Effective altruism is an engine that runs on ideas. To be engaged in EA is to be on the lookout for new ways to do good. To be learning, questioning, testing and comparing, across an incredibly broad array of fields.
My fear is that this breadth makes EAs vulnerable to a particular type of false friend: the good story. A good story is not necessarily true. Rather, a good story’s appeal lies in the satisfaction it offers in the telling, which might be driven by a twist or turn, or some elegant simplicity. Above all, good stories are memorable.
Good stories are particularly tempting to EAs because of the enormous amount of ground they try to cover. If you are trying to spend very large amounts of money at pace, on anything from better PPE to space governance, then narratives and heuristics can help you navigate the terrain. This is a totally understandable approach, which Will Macaskill described on a recent 80,000 Hours podcast.
The idea that a highly effective way to improve educational outcomes in low-income countries is to give deworming children is an example of a good story. Its conclusion is unexpected, but explainable with a series of logical steps. It doesn’t matter how good a nation’s schools are if children are absent through illness. Eradicating intestinal worms could improve learning by ensuring high attendance.
Once students are present, let’s try to ensure they stay awake. Another good story holds that anywhere in the world we could boost learning by shifting school start times to better align with the teenage body clock. This one is easy for listeners to identify with and feels a simple fix, certainly more straightforward than getting into the details on teaching quality or targeted support.
The problem is that good stories are not always accurate guides. The deworming finding that captured so many headlines did not stand up under re-analysis. The idea that a later start would improve academic outcomes has not been rigorously tested, but this fact is in itself instructive. In England, the Wellcome Trust and the Education Endowment Foundation could not recruit enough schools to participate in a trial, indicating that the idea is unlikely to be practical at scale.
Of course, the EA community does not ignore evidence, and it is not the intention of this critique to ignore the enormous effort EAs invest in evaluating ideas. But once a good story has stuck, it can be very hard to set it aside, even in the face of increasing challenge, as has occurred in the case of deworming, which was one of the earliest causes championed by charities such as Givewell and Giving What We Can.
There is also a softer version of this risk, where a good story is true, but something boring is better. It may be the case that changing start times would help teenagers learn more in school. But it is highly unlikely that the effect would be larger than overcoming more prosaic challenges, like improving the quality of teaching or increasing the availability of high-quality tuition.
I don’t have a straightforward remedy to problem of good stories. I think that generalists – or specialists exploring new fields – are more likely to be tempted by good stories than domain experts, because generalists have less background knowledge and are less likely to encounter timely reality checks. Good stories also offer a neatness that can help mobilise support and appeal to funders.
But I think it must be possible to develop some early warning systems within EA that minimises the risk good stories pose, without compromising the movement’s ambition and urgency.
The danger of good stories
Effective altruism is an engine that runs on ideas. To be engaged in EA is to be on the lookout for new ways to do good. To be learning, questioning, testing and comparing, across an incredibly broad array of fields.
My fear is that this breadth makes EAs vulnerable to a particular type of false friend: the good story. A good story is not necessarily true. Rather, a good story’s appeal lies in the satisfaction it offers in the telling, which might be driven by a twist or turn, or some elegant simplicity. Above all, good stories are memorable.
Good stories are particularly tempting to EAs because of the enormous amount of ground they try to cover. If you are trying to spend very large amounts of money at pace, on anything from better PPE to space governance, then narratives and heuristics can help you navigate the terrain. This is a totally understandable approach, which Will Macaskill described on a recent 80,000 Hours podcast.
The idea that a highly effective way to improve educational outcomes in low-income countries is to give deworming children is an example of a good story. Its conclusion is unexpected, but explainable with a series of logical steps. It doesn’t matter how good a nation’s schools are if children are absent through illness. Eradicating intestinal worms could improve learning by ensuring high attendance.
Once students are present, let’s try to ensure they stay awake. Another good story holds that anywhere in the world we could boost learning by shifting school start times to better align with the teenage body clock. This one is easy for listeners to identify with and feels a simple fix, certainly more straightforward than getting into the details on teaching quality or targeted support.
The problem is that good stories are not always accurate guides. The deworming finding that captured so many headlines did not stand up under re-analysis. The idea that a later start would improve academic outcomes has not been rigorously tested, but this fact is in itself instructive. In England, the Wellcome Trust and the Education Endowment Foundation could not recruit enough schools to participate in a trial, indicating that the idea is unlikely to be practical at scale.
Of course, the EA community does not ignore evidence, and it is not the intention of this critique to ignore the enormous effort EAs invest in evaluating ideas. But once a good story has stuck, it can be very hard to set it aside, even in the face of increasing challenge, as has occurred in the case of deworming, which was one of the earliest causes championed by charities such as Givewell and Giving What We Can.
There is also a softer version of this risk, where a good story is true, but something boring is better. It may be the case that changing start times would help teenagers learn more in school. But it is highly unlikely that the effect would be larger than overcoming more prosaic challenges, like improving the quality of teaching or increasing the availability of high-quality tuition.
I don’t have a straightforward remedy to problem of good stories. I think that generalists – or specialists exploring new fields – are more likely to be tempted by good stories than domain experts, because generalists have less background knowledge and are less likely to encounter timely reality checks. Good stories also offer a neatness that can help mobilise support and appeal to funders.
But I think it must be possible to develop some early warning systems within EA that minimises the risk good stories pose, without compromising the movement’s ambition and urgency.