Work in a morally questionable part of finance in order to make large donations to charity (where you think the donations will have a greater positive impact than the harms done by the work)?
Work in a factory farm and make the conditions less bad, causing less suffering overall?
Join a political campaign you think might be harmful in order to gain connections (where you think the connections will let you have more positive impact than the harm done by working for the campaign)?
Work at a lab developing dangerous biotech so that you can blow the whistle if you see something particularly dangerous happening?
This post sets out 80,000 Hours’s views on these issues.
We look at how to analyse these situations using moral philosophy, and then apply the results to some common options, like finance, law, and the oil industry. We assume you’re capable of taking the harmful option, and focus on the question: “Is it right to take it?”
We believe that in the vast majority of cases, it’s a mistake to pursue a career in which the direct effects of the work are seriously harmful, even if the overall benefits of that work seem greater than the harms. And within a job, we think you should avoid actions that seem very wrong from a commonsense perspective, even if you think it’ll do more good overall.
We think that this position is justified even if all you value, morally, are the consequences of your actions. But we also think that we should put weight on the commonsense view that the ends don’t always justify the means.
Having said that, it’s important to bear in mind that all careers will involve some degree of negative impact. So the question to think about is not whether your career involves making others worse off, but how much it does and in what way. We think you should be wary of treating classes of career paths monolithically, and instead should pay close attention to the specifics of the job in which you would be working. In the post, we’ll outline a process you can use to judge individual cases.
For many, the first rule of ethics is “do no harm.”
But this is not a good guide to career choice, because every career causes some negative impact.
For instance, imagine that you become a doctor with the aim of helping people. Everyone makes mistakes, so at some point you’re going to mess up, and people might die as a result. In some cases, these mistakes will be your fault, plain and simple — maybe you had too little sleep, or maybe you could have studied a bit harder. But just because this scenario is extremely likely to happen doesn’t mean that it’s unethical to become a doctor.
Here’s another example. Suppose you become vegan in order to avoid hurting animals. You’ll still have to eat grains, and when those grains are harvested, typically some field mice will get killed in the process. In our modern economy, almost every action has some negative consequences.
To really minimise your negative impact, you could become a hermit and live alone in the woods. But that hardly seems like the most ethical life, because you’re forgoing huge opportunities to help others.
So, merely pointing out that a career path has negative effects doesn’t mean much, because some negative effects are unavoidable.
Another problem is that it really matters exactly how you cause a negative impact.
These days when we think of harmful careers, finance is often the first that comes to mind. But the financial sector employs almost 4% of the workforce, or 6 million people in the US.1 These people do a huge variety of tasks, from legal compliance, to IT, to trading, to banking.
First, it’s highly unlikely that all of these positions are net negative. Rather, some do harm, others do good, and many are roughly neutral.
Second, it matters exactly how the negative consequences come about. If a banker is overpaid, that may be socially wasteful and so in effect have a negative impact on others, but ethically it’s very different from a banker who commits fraud.
When we get serious about the ethics of career choice, we cannot simply do no harm, or treat entire industries as a monolith. Rather, we should aim to:
Compare positive and negative impact, and seek to do far more good than bad overall.
Understand exactly what the harm consists of in the specific job at hand, and then distinguish between permissible negative impact, like the doctor making a mistake, and impermissible negative impact, like the banker committing fraud.
Career advice needs to be tailored to your situation, and that’s even more true when there’s a significant risk of doing harm.
This is what we’ll cover in this article.
Why to avoid careers with significant negative impact, even if you think you’ll do more good
As we’ve shown, entirely avoiding negative impact isn’t possible. But that doesn’t mean you should take a harmful option lightly, even if you think you’ll do more good overall by donating or other means.
Our default position is: don’t take a career for the greater good if that career directly causes significant harm.
And for very similar reasons, we’d oppose taking actions within a career that seem very wrong from a common sense perspective, even if you think they’ll do a lot of good.
