For the full version of this essay, go here.
As you may know, Sam Bankman-Fried ("SBF") was convicted of seven counts of fraud and conspiracy. He now faces the potential of more than 100 years in prison.
I've been trying to figure out how someone who appears to believe deeply in the principles of effective altruism could do what SBF did. It has been no surprise to me to see that the actions he was convicted of are nearly universally condemned by the EA community. Could it be that he did not actually believe in EA ideas despite promoting EA and claiming to believe in it? If he does believe in EA principles, there seems to be a genuine mystery here as to why he took those actions.
There are a few theories that could potentially explain the seeming mystery. In this post, I'll discuss the strongest evidence I've been able to find for and against each of the three theories that I find most plausible.
It seems important to me to seek an understanding of the deeper causes of this disaster to help prevent future such disasters. It also seems to me to be essential for the EA community, in particular, to understand why this happened. An understanding of the disaster and the person behind it might be necessary (though probably not sufficient) for the community to prevent similar events from happening in the future.
A few important things before we begin the analysis
In this piece, I assume that SBF committed all the crimes that he was convicted of. If it somehow turns out that SBF isn't guilty of these crimes, then some parts of this post would not apply (and you should consider most of this post withdrawn).
It's also important to note that the opinions I express in this post are, for the most part, informed by studying publicly available details about SBF and the FTX collapse, as well as confidential conversations I've had with a number of different people who knew SBF (some who worked with him, some who knew him as a friend). I promised confidentiality to these people to help them be more comfortable sharing information honestly with me, so I won't use their names or other indications of how they know him. I shared this post with them prior to publishing it to help reduce the chance that I introduced errors in what they said to me.
I've also pulled quotes from the new book about SBF, Going Infinite, and from podcast interviews with its author, Michael Lewis. Lewis spent a lot of time with SBF (starting from late April 2022 and continuing into SBF's period under house arrest), so he had a lot of time to form impressions of him.
I also had some interactions with SBF myself, which I discuss in more detail in my podcast episode about the FTX disaster. The podcast episode is a good place to start if you are fuzzy on the basic facts of what happened during the FTX disaster and want to know more. I also recorded an earlier podcast episode with SBF about crypto tech (prior to accusations of wrongdoing against him), but it doesn't provide much information relevant to the topic of this post. My first-hand experience with him was limited; it informs my viewpoint on him much less than other evidence I've collected.
I am very interested in hearing your own arguments or evidence with regard to which theory you think is most likely about the FTX calamity (whether it is one of the three outlined below or another theory altogether).
Defining DAE
Throughout this post, I'll use the term DAE ("deficient affective experience") to refer to anyone who has at least one of these two traits:
- Little or no ability or tendency to experience affective (i.e., emotional) empathy in response to someone else's suffering
- Little or no ability or tendency to experience the emotion of guilt
For instance, whereas most people would experience a strong empathetic emotional response to watching a dog being tortured, someone with DAE may have no emotional response to it. And whereas most people would feel very guilty if they learned that they had badly injured someone by accident, a person with DAE may have no emotional experience of guilt upon learning that.
Someone with DAE may still have the capacity to be abstractly convinced that it is bad that others suffer (i.e., they may logically agree that suffering is bad or believe in a philosophical theory that says suffering is bad) and may still have personal moral principles that they try to avoid violating, even though they may not feel guilty if they violate them.
To be characterized as having DAE in the way that I'm using the term (which, for the purpose of this post, I'm attempting to define in a more precise way than the typical usage), someone wouldn't have to have a complete absence of such experiences (affective empathy and/or guilt), but they would have to have a very low amount of one or more of them relative to the general population.
It's also important to note the distinction between "emotional empathy" - which a person with DAE lacks or has little of (e.g., feeling the emotions of another person when you see them suffering) - and "cognitive empathy," which a person with DAE may be fully capable of (e.g., rationally understanding that a person is suffering when you see them suffering, and knowing what to expect about their behavior as a consequence).
