Ideological tensions as an Effective Altruist in the civil service

The civil service delivers public services and supports the government of the day to implement policies. At first glance, working in the civil service seems to align quite strongly with EA principles and values as there's a common motivation to do good through evidence based decision making. 

In this post I want to discuss what I see as an ideological tension between working as a civil servant (in the UK government) and being an effective altruist. My main motivation for writing this is to help articulate a real frustration. On the one hand, working in the civil service can have enormous impact - arguably much more than relevant alternatives. On the other hand, there are serious ideological clashes between civil service and EA principles that any EA civil servant would need to internally reconcile.

The main structure of this post will be as follows:

1) A brief introduction outlining my experience working in the UK civil service, my experience with EA and why now's the time for me to write this.

2) An argument for why it might seem logical to work as a civil servant.

3) An argument for why one might think working as a civil servant is a more impactful than relevant alternatives in the private sector, research centres or charities.

4) An explication of various tensions between EA values and government values

5) Reconciliation and concluding remarks

 

1 - Introduction

I've worked as a civil servant in the UK government for four years now and this post mainly refers to the UK and other countries that use the Westminster political system. For those not aware, civil servants in the UK are apolitical. They work alongside politicians - ministers - but they themselves have no political affliation. The theory is that ministers are democratically elected to make decisions for the benefit of the country, and that civil servants are the ones that implement those decisions. In practice the line can be somewhat blurred - it is usually the civil service who suggest what decisions to take based on policy analysis, with the minister having the final say. It is worth noting that the civil service is wide and covers a range of roles - including key operational roles that have minimal ministerial engagement. This post will primarily focus on policy roles.

I took part in the civil service graduate programme, which involved 5 rotations across government. I've worked in the Department for Transport, the Cabinet Office, the Home Office, The Alan Turing Institute (the UK's national AI institute) and BEIS (Department for Business, Energy and Industrial Strategy). In my current role I work as a technical lead on policy around law and AI. My experience I think is relatively wide, although not very deep - but hopefully this provides context on where I'm coming from.

I've been aware of EA for a while, although started looking at it more deeply in the last year as I have been considering different career options. This naturally led me to 80,000 hours, as I've considered moving towards the AI policy space (the second highest rated career according to their analysis!). This led me to ask the question - where's the best place to do this kind of policy work. Should I stay in government, or should I go elsewhere? What's the best option based on traditional EA analysis?

There's also three observations I've had that surprised me and motivated this analysis:

• How few civil servants there are in EA
• How few civil servants care about the magnitude of their impact (compared just to the fact they are impacting society)
  
• How (comparatively) little I see civil service discussed as a EA career path
  

This post should hopefully shed some light on these questions and observations, detailing my thoughts on where I should go to make an impact. It will hopefully convince the reader of the potential impact of being a civil servant, and also explicate what I see as key tensions being an EA in the civil service.

My thoughts here were greatly informed from various discussions I've had through the Civil Service EA network, for which I'm very thankful.

2 - Why be a civil servant?

When you ask a civil servant why they work in government, you're likely to hear a few things. Flexibility, security and most relevant to this discussion - impact to society. Many people work for government because they want to change citizen lives for the better and to contribute towards long term change.

Now if you're like me, the next question you'll have in mind is this: do civil servants actually have impact? And perhaps more relevant - are civil service jobs more impactful than relevant alternatives in the private sector, research centres or charities?

Let's focus on the first question for now. In some sense the answer should be an affirmative yes. Civil servants work to implement government decisions, which are meant to support and thereby positively impact the citizens of the country. In this sense, the roles is designed to have social impact.

In response, consider the following argument: there are many 'well-motivated' roles in this world, but not all of them will have significant impact under 'traditional EA metrics (QALY, DALY etc). The main reason for this is usually the scale of the work. From my personal experience, there are two 'roles' volunteering in charities that I've done that I'd like to think were well motivated, but probably had rather minimal overall impact. The first role was volunteering in a charity shop - organising goods, steaming clothes and talking to customers. The second role was volunteering at a community centre helping the elderly with IT problems. My view is that both roles were impact constrained as they could only focus on supporting a few people. It felt good to support the old man down the road to use his phone to call his children, but the two hours spent may represent a large opportunity cost.

EA analysis I think can be heartless at times, but the numbers should be clear - helping individuals on such a local level lacks the same scalability as working in a larger organision to influence how society is run.  As EAs we'd ideally like to positively impact as many people as possible - scaling is critical.