To see why, consider this question:
Is it ethical to kill one person, harvest their organs, and use those organs to save the lives of five people?
Almost everyone agrees this isn’t ethical, at least not in the actual world we live in. And this helps to explain why we would not generally recommend taking a career that causes significant negative impact in order to do more good.
There are five reasons not to take careers that cause significant harm:
1. It might violate rights (non-consequentialist reasons)
Many people think killing the one person is prohibited by the basic rules of moral behaviour, such as “do not kill,” which is one of the Ten Commandments.
Early modern philosophers often also appealed to basic principles of morality that would rule this out, like Kant’s categorical imperative.
Modern moral philosophers would probably focus on the person’s right to life. Killing someone against their will violates their rights, and so is probably not permissible, even if it is for the benefit of more people overall.
Almost all ethicists agree that these rights and rules are not absolute. If you had to kill one person to save 100,000 others, most would agree that it would be the right thing to do (although you might still not be able to bring yourself to do it). However, they would say that such violations should only be made in exceptional circumstances.
Although they’d presumably fall short of literally killing people, jobs with a large negative impact might involve violating rights or other basic moral principles (such as truth-telling), which means that they will usually not be morally permissible on these grounds.
Even if you’re sceptical of Kantian or other rules-based morality, it’s hard to be confident these are wrong — more on this below.
Later, we’ll consider which jobs in particular might be ruled out for these reasons.
2. It’s probably not the highest-impact path due to hidden harms (consequentialist reasons)
Other moral philosophers think that what ultimately matters is only whether your actions produce good or bad effects overall. These are called “consequentialist” theories.
These philosophers think that in principle it would be justified to kill one person in order to save the lives of five. However, they generally don’t think that in practice people should go around harvesting organs. That’s usually because such activities would lead to indirect harms that would make everyone worse off in the long-term. So the means also affect the ends. (For more detail, see the distinction between a decision procedure vs. criterion of rightness).
Turning specifically to career choice, we think there are many ways that taking a harmful job could have overall negative consequences, even if it seems higher impact in the short run (and even if you think you’d be ‘replaced’ anyway). For instance:
Reputation: It’ll harm your reputation and the reputation of the people you associate with. This makes it harder to achieve good in the future. This effect could be very significant if you’re part of a social movement, such as effective altruism, veganism, feminism or environmentalism, because your actions could tar the reputation of the whole group. (And note your reputation can be ruined merely by doing a job that’s widely seen as harmful, even if it actually isn’t.)
Character: Habit and norms are a huge driver of behaviour. Acting in ways that violate important ethical norms (even in small ways) and/or being around unethical people all today probably has a negative impact on your character, making you less able and likely to act morally in the future.
Motivation: Helping people is motivating, and a job that seems harmful day-to-day could be extremely demotivating, raising your chances of burning out.
Career prospects: The skills you learn will be less useful, because industries widely seen as harmful often shrink or perish. For instance, the earnings in investment banking went down significantly after new regulations were introduced after the financial crisis. Likewise, if you want to do good, it’s helpful to make connections with other people who want to do good, and it will be hard to do this if you work in an industry that’s widely seen as unethical.
Signalling: You’ll implicitly make it more socially acceptable for other people to take these jobs, and they probably won’t try to use the position for good (e.g. they probably won’t donate a large fraction of their income). This effect is most significant when (i) you’re doing something unusual and (ii) you have a large network or public platform.
There will probably be other harms, too,2 and these harms will likely offset any indirect benefits. We think that commonsense moral intuitions often have important insights behind then, even if they’re not immediately obvious. All of this means that it’s usually better to avoid taking the harmful position even if you only care about the consequences of your actions.
3. You’re probably wrong about the benefits
Suppose you’ve tried to weigh the harms against the benefits and have concluded the benefits are so great, that they outweigh the harms.
But not only might you be ignoring hidden harms (as above), your estimate of the benefits might be way off.