Similarly, some researchers distinguish between affective guilt (based on empathy) and cognitive guilt (based on an acknowledgment of responsibility). A person with DAE may be capable of regretting having taken certain actions (e.g., because they think things would have turned out better if they had taken other actions) without having the emotional experience of guilt that most people would feel.
I believe that DAE is likely positively correlated with other personality tendencies, including:
- An absence or reduction in prosocial emotions related to attachment with others, such as love
- Manipulativeness, deceptiveness, and willingness to use people as a means to an end
- An increased frequency and intensity of anger
- A dampened fear response (perhaps especially dampened in relation to the fear most people experience at the thought of violating strong societal norms, rules, and laws)
- A sense of being superior to most or all other people
Of course, this doesn’t mean that everyone with DAE has these traits.
Ethics for someone with DAE
There is an important distinction between whether someone has DAE and whether they act unethically. Someone who lacks empathy and guilt may still act ethically because, for instance, they distinguish between right and wrong and care about doing what is right. On the other hand, a lack of empathy is correlated with criminality and causing harm.
One way to think about this is that there are multiple reasons why most people try to avoid unethical behavior. In particular:
- Affective Empathy: They feel empathy for those who would be harmed
- Guilt: They want to avoid feelings of guilt, or they feel motivated by past feelings of guilt to avoid actions that were similar to those past actions that produced guilt
- Mimicry: They have been taught certain default behaviors that involve good behavior, or they may mimic others because it makes them fearful to stand out or seem weird, or they mimic the good behavior of others around them merely because of the natural social mimicry instincts that most people seem to possess (e.g., in many social circles, it is not socially acceptable to punch someone during a disagreement, so most people in these social circles won't do that regardless of their other traits)
- Belief systems about right and wrong: According to their belief system, an action is wrong, and they don't want to take actions that they believe are wrong (e.g., a Christian might try to avoid breaking the Ten Commandments even if they aren't worried about going to hell, or a utilitarian might try to avoid taking actions that cause suffering because they cognitively believe it's wrong to cause suffering)
- Punishment avoidance: They don't want to be punished (by others, by society, or by the legal system) for doing harm
People with DAE don't want to be punished and may still adhere to strong belief systems about right and wrong, so, like everyone else, they may be motivated by (4) and (5) to avoid taking unethical actions. They may also engage in social mimicry (though, perhaps less so than the average person, due to less social fear), so (3) may still motivate them to avoid certain kinds of unethical behavior. However, someone with DAE is likely to be much less motivated by (1) and/or (2) than most people are.
So it's not that people with DAE have nothing motivating them to avoid causing harm or acting unethically, but they have fewer forces pushing them not to do so than most people do. With fewer such forces pushing them away from unethical actions, they are more likely to harm others, even though there are plenty of people with DAE who act ethically and are productive members of society.
I know a number of people who I believe have DAE. Some of them are good and work to make the world better. Other people I know with DAE are genuinely frightening (for instance, one of them told me that he enjoyed it when his girlfriend would catch him cheating because it would give him an opportunity to trick her, which, to him, is a fun game).
Theories of how SBF could have caused the FTX disaster
While other theories about what happened are possible (e.g., "perhaps one or more of the people he worked with did all the bad stuff, and he was just a victim, despite the conviction"), I'm going to limit this post to considering just the three theories in the diagram above because I currently think they are the most likely ones. If you think there is a theory that's more likely than any of these three, please let me know in the comments.
I also want to point out an obvious theory that is not included here, namely, the one that would be in the missing bottom left quadrant of the chart above (i.e., SBF does not have DAE, but also, he did not genuinely believe in EA). This theory seems unlikely to me because it would require a non-DAE-based explanation for why he constantly lied about believing in EA (even going back to his pre-FTX days), and I don't have such an explanation. For reasons that will become clear later in this post, if he didn't believe in EA (and was just publicly pretending to believe in it), I think it's likely he had DAE.
Let's briefly review the three theories.