I think this kind of argument forces a certain nuance in the analysis of being a civil servant. Strategy and policy roles are often criticised because they are very far away from the beneficiary - but I think there's an important trade-off. Macro roles allow for more potential impact - perhaps at the loss of feeling good about what one does. Writing the cost of living rebate policy probably doesn't give you the same emotions inside compared to being an assylum case worker supporting refugees but it's overall impact is likely higher. I will say of course that while policy roles may be the most obvious place to have this macro impact, there are many relevant alternatives. For example, as poor as the UK's overall Covid19 approach was, we could argue that the commercial team delivering on vaccines development and roll out had an extraordinarily high impact in reducing Covid deaths. A similar point could be made about technical teams building contract tracing applications.

Thus I argue that at least at first glance (and with the right considerations), working as a civil servant can have significant societal impact. Governments have enormous budgets and scope - there's a lot of potential.

3) How does the impact compare to relevant alternatives?

Now for the other question. Let's suppose that you're someone who wants to change how society functions. There are different ways to do this, but in the UK (I don't have experience of other governments, but I assume it is somewhat similar) this is usually done by influencing goverment law and regulation. Since I've already mentioned the 80,00 hours career review, let's just use the example that you care about AI governance or strategy.

The question now becomes - where should you go to have the most influence?

This is actually a real question I've been considering. Here is a list of roles that I've considered so far - I appreciate it is not comprehensive. We can consider this as our comparison space - we want to determine the most impactful role of the below.

• 1) Minister
  
• In the UK this would likely be in the Department of Culture, Media and Sport, as a the minister for Science, Research and Innovation.

• 2) Civil Service
  
• Policy Lead in the same department.
  
•  3) Research Centre
• For example the Alan Turing Institute or something like the Centre for the Governance of AI. Research associates or other roles.
  

• 4) Academic Institution
  
• From a PhD / Post-Doc going up towards Professorship.
  

• 5) Private Sector
  
• Incredibly variable - this could include policy leads in big tech companies or product owners for AI safety products. It could also just amount to working a role that requires some degree of AI governance.
  
•  6) Charities
• I think there's overlap here with research centres -  some roles here might involve policy research and lobbying.
  

Now there's extraordinary nuance here and I really say comparison in the loosest sense. The barriers to entry for each role are clearly different and represent different levels of seniority. Furthermore, I lack personal experience in many of these options - unsurprisingly I haven't worked as a minister before! The argument I'm going to raise is primarily rooted in my own experience working on AI-related policy in government.

I argue now that while there are 6 kinds of roles, there are in practice only three relevant categories - decision makers, implementors and advisors. Using our above example, we can think that the decision maker signs off on the policy, the implementor ensures it happens (e.g. manages the law making process, communicates it and ensures compliance) and the advisor feeds into the evidence base (e.g. through a consultation).

Here, the decision maker is clearly the minister as they have formal authority. However one could argue that the civil servant also has considerable decision making capability - they are providing the formal recommendation to the minister, and it's fairly common for ministers to just do as their civil servants recommend.

The implementor is most obviously the civil servant as it is their job to implement ministerial decisions. In practice though, some of the implementation can be outsourced or decentralised - for example technical delivery is often done by private sector consultancies. Vaccines for example are produced outside of government, and vaccine centres decentralised to the NHS / local councils.

Who are the advisors? Definitely not the minister, and probably not the civil servant - but likely anyone else. As an example - the civil servant might release a consultation asking for views on various topics. It is common for various companies, research centres, charities and academics to feed into this work. The civil servant will analyse the responses, writes up a strategy and then tries to get broader agreement and eventually ministerial sign off. Now, I don't know how long it takes stakeholders to prepare such comments, but I'm fairly sure the amount of time civil servants spend on each one probably doesn't do it justice.

Another common thing is to run roundtables, where various academics or senior industry leaders are invited to make something happen. From my own experiences it is still surprising in some sense that these academics (who are far more intelligent than myself) are just feeding in their comments, for me to accept into our approach. I'd highlight of course that reality is more nuanced than this - certain stakeholders (e.g large tech companies) may have disproportionally more influence than the average advisor.

Now I would argue that the role with the most impact tends to be decision maker. Note that this can be a positive or negative impact but for sake of argument we will also assume that all role holders here are well motivated and competent. The second most impactful role is the implementor as they make it happen. The least impactful role is the advisor, although they still do considerable good for society.

If you've been following this argument you can see where this leads.