In particular, if you think the benefits of an action are particularly high, due to regression to the mean, you’re probably too optimistic about them.
The problem gets worse if you throw in the natural human tendency for self-delusion and overconfidence.
And the track record of people who believed they were “breaking a few eggs” in order to promote “the greater good” seems terrible and includes many of the worst people in history. So taking an outside view should also give us pause.
All things considered, there’s a good chance you’re simply wrong, and the benefits don’t actually outweigh the harms — even though you think they do.
You should be especially cautious in any circumstance where you also personally stand to benefit from the harm, such as via holding onto influence, money or status.
We think that in general, in real life, it’s better to stick to simple rules — and just avoid doing anything that seems seriously wrong from a commonsense perspective.
4. Moral uncertainty and cooperation
Suppose you think only consequences matter, and you’re convinced that the benefits of taking a job are larger than the harms, even taking into account all the hidden harms and all the arguments above that imply you’re probably wrong. Should you do it?
It’s very hard to settle ethical questions, so 100% confidence seems crazy. A survey of professional philosophers found they are very divided, and a majority are not consequentialist.
Number of responses
301 / 931 (32.3%)
Accept or lean toward: deontology
241 / 931 (25.9%)
Accept or lean toward: consequentialism
220 / 931 (23.6%)
Accept or lean toward: virtue ethics
169 / 931 (18.2%)
Even if you think consequences are likely all that matters, you should take into account the fact that there’s some chance that it’s unethical to ignore other considerations. And so, out of caution, you might still not want to take a harmful career. And, for non-consequentialist views, it’s clearly morally wrong to take a career that does a lot of harm — potentially seriously wrong.
It’s also good to take other moral views seriously so you can maintain cooperation and friendly relations with others who have those views. For instance, if you do something that lots of other people think is unethical, those people aren’t going to want to work with you or support you in the future, reducing your long-term impact.
When we look at real examples of harmful careers, there are often nearby options that are not harmful and almost as good on other dimensions. When this is true, you may as well just take the non-harmful option.
For instance, while some areas of finance might involve deceptively marketing financial products, other parts (like retail banking or venture capital) seem neutral or positive. Venture capital is also very high-earning, so if you’re thinking of going into finance to earn to give, it seems better just to avoid the harmful areas.
There won’t always be nearby alternatives, and sometimes you’ll just face a sharp tradeoff. But our guess is that this won’t happen very often.
When might it be justified to take a job that has significant negative impact?
We think putting the last five points together makes a good case for mostly avoiding jobs that have a significant negative impact. Doing so could mean violating rights and other moral principles, which should probably concern you even if you think consequences are all that ultimately matters (due to moral uncertainty). The net impact is also probably worse than you think due to hidden harms and because your estimates are wrong. Finally, there are probably better nearby alternative options to be considered.
That said, there might be exceptional circumstances in which you should take an option with serious negative impact, even if you’re not a consequentialist. As an extreme example, Oskar Schindler ran munitions factories for the Nazis, producing mess kits and, later, ammunition for Nazi soldiers, but in so doing was able to earn enough money to ensure the safety of 1,200 of his Jewish workers. He also deliberately ran the factory inefficiently, so that fewer munitions were produced. It’s hard to imagine a more unethical job than running a Nazi munitions factory, but Schindler is widely seen as a hero, and was celebrated in the movie Schindler’s List.
In an emergency, like Schindler found himself in, it can be permissible to have a significant negative impact in order to achieve a greater positive impact.
There are also clearly cases where it’s morally permissible to make individuals worse off. For instance, it’s permissible to fire an employee who’s performing badly, even though that will make their life worse. Earlier we also mentioned the example of a doctor who makes a mistake.
In what follows, we’ll show how to think through these cases, and why it could be morally acceptable to work in finance, even if the industry causes a serious negative impact.
First, clarify exactly what situation you’re in
When you’re concerned about taking a job with negative impact, you could be in one of the four situations in the following table. Each situation requires a different response.