Theory A says that SBF genuinely believed in effective altruism when running FTX, and he does not have DAE. Additionally, this theory says that his unethical behaviors were just the result of some mixture of incompetence and bad luck (possibly, though not necessarily, exacerbated by a belief in naïve utilitarianism). In this theory, he did not intentionally defraud anyone, though he was, at the very minimum, extremely reckless. This theory says that he was either so incompetent that he didn't know he was behaving unethically and breaking the law, or else he was able to justify his behaviors to himself (even if he felt very bad/guilty about his actions) via a naïve utilitarian calculus based on his guess that doing so would yield more utility in the world in the long-run for all conscious beings (compared to if he followed the law and behaved ethically).
This theory seems to be in line with what SBF himself said about the disaster. As Fortune described it:
"I didn't ever try to commit fraud," Bankman-Fried told reporter Andrew Sorkin at the New York Times Dealbook Summit. "I was excited about the prospects of FTX a month ago. I saw it as a thriving, growing business. I was shocked by what happened this month. And reconstructing it, there are things that I wish I had done differently."
...SBF admitted that he "made a lot of mistakes" as CEO but denied that he used FTX's funds at Alameda.
"I didn't knowingly commingle funds," he said, arguing that it was a "failure of oversight" rather than anything malicious.
SBF went on to distance himself from Alameda altogether in the interview.
"I wasn't running Alameda," he said. "I was nervous because of the conflict of interest of being too involved."
SBF went on to claim that he was not aware of the depth of the relationship between FTX and Alameda Research, nor the sizable amount of funds transferred between the exchange and trading house.
"I didn't know exactly what was going on. I didn't know the size of their position," he said.
Theory B says that SBF has DAE but that he also genuinely believed in EA when running FTX and that his behavior and choices were guided by a combination of the two (a lack of emotional empathy and/or guilt, plus a belief system genuinely based on EA principles). While bad luck and/or incompetence may also have played a role (perhaps a substantial role), these other factors (DAE and a genuine belief in EA) played a critically important role in the calamity.
You may immediately object to Theory B, wondering whether a person with DAE could (or would) ever believe in EA, given that EA is a philosophy about doing good in the world. But EA is a cluster of beliefs, and while one's emotional capacities and tendencies can influence what beliefs a person has, I think that almost any set of emotional capacities can co-occur with almost any set of beliefs. For instance, someone who is born incapable of experiencing empathy and guilt could be raised Christian, Muslim, secular humanist, etc., and, like most people, I think would typically continue with whatever belief system they were raised with into their adulthood.
I will also add that I have met a few different people with DAE who seem to strongly believe in specific abstract belief systems (such as utilitarianism). Of course, they could be lying, but for whatever it's worth, I was pretty convinced (when talking to them) that they believed in these systems, and their behavior seemed to back that up.
Theory C says that SBF has DAE, that his DAE is an important factor in his behavior with FTX, and that he was only pretending to believe in effective altruism when running FTX.
Basically, this theory says that he was playing a long con, presumably aimed primarily at gaining some combination of wealth, status, pleasure, and/or power for himself (and perhaps also for his close companions, although it's possible that his companions were also just a means to an end for him).
In this theory, EA was just another part of that con. While luck and/or incompetence may have played a role in the disaster (perhaps even a substantial role), this theory says that he has DAE and that understanding him as having DAE is necessary to understand the calamity (and that his claims of EA beliefs were just a front).
Note that it's important to distinguish between having DAE and being broadly sadistic. I believe that most people with DAE do not actively enjoy harming people who they have no special negative feelings towards (though they might enjoy harming people who they feel have wronged them or who they think are bad in some way).
So, Theory C (which hinges on DAE) is not a hypothesis that SBF hurt people merely because he enjoyed hurting people. I think that, thankfully, it's very, very rare to find people who indiscriminately enjoy creating suffering. Even in cases of school shooters and terrorist attacks, if the attackers express joy in hurting people, I believe it's usually because they view the people they are hurting as evil or lump together the people they are hurting as all being from a group that they feel they or their group was wronged by. So, if Theory C is true, I think the most likely explanation of SBF's behavior would be selfish motivations (along with indifference towards harming others), not motivations specifically to harm other people. It's a theory about DAE, not a theory about sadism.
This is not the full post - for the rest of it, including an in-depth discussion of the evidence for and against each of these theories, you can find the full version of this post on my blog.
