- Hadn't read the short story but thankfully it was very short! Yeah so I think you're getting at the difficulty of summarizing information or something like Goodhart's principle, which is about how can metrics detach themselves from the thing you intend on measuring (specifically when you optimize for them). But what exactly causes any …
- Hadn't read the short story but thankfully it was very short! Yeah so I think you're getting at the difficulty of summarizing information or something like Goodhart's principle, which is about how can metrics detach themselves from the thing you intend on measuring (specifically when you optimize for them). But what exactly causes any possible metric to elide some factor that is important to some stakeholder? To keep the running community resiliency example, what's the issue with surveys causing them to always misrepresent/fail to represent some stakeholder? Why can't the survey accurately represent what people think? Besides, I think that even if the metric missed something we care about, as long as it still correlates strongly with the thing we truly care about, it would be worth measuring and caring about as a metric. This seems much more likely to be the case about our attempts to quantify problems/risks/good things.
- Ah interesting about the generations. Hadn't thought about that
- Hadn't read the short story but thankfully it was very short! Yeah so I think you're getting at the difficulty of summarizing information or something like Goodhart's principle, which is about how can metrics detach themselves from the thing you intend on measuring (specifically when you optimize for them). But what exactly causes any possible metric to elide some factor that is important to some stakeholder? To keep the running community resiliency example, what's the issue with surveys causing them to always misrepresent/fail to represent some stakeholder? Why can't the survey accurately represent what people think? Besides, I think that even if the metric missed something we care about, as long as it still correlates strongly with the thing we truly care about, it would be worth measuring and caring about as a metric. This seems much more likely to be the case about our attempts to quantify problems/risks/good things.
- Ah interesting about the generations. Hadn't thought about that