We’ve illustrated the four situations for the decision of whether to earn to give in an industry with some substantial negative impacts.
(We just focus on the direct impact of the work and the impact of the donations, whereas in reality you’d want to look at all the types of impact, including career capital and the indirect harms we mentioned earlier.)
Impact of the work
Impact of donations
Should you do the job?
1. Net harm.
Large negative and morally impermissible.
No. The net impact of taking the job is negative, so everyone agrees you shouldn’t do it.
Killing 3 people to save 1 person.
2. Net positive, but arguably morally impermissible.
Small negative and violates a moral rule, e.g. against harming an innocent person.
No. Your net impact is positive, but taking the job may well still be morally impermissible.
Harvesting the organs of 1 person to save to lives of 5.
3. Net positive, permissible.
Small negative but morally permissible (e.g. the positive effects are sufficiently large that they outweigh the negative).
Maybe. It depends on your other opportunities. Maybe there’s something even better.
Telling a lie to save 3 people’s lives.
4. Industry isn’t actually harmful.
Maybe. It also depends on your other opportunities.
Saving 3 lives when you could have saved 4.
If taking the job is net negative, then obviously you shouldn’t take it. The job could also be net positive, but not worth taking because you have a better option.
The tricky part is deciding whether a path with net positive impact is morally impermissible or not — deciding between situations 2 and 3 above. This is what we explore in the next section.
Where’s the line between permissible and impermissible negative impact? A step-by-step process for deciding.
Below are some of the key considerations that moral philosophers have proposed to decide between permissible and impermissible harms. We think they’re a reasonable place to start when doing the analysis, although all of them are debated within the literature. If you favour non-consequentialism, you might interpret these as morally relevant rules of behaviour. If you favour consequentialism, you might interpret them as rules of thumb for when the indirect harms are relatively small.
These rules of thumb are roughly grouped into four categories.
Emergency situations: are the benefits much larger than the harms? Most agree that large enough benefits can outweigh harms. For instance, imagine an axe murderer came to your door and asked where your family is. You lie and tell him they’re out of town. Deceiving people is (usually) wrong, so was that the wrong thing to do? No — almost everyone would agree that it’s OK to lie to save the lives of your family. Protecting the lives of your family outweighs the wrong of lying. How large do the stakes have to be before it could be permissible to do harm, or violate someone’s rights? There’s no clear line, but our rough impression of the literature is that many people think it becomes permissible when the benefits are 100 times larger than the costs. Those with more consequentialist views might think that 10 times larger is already enough, while those with more deontological views might think over 1,000 times is required. Almost everyone agrees that there is some point at which the harms are immediately outweighed.
(Note, though, that these hypothetical figures assume certainty about the size of the harms and benefits. As covered, in the real world, we likely face huge uncertainty about the benefits, which would suggest taking significantly more caution.)
Are the harms a means to achieving the good thing, or a side effect? Deliberately causing harm in order to accomplish a greater good is widely thought to be wrong. For instance, killing innocent people to promote a cause (like in a terrorist bombing) is wrong, even if you expect to do more good in the long term. However, if the harm is an unintended consequence, it’s more likely to be permissible. For instance, it could be justified to tolerate civilian casualties to win a just war as long as you make an effort to avoid them, because, unlike the terrorism case, you don’t try to kill the civilians in order to achieve the end. If the harms are a side effect, then it’s likely permissible to cause a negative impact if the benefits outweigh the harms substantially — even if by much less than 100x (e.g. 2–5x).
Do the people involved consent to being made worse off? Winning money from a friend in a poker game is not immoral, even though it makes your friend worse off. Part of this is because your friend consented to being in the game, knowing that they might lose. If you cheated, however, then this would be wrong, because the friend doesn’t consent to play against a cheater. Or if your friend was addicted to poker, then they might not be able to give true consent. With the doctor who makes an error, their patients consent to be treated with full knowledge that the treatment might go wrong. Likewise, the ratio of benefits to harms can again be much lower if both parties consent. (A non-consequentialist might even think it’s permissible to make everyone worse off if they all consent.)