I don’t think that observing lots of condemnation and little support is all that much evidence for the premise you take as given—that SBF’s actions were near-universally condemned by the EA community—compared to meaningfully different hypotheses like “50% of EAs condemned SBF’s actions.”
There was, and still is, a strong incentive to hide any opinion other than condemnation (e.g., support, genuine uncertainty) over SBF’s fraud-for-good ideology, out of legitimate fear of becoming a witch-hunt victim. By the law of prevalence, I therefore expect the number of EAs who don’t fully condemn SBF’s actions to be far greater than the number who publicly express opinions other than full condemnation.
(Note: I’m focusing on the morality of SBF’s actions, and not on executional incompetence.)
Anecdotally, of the EAs I’ve spoken to about the FTX collapse with whom I’m close—and who therefore have less incentive to hide what they truly believe from me—I’d say that between a third and a half fall into the genuinely uncertain camp (on the moral question of fraud for good causes), while the number in the support camp is small but not zero.[1]
And of those in my sample in the condemn camp, by far the most commonly-cited reason is timeless decision theory / pre-committing to cooperative actions, which I don’t think is the kind of reason one jumps to when one hears that EAs condemn fraud for good-type thinking.
'By the law of prevalence, I therefore expect the number of EAs who don’t fully condemn SBF’s actions to be far greater than the number who publicly express opinions other than full condemnation.'
It seems like you can always use this to claim "sure, everyone says they don't think X, but they could be lying for reputational purposes" whenever X is taboo, but some (most!) taboo things are genuinely unpopular. That seems dangerously unfalsifiable. Of course, the actual conclusion of "more people believe taboo things than will admit to it" is true in almost all cases. But if almost no one will admit to having an opinion, the real prevalence can still be low even if higher than the visible prevalence.
I agree that on things that are taboo to support (like this) we should expect support to be greater than publicly acknowledged support. However, a near-universal lack of public support is still evidence of a genuine lack of support. We could debate how much evidence it is. Talking to EAs 1-on-1 I also have barely found any that say they support the kind of actions that SBF was accused of, but many that condemn those actions. Again, not perfect evidence, but it provides a bit of additional evidence.
Zvi's take on this is good: https://thezvi.wordpress.com/2023/10/24/book-review-going-infinite
Possibly nitpicky, but I don't think "In this theory, he did not intentionally defraud anyone" is necessarily the case even if we assume theory A (SBF genuinely believed in EA principles and did not have DAE).
I'm not sure if there's something here along the line of "if you believe in the ideas of EA then you would never intentionally defraud anyone / behave unethically under any reasonable interpretation / break the law", which I think I feel less sure about, but at the minimum it's plausible to think (as you point out) that SBF could have been a naive utilitarian, which could have made him more likely to intentionally defraud someone than if he wasn't, all else equal.
Alternatively, you do mean this to be the "unintentional" theory, but in this case there should probably also be room to explore the hypothesis that he was acting more intentionally and as a naive utilitarian, and I think there is definitely some evidence out there to support this (and I would find this more compelling and probably more likely than theory C).
I do mean theory A to specifically be a theory where he did not intentionally defraud people. I don't think the EA + non-DAE + naive utilitarianism + intentional fraud theory is one of the three most likely, that's why I didn't discuss it in detail, but I'd be interested in evidence for it (if you think there is evidence for that theory in particular that is not already mentioned in my post)
I think ultimately this is just "how much do you trust Sam's testimony about himself (e.g. how plausible is it that he thinks it's OK to take billions of customer money). Given his willingness to be misleading or lie about other things (e.g. whether or not Alameda received special treatment, the frugal image etc), this might be some reason to discount his own testimony.
I also think generally with large scale financial fraud you should expect that it is much easier to have nonzero idea of the fraud occurring than have zero knowledge of the fraud occurring; it seems not super relevant to me whether this is something SBF "let happen" or defrauding as an action was what SBF was actively pursuing, though you may disagree.
https://twitter.com/alexeyguzey/status/1728549209949995299
I'm not sure how this tweet you linked to is related to my post above?