as effective as one relying on a heavy dose of propaganda. But are EA advocates willing to rely on more questionable methods? In the past, people have tried using misinformation, deception, leaks, outing people, assasination attempts, staging coups for what they may have thought to lead to a greater good. Under what circumstances should these methods be advocated for? (See this post on why leaks are more effective than transparency initiatives to catch crooks). Autocrats are by definition creating an uneven playing field, hence tipping the balance may well require questionable methods, possibly violating or at least finding loopholes around questionable regulations. (See this intervention providing students with VPN to circumvent the Great Firewall in China). Political interventions often involve both good and bad, a moral puzzle dreaded by many utilitarians.

Think global, act local - Some EA advocates suggest one of the best ways to do good is to earn lots of money in the richest parts of the world and re-channel those funds to the poorest ones where those dollars go a longer way. While this certainly makes sense in some fields, it may not be as applicable to the world of politics. Working on political change requires building trust and cooperation across large domestic networks. Foreign interventions and foreign funding may be viewed with suspicion and/or of lesser legitimacy. It can also become an easy target of state crackdown, with 44% of countries worldwide having adopted legislation that restricts foreign NGOs and/or foreign funding flows (See case study from Ethiopia). It is also unclear to me whether money is the most scarce factor, as opposed to able actors willing to invest time and take risks. While EA has spear-headed a global and donation focused approach to charity, there are reasons to believe that working for political change is primarily a national and labor intensive project.  

The above challenges listed are by no means exhaustive, nor are they insurmountable. But they may have held back ambitions to design political interventions using an effective altruist lens. What would it take for EA advocates to embrace funding programs aiming at toppling dictators? Do we need more tools and experiments or more money going into political science research? Is there a substitutability between EA donation efforts and their time as political activists? Are there already a subset of interventions in this field who’s effectiveness could be compared rigorously using an EA framework?

As it stands, I see a blindspot in the EA community’s current efforts in tackling autocracies. This post  is meant as a call to encourage (1) collaborative thinking on evidence based approaches to furthering democratization (2) understand the current reluctance in engaging on this important issue.