Is no one being exploited? Is the arrangement just? Both parties might gain, but if the gains go disproportionately to one side rather than another, and that side is much more powerful than the other, then it might be considered exploitation, and therefore impermissible. This is why many people consider it wrong to run a sweatshop factory, even if the jobs are better for the workers than what they would have had otherwise (among many other factors).
Based on the above, here’s a step-by-step process for deciding whether a job is morally permissible or not:
Are the total benefits larger than the harms, including the hidden harms? If not, don’t do it.
Are the (counterfactual) benefits 100 times larger than the harms with high-confidence (or 10–1,000 times larger, depending on your moral views)? If yes, the non-consequentialist factors are plausibly outweighed.
If the benefits are between 1 and 100 times larger than the harms, check the following: a. Would you be harming people as a means to your end? b. Do the people who get made worse off not consent? c. Is one group exploited? d. Is it unjust?
If you answer yes to any of these, it’s probably best not to do it (though it still depends on the situation).
Now we’ll apply this process to a real case: finance.
Case study: earning to give in finance
We’ve promoted the idea of taking a highly paid job in finance in order to donate to effective charities — an example of a path we call earning to give.
Suppose you could take a job in finance where you’d expect to earn $200,000 per year (which is below the mean), of which you’d donate $50,000 per year to GiveWell’s recommended charities. Is this ethically permissible?
As we’ve said, finance is a huge sector, so it depends on exactly what you’re doing. While some jobs in finance might have a negative impact, others are likely neutral or positive. A job that generally has a positive impact can end up having a negative impact if you do harmful or unethical actions within it.
Let’s do a rough application of the steps above to finance to show how you might think through the case.
How large are the benefits compared to the harms?
If we define the value of giving $1 to a random American as one “unit” of impact, then we think each dollar donated to GiveDirectly creates at least 20 units of impact. This is mainly because GiveDirectly’s recipients are about 100 times poorer, and money goes further the less you have. We’re using 20 rather than 100 to be conservative, in case there are some spillover benefits of making rich countries wealthier.
On the basis of GiveWell’s review in 2016, we think Against Malaria Foundation is about four times more cost effective than GiveDirectly, and so creates 80 units of impact per dollar of donations.
This means that donations of $50,000 per year produce as much value as giving $4 million to randomly chosen Americans each year, and in concrete terms would be enough to save about seven lives per year. So the question is whether a single individual working in finance and earning $200,000 per year causes a comparable amount of harm.
In a 2013 article, we argued that for the financial sector on average, it’s implausible that the harms are anywhere near this large. It would mean that the finance sector as a whole was doing harm on the scale of killing tens of millions of people each year.
Moreover, since then we’ve seen other quantitative estimates, which generally find that the harms are smaller than the salary. For instance, we found a rough estimate that workers in finance cause negative externalities to the American economy of about 30% of their salary, which would be ~$70,000 per year. And this seems more likely to be an overestimate than an underestimate, since, among other reasons, it doesn’t fully take account of replaceability.
We also made a rough estimate of the damage caused by jobs in the financial sector that increase the chance of a financial crisis, and found a figure of $42,000 per year.
This negative impact is large enough to make working in finance in a way that increases the chance of a financial crisis arguably ethically wrong if you don’t donate to effective charities. But if you donate 25% of your income to Against Malaria Foundation, the benefits are over 50 times larger than the harms.
However, bear in mind some caveats:
First, the estimates are highly uncertain, and only apply to finance jobs on average. If you picked the most harmful jobs in finance, they could be much worse. For instance, in the leadup to the financial crisis, certain executives in investment banks who oversaw poor lending practices probably had a far greater negative impact than typical workers. It seems plausible that many caused more harm than the good they could have done by donating their income to charity (which, by and large, they didn’t…).