I'm arguing that if you want to make an impact you should be the minister, and if not then be the civil servant. If you can't do those then you should work as an advisor - providing feedback to the minister and civil servants.

Being a minister of course is tricky - there only say 100 such roles available at any one time. It requires one to get elected and chances of success are low. By all means would I encourage people to do so, but I don't think it's an entirely realistic path to pursue. 

There are hundreds of thousands of civil servants (with varying levels of impact), but it's clearly much easier to get into compared to being a minister. My personal view is that the entrance barrier to join the UK Civil Service is not particularly high. I recognise that my perspective might be a bit skewed here as I joined government four years ago. I also recognise that the roles I've mentioned have limited availability, although my experience running recruitment is that many roles are so niche that they receive few applications.

I think certain advisors have serious impact, however I think the investment in becoming an advisor with serious impact comes at a large time cost. As an example, most academics we engage with are professors with more than twenty years of experience in the field compared to our policy leads with potentially one to two. I'd argue the same point for other advisors - it takes to get into a position where you can have a large influence on policy. Of course it's worth mentioning that the policy eco-system relies on a range of stakeholders - I'm certainly not arguing that everyone should suddenly transition into the civil service. It's clearly important for example that there are people doing active research on the relevant questions that can advise government. I'm just arguing the marginal individual would likely make more impact in government than elsewhere.

Anyhow, let me just drive forward my point: if one is interested in changing government policy, usually the best route is to do it from the inside. In this sense, I'd argue that if you want to make an impact, being a civil servant is usually better than relevant alternatives.

4) Tensions between EA and the civil service

Hopefully I've given you at least some reason to think that being a civil servant is not only a suitable choice for an effetive altruist, but potentially even an optimal one. My hope is that such a recognition will make my next argument hit more deeply (and frustratingly), as I discuss some ideological tensions between the two.

I think it is fair to say that EA is value laden, in the sense it presupposes the acceptance of a particular set of values. Now it's entirely possible that these values are objectively true - and as a moral realist I'd be open to this - but epistemically it is difficult to know whether that's the case. What this means is that - potentially - the EA values are wrong, and we aren't yet sure if that is the case. Other values might therefore be more important, leading to the issue of value conflict.  

The centre for effective altruism website outlines principles such as a committment to others, a scientific mindset (evidence based decision making) and openess (transparency). Marginally more controversial, but lets also add some degree of acceptance of long-termism and placing great value on future life. In practice it might mean focusing only on policy areas that have high predicted impact.

Now, I think generally government should and does endorse a committment to others, evidence based decision making and transparency. These are good things. I will shortly however be covering two values that I think government would consider even more important - democracy and comprehensivity.

Before I go into these values in detail, I first want to outline and distinguish between two relevant but slightly different notions.

1) Whether government should follow Effective Altruism (the macro)

2) Whether a civil servant working in government should follow Effective Altruism (the micro)

The first notion essentially means that government should fully embrace EA principles (I'm using principles interchangeably with values) in decision making and prioritisation.

The second notion focuses on the individual working in government. It places greater emphasis on such an individual to seek high impact areas of work. It might additionally mean embodying the various principles mentioned above in one's day to day practice.

Onto the value conflict.

Democracy

The (UK) government is democratically appointed and ministers (with a few rare exceptions) are politicians who have the democratic mandate. The government prides itself (and I think rightly so) on this mandate - it has the power to make decisions because the people elected it to do so. Civil servants in comparison are hired by meritocractic principles - their authority is indirectly granted by the political establishment. This means that ministers have overriding decision making power - they can (and often will) disagree with their civil servants, even if it goes against the evidence.

It is actually somewhat ironic that having this democratic mandate can lead to surprisingly tryannical decision making.

While civil servants are (in theory) apolotical, and focused purely on the public good - ministers are not. They have pressures to get re-elected and must consider positions that are popular as opposed to merely effective. Ministers frequently worry about how they might be portrayed in the media and are (perhaps rightly) concerned that they might lose their jobs if things go poorly.  Ministers themselves might have a slightly warped sense of risk, at times taking an overly risk averse approach to avoid controversy and other times adopting an overly risk loving approach, taking advantage of generally short tenures in post.  

The democratic mandate means that government decision making can at times conflict with the scientific, evidence based decision making espoused in effective altruism theory.  Furthermore, the fact that ministers can change so frequently and priorities so radically can often undermine the capability to seriously evaluate evidence and make informed decisions. The shortness and uncertainty of ministerial post mean that government strategies are often focused on the immediate future - usually less than 5 years at a time. This represents further conflicts with the long-termist agenda often espoused by EAs.

