- cross-posted to:
- hackernews@derp.foo
- cross-posted to:
- hackernews@derp.foo
I posted the Forbes article that’s nothing but a link to this article instead.
Meta: Based on how Lemmy works (IE: articles that get cross-posted link to each other sub-communities), I’m thinking that we might want to err on posting more articles on this software? Anyone have an opinion on this?
I’d be cautious about interpreting too much causality here, other than for Tesla (which has already resulted in a recall). I didn’t see if they controlled for additional factors (like age of driver or age of car), and it seems to beg some questions (are there differences between the F-150 and the Ram trucks?). There’s also questions about the influence of weather (we’d expect this to average out within state but not nationally. NorCal has very different weather than Florida or Minnesota and weather-specific performance might have some influence. You might be able to disentangle that using their state specific stats but I didn’t see it if they did it.
In the pickup example though, maybe it’s caused by swapping the Dodge badge with the word RAM.
Insurance companies don’t care about causality. Correlation is good enough for insurance. They don’t need the root cause, they just need to know how much it all costs at the end of the day.
From my perspective, this is a correlation that utterly destroys Tesla’s claims of autopilot and/or FSD being “safer” for drivers though. Correlation is certainly not causation, but its difficult to prove causation when the correlations are inverted from the thesis.
Oh, I know. I’m a scientist and I care about causality.
Also, people love to make causal inferences from data like these about BMW drivers being jerks or whatever.
When I first moved out to CA I ran across a random news article about a guy wrapping his just-purchased lambo around a tree on a twisty mountain road, killing himself and his passenger. I think the naive causal inference (too much car for his level of experience) is probably pretty safe. When dealing with highly aggregated, wide ranging, and deliberately assembled stats, we just need to exercise more caution.
Plus, I needed an excuse to post my joke.