This isn’t, however, an argument against finance in general. It’s possible to do huge amounts of harm in many professions, and obviously you should avoid this.
Second, a 50 times difference isn’t obviously enough to put you in an outweighing scenario, and so we still need to consider whether the negative impact is ethically permissible, as we do in the next section.
Third, we’ve ignored hidden harms.
For instance, you could have an advocacy impact, such as making it more socially acceptable for others to take jobs like yours, causing additional harms. However, going into finance is relatively common, while donating a large fraction of your income to effective charities is not. So it seems likely the bigger effect is to encourage others to donate more.
A more difficult case is with reputation. The effective altruism community has to some extent become associated with earning to give in finance, which has discouraged some people from getting involved. This could have done more harm than good (though it seems like a relatively small factor going forward, since those costs have already been paid).
On the other hand, we’ve ignored the possibility that some jobs in finance have a positive impact. For instance, value investors like Warren Buffett do intense research to identify undervalued firms. This brings information into the market, helping to make the price of company shares more accurate. This means companies get investment in proportion to their needs, helping the economy to grow.
Moreover, Buffett often invests after major crashes. This helps to prop up falling prices, making markets less volatile. So, it’s at least plausible that Buffett’s impact is positive, though it’s hard to say.
Are the negative impacts permissible or not?
Let’s suppose the benefits of the donations are only 1 to 100 times larger than the negative impacts of finance, and that we don’t count that as an outweighing situation. Then, would the path be morally permissible? Would we be in situation 2 or 3 from the table above?
We want to know:
Are the harms intended?
Are people used as means?
Do the people who get made worse off not consent to it?
On the other hand, suppose you work at a hedge fund that trades the markets and makes large profits at the expense of other investors who trade frequently. This makes other people worse off, but it doesn’t seem morally impermissible. Rather, it seems more like the example of winning money off your friend while playing poker, or haggling to get a cheap product.
Some other jobs in finance are probably just socially wasteful. For instance, investors employ asset managers who don’t beat the market on average, so don’t produce any value for them. Many of these asset managers also probably don’t create significant positive externalities by making the market more efficient (though perhaps some do, like Warren Buffett as we mentioned earlier). This is a shame, but it doesn’t seem like it’s morally impermissible to take the job if you’re going to donate the money. If investors want to waste their money on useless managers, that’s OK, so long as they’re not being deceived. If you then redirect these funds into socially positive activities, that’s for the better.
Are there better alternatives?
Suppose you restrict yourself to a part of finance that has neutral or small positive direct impact. In that case, we’re in situation 4 from the table above.
We still think it’s often not best to pursue finance to earn to give, because there are other options that have greater impact or career capital.
For instance, the expected earnings in tech entrepreneurship are similar (though higher risk), but tech entrepreneurship potentially has a greater direct impact, and likely gives you better career capital too. So, if you’re a reasonable personal fit for both finance and tech entrepreneurship, then there’s a good argument that tech entrepreneurship is better.
It’s hard to generalise, because there are so many different types of activities within finance. Some parts seem net harmful, some morally impermissible, and some are just outweighed by alternatives. However, there are probably also options that are reasonable contenders for someone with good personal fit.
Which jobs might get ruled out?
We made some guesses at jobs that involve causing a significant amount of negative impact. In these jobs, the negatives are likely too great compared to the benefits, even if you donate. Some might also be morally impermissible for non-consequentialist reasons.
Marketing and R&D for compulsive behaviours such as smoking, alcoholism, gambling, and payday loans (and see this post on being a tobacco CEO)
Lobbying for rent-seeking industries
Activities that make financial firms highly risky
Fundraising for a charity that’s less effective than average
Tax minimisation for the rich
Working in the oil and gas industries might also qualify, but we haven’t looked into them much.
Bear in mind that the list isn’t the result of thorough research, and any of the above could easily be wrong. Making good estimates of the size of the harms and benefits is very hard, as is analysing the ethics of the situation. (For example, read our analysis of whether it’s harmful to work in a leading AI lab.)