I'd also like to discuss a real example that really hit home to me regarding this tension. The UK government has immense capital budgets, spending hundreds of billions of pounds on for example the net zero agenda. The government loves to talk about how much this capital spend. On the flip side, the government is making serious administrative cuts - they are trying to cut 90,000 civil servants by 2025. I appreciate here that a different government might slightly differ in approach, but I think the following argument still holds. Government is often evaluated on much it can spend on capital with a limited operational budget. Hopefully the parallel here is clear: this is exactly the same as the EA argument against evaluating charities based on what portion of one's donation goes towards the cause (instead of the operations). What should matter is not how much money is being spent but how much impact is created!

The democratic mandate can at times be very challenging and frustrating for civil servants themselves. We've seen recent news cases where civil servants have actively pushed back against government policies. Whether this is the ethical choice is an interesting question. I would argue that they're certainly not meeting their civil servant responsibilities, and maybe that they're even undermining democratic values. I still however have some sympathy for them - it really can be a struggle working on policy that one does not believe in.

Suppose a civil servant works in a policy area but really disagrees with the ministerial steer. This might be because of ethical reasons, or it could be because the steer goes completely against what the evidence suggests. This can be a really challenging experience and I think one that many civil servants experience in their career. 

One option is of course to go through with it nevertheless, but this can feel quite powerless and to some extent undermines the reasoning for joining the civil service. I argued earlier that one should join the civil service because it is better to influence the policy from within. If we're unable to leverage this influence at all it might feel pointless doing the role. Furthermore, if you have serious ethical qualms with the policy area you're dealing with, then it might even feel like you're selling out - not only are you not creating good, but you're actively implementing bad.

I've also seen cases where people try to actively block work. This seems somewhat problematic as well - you're actively going against your responsibilities as a civil servant, you're (arguably) taking a politically biased position and undermining the democratic mandate. You're also contributing to civil service beaurocracy, making it more difficult for things to get done - while being paid by the public purse. One argument to consider is whether it is really that bad to undermine democratic principles - especially if it is leading to negative outcomes. One of my colleagues has talked to me (quite proudly) about their (successful) attempts to block new policies that would reduce benefits for pregnant women. If the policies are really that bad, then maybe it's really worth fighting for - and democracy be damned.

Finally, you could just quit. This saves one's own moral character to some degree, but there are worries that your replacement will probably still make it happen, and likely worse than yourself (as it takes time to learn a role). This also feels a bit like giving up.

Comprehensivity

Government cannot just focus on a select few problems - in some sense, everything is its responsibility. Traditional EA theory might suggest that we focus on impact - donate to the best charities and work in the most critical areas. If government took this approach then maybe it focuses on say net zero, AI safety or pandemic preparedness - investing all of its resources.

What however happens to other areas - the one's that are less obviously impactful to our long term future but still need someone to work on. Presumably someone still needs to work on trade policies to ensure that there sufficient exports to meet the needs of UK citizens. And presumably someone needs to control who can immigrate or ensure that local councils have the funding for constitutent bins are collected.

 One thought is that if government don't do this, then someone will. Perhaps the free market will pick up on it and provide the relevant service. I don't think this a particularly convincing argument. Government takes responsibility to address market externalities and ensure fair access. There are specific work areas where it just isn't financially viable for private sector parties to respond to, and if pushed - would just lead to only wealthy gaining benefit. Moving on to the regulative side - we might think there needs to be some kind of central authority to ensure market compliance. Presumably we don't want to just let big coorporations decide immigration policy.

It's worth mentioning that comprehensivity is linked to democracy. In so far as government has a democractic mandate, it also has a responsibility to ensure that all citizen needs are accounted for. Some of these needs can certainly be provided by the private sector - and government invests a lot of resources to stimulate these markets. The government is voted for by the public and is there to serve their needs. If certain needs aren't being met elsewhere then it is up to the government to ensure that it is.

Just to mention a few nuances here. Government does not deal with everything, and in fact a lot of implementation work is outsourced. Thus the notion of comphrensivity is bounded primarily to work mentioned above that addresses externalities or requires central authority. Additionally, one might argue that EA theory places emphasis on supporting areas that are neglected. If government suddenly eschews responsibility for a workstream, it would become more neglected and thereby shift the priority areas. I suspect however that this new equilbrium would be probably be inadequate - there would likely be new priorities but I think there would still be significant gaps of neglect from government departure.

