4 ms·
That’s the thing: my guess is that the reason your model works well is because (respectfully) very few people stand to gain enough from gaming it. It’s an insta
by samhw 5y ago
That’s the thing: my guess is that the reason your model works well is because (respectfully) very few people stand to gain enough from gaming it. It’s an instance of Goodhart’s law: when a measure becomes a target, it ceases to be a good measure. (In other words, search result quality is probably a convex function of popularity.)
I wonder about the approach @Jack was reputedly considering for Twitter, where users would select algorithms[0] out of a marketplace of options (perhaps with an interface similar to fonts.google.com, where they could sample the results of each algorithm for a given query). I don’t know if that would run within Google or would have to rely on something like Common Crawl. Google has a huge advantage not only because of its algorithms but also - as I’m sure you know - because of its index, and other miscellanea like click-and-query data.
[0] In that case for filtering/ordering, in this case for searching, though one could argue it’s basically the same thing except the latter has not only a user but also a search query as a parameter.
- marginalia_nu 5y agoMy point is that this is a basic dynamic that exists in biology as well (or any ecosystem). If you have extreme monoculture, specialists can and will flourish. Specialization as a strategy is much less effective in a diverse ecosystem. If you target one algorithm successfully, the others will work better as a result; and vice versa. It's fairly easy to construct mutually incompatible scoring systems. Even minor differences in scoring will limit the success of specialization. It doesn't matter if it's a search engine or half a continent growing nothing but wheat, the dynamic is the same.
- samhw 5y agoSorry for the late reply - and thank you for responding! That makes perfect sense. I share your thought about this. I hope my comment wasn’t too unclear. I was just considering that a more diverse ecosystem of algorithms doesn’t necessarily imply a more diverse ecosystem of companies. Hence my thought experiment about a marketplace of algorithms within one search provider. But of course I’m not wedded to that, and I have no reason to be. Both situations seem equally improbable to me: (a) Google’s losing dominance, or (b) Google’s providing a marketplace of different algorithms. I’d be interested to hear any more thoughts you might have! Thanks again for your reply.