What actions are ruled out?
Doing an analysis at the level of jobs is still very broad — for most people, analysing specific actions will be more useful.
Unfortunately, it’s hard to give clear guidelines that everyone can agree with. However, to give you a flavour, here are some examples of actions that could become possibilities in your career, which we think should be ruled out:
Being dishonest (including withholding important negative information) in order to have a stronger pitch for a charity or an idea
Covering up unethical behaviour committed by someone doing important work
Acts of violence as a form of protest, even if you think it might be justified to draw attention to a crisis like climate change or a potential AI disaster
Should you go into a harmful industry in order to make it better?
For instance, could it be justified to work on a factory farm for a year in order to publicise the conditions, or develop better welfare standards that could be applied across the industry?
This is a tricky situation, and should not be pursued without a lot of thought. Our main advice is to be cautious and seek advice from people with an outside perspective.
When might it be justified? Applying what came earlier, one situation might be outweighing — if the benefits are expected to be much larger (such as 100 times) than the harms. This seems plausible with the example suggested above.
Second, it seems more acceptable if you can work in a position that doesn’t directly harm the animals, since then the harm seems more like a side effect, rather than intentional.
If you’re considering taking a career with a substantial negative impact (even if you think it’ll be net positive), you probably shouldn’t, for a few reasons:
There are likely to be better alternatives.
The option is probably less positive than it first looks, due to hidden harms.
Even if it’s net positive, you probably shouldn’t violate common standards of morality. Even if you mainly care about the consequences of your actions, it’s worth putting some weight on the commonsense position that the ends don’t justify the means.
However, that doesn’t mean you should never take a career that causes some harm. For one thing, all careers have some negative impacts. In exceptional circumstances, it might well be justified, such as when the benefits outweigh the harms by more than 100 times (or 10–1,000 times, depending on how consequentialist you are). It also might be justified to cause the negative impact even if the benefits outweigh the harms by a much smaller factor (say 2–5 times), if the harms aren’t especially morally problematic, like being an overpaid asset manager.
We think a combination of these factors can apply in high-stakes areas such as national security, where it’s very hard to guarantee you’ll never do harm — even very significant harm — but you might also have opportunities to prevent a lot of harm, and the moral rules are pretty complicated.
For example, the national security establishment in the US has a big effect on the development of AI – and we think risks from AI are one of the world’s most pressing problems. But it’s important to take seriously the substantial harm that’s sometimes been done in the name of national security. You will find yourself in an ethically complicated situation if you’re professionally obligated to protect one country’s national security — potentially at the expense of another’s. Sufficient progress reducing those risks may well put you in an ‘outweighing’ situation, and if there are no other better options, the harms of a career in national security are also more likely to be permissible.
So, if you are considering taking a harmful job that causes some harm because you think overall it can help you make the world better, be really cautious, and explore all the arguments before you go ahead. Much comes down to the details of the situation, so you need to make a case-by-case analysis.
As of March 2017, according to the Bureau of Labor Statistics, the “Finance and Insurance” industry employed 6,223,100 people. Archived link, retrieved 8-April-2017.↩
Another type of argument concerns uncertainty in your estimates of the positive and negative impact of different options. We’re a bit more unsure about these, so they’re in a footnote.
If ‘fat-tailed’ negatives are more likely than fat-tailed positives (i.e. there are more ways for things to go wrong than to go well), then your estimate of the negative impact is likely to be underestimated relative to positive impact. This could mean that a good rule of thumb for impact is to focus on ‘robustly good’ paths, and avoid significant negative impacts.
Relatedly, if the argument is that you can cause a small guaranteed harm in order to have a larger positive impact, this relies on your estimate of the large positive impact being accurate. But your estimate of positive impact is probably too optimistic, which again makes “ends justifies the means” situations less attractive than they first seem.↩