It's also worth considering what happens if government does suddenly place great focus on a few key areas. In fact, it's not that hard to imagine - there's been an extraodinary push in recently towards the net zero agenda and significant amounts of investment given towards it. The important question is whether these fields can properly scale with more investment  - which would likely be monetary, as opposed to operational resource.

So now suppose we grant that government should take responsibility for a wide range of areas. As an EA in government you'd presumably be going to the highest impact policy areas. There are concerns about a potential bottleneck, where all the EAs of government all pivot towards the roles - massively increasing competition in the space while leaving other areas neglected. One solution of course is to suggest that people place value on areas that they might be suitable for. For example, clearly not all policy advisors in government have advanced knowledge on AI, so maybe they shouldn't apply.

There's a few problem with this solution. First thing is to note that the civil service tends to pride generalism - it's not about knowledge of a specific policy, but knowledge of being a policy person. It's very common for people to enter policy spaces with no prior knowledge of the field. If this view is taken to the extreme then the suitability argument weakens (although still holds to some degree). A big follow up problem to this is the fact that civil servants rotate careers incredibly frequently - with most policy leads switching jobs every 1-2 years (not that dissimilar to ministers). Now I think these are primarily practical issues with the running of government, as opposed to a fundamental clash between civil service and EA values, but I think it is important to note.

I'd also highlight that the range of roles in the civil service and its generalism can lead to a pressure to always be searching for new opportunities. It's no surprise that some of the most senior civil servants I've met have rotated between crisis jobs yearly. I do think that the range and ease of movement can create a pressure on effective altruists within government as they provide a push to always do the job with more impact - further driving the problem of churn. One should also consider their own marginal impact in pursuing such roles. If they are so high in demand then it's likely that another person could do the role to a good standard as well. It's important not to ignore neglected areas that might need people the most!

Reconciliation and concluding remarks

I'd ideally have ended up this post with a comprehensive and tangible solution. It would have explained exactly how I would solve all the problems of government and outlined a direct approach for a civil servant navigating the process. I think however the tensions are at a fundamental level that there is no easy solution. I will however express a few comments on how I and I think other civil servants can start to reconcile these tensions.

Let's start with democracy. There's a whole raft of political philosophy literature discussing potential problems with the model and a whole lot more discussing why it might seem better than alternatives. I'll instead focus on the civil servant perspective.

I mentioned an irony that the democratic mandate can lead to tyrannical and anti-scientific decision making. I think it's worth considering a notion of a democracy from a macro as well as local lens. If we choose to block work getting done we may be undermining 'macro' democracy. If however we have considerabl evidence from stakeholders then choosing to block such work might uphold a more 'local' democracy. Such behaviour could be consistent with EA principles such as consideration towards others or evidence based reasoning. I'd of course be remiss to highlight the possibility that our own views might wrong when disagreeing with ministers.

Another approach to consider is to simply avoid policy areas with significant ministerial opinion. A lot of new areas such as AI governance are actually quite good in this regard - most ministers lack the technical knowledge to effectively comment on policy in this space, meaning there is less likely to be conflict. Getting the right balance here can be tricky as you want them to be interested enough to have skin in the game and push it forward, but not so interested that they too many opinions themselves.

On the point around comprehensivity, I think there are a few questions to consider. I think it's generally quite damaging to government for policy people to move so frequently. On the flip side, it's entirely possible that certain policy areas are just so more immensely more impactful that it's worth investing whatever you can do to get there. 

I wonder if the best approach is more to stay in a specific field and develop expertise, as your marginal replacement (the best person who would be next to do the job after you) would just be radically worse than yourself in terms of impact. While in the same role, one could then try and bring in EA principles to the day to day work. I think there's already lot of demand in government to write business cases with the right KPIs (Key performance indicators). A lot of EA theory is well suited to government decision making. 

I'm going to draw this post to a close now. In summary I have argued for three main things. First, that EA and civil service principles are at first glance very much aligned, and that civil servants can have really high impact to society if they work in the write space. Second, that someone interested in making an impact should consider the civil service even over other relevant opportunities as it is more impactful to be an implementor over being a advisor. Finally I have argued that there are theoretical tensions between EA values and CS values, and that an EA civil servant will need to properly reconcile them.

I hope you can see that I'm still on my own journey in reconciling these issues, and I'd really welcome thoughts and feedback.