3 ms·

> As a result there was no optimal way for them to respond in this situation. The optimal way from a user's perspective would be to let surge pricing do what i

> As a result there was no optimal way for them to respond in this situation. The optimal way from a user's perspective would be to let surge pricing do what it does. The optimal way from a PR perspective would've been to cap surge pricing (the hysterical lunatics currently screaming about this on Twitter et al would have a lot more trouble getting people to blame Uber for "there's not enough Ubers around!"). It seems in this case Uber chose what was optimal for users, and they were vilified for it. Now obviously I'm talking about a sustainable optimal solution: what they ended up doing (free rides and refunds) is the most optimal for users but obviously not a steady-state solution (at the very least it would require them to keep tabs on every situation that might be considered an emergency and devote time to classifying those: everything from strong weather events to more anomalous things like the current hostage situation).

dllthomas 12y ago
Further, the "refund you after the fact" opens people up to the risk that they believe an event will motivate free rides when in fact it fails to. Definitely tough. That said, "at the very least it would require them to keep tabs on every situation that might be considered an emergency and devote time to classifying those: everything from strong weather events to more anomalous things like the current hostage situation" You mean a 50B company with the ability to massively shift allocation of resources needs to pay attention to what's going on in the world, to deploy such resources effectively and ethically?
- mikeash 12y ago
  There's a tendency among programmers to reject any solution that requires O(n) human work. Either automate it or don't do it at all. This seems to drive Google's approach (or non-approach) to customer service, as an example, and also leads to a lot of "why didn't I think of that?" moments when some company figures out they can make a lot of money without automation. (For a good example of the latter, look at all the attempts at electronic bill pay that tried and failed because of network effects before somebody finally realized that they could bridge the gap by mailing checks to anyone who wasn't yet in the system.) Here, there's an inherent resistance to anything that would require Uber to deploy humans in a way that scales linearly with their business. Surge pricing can be automated, so it's OK. Modifying surge pricing based on the nature of current events requires a number of people that increases linearly with the number of locales that Uber serves, so it gets rejected.
  - wutbrodo 12y ago
    Just because you evidently don't understand the matter at hand and the infeasibility of pushing certain decisions down to a huge employee base, doesn't mean it's due to some irrational mental block on the part of "programmers". Honestly that's the laziest form of ad hominem: "This conversation is hopelessly beyond my comprehension so I'm going to fall back on a lazy stereotype of a category that I assume the other person falls into". Hint: it's not just about labor costs. Imagine Google ranked their search results by hand, and had to defend themselves against every website owner who thinks there's foul play going on? Being able to point to a ranking algorithm which they swear they don't bias has saved them countless PR headaches (and it's not a coincidence that when other Google properties starting showing up prominently is when they started having to deal with these kind of complaints from govts and others). Uber's case is even simpler, since the surge factor algorithm is much simpler and much less secret that Google' ranking logic. If you hire a huge base of employees and task them with making tons of potentially PR nightmare decisions day in and day out, you'll be dealing with explosion after explosion and in every case, you can't point to an algorithm, but instead "Uber CHOSE not to count this weather event as an emergency and now some guy got charged a crazy amount of money to get his sick kid to the hospital!!!!".
- wutbrodo 12y ago
  That's cute, but I still maintain that it's not feasible. It doesn't matter how much manpower you have, determining whether every single surge is worthy of a price increase or an emergency that people should be protected from is literally impossible, given that there will be disagreements on every case that you choose to ignore ("there was heavy rains and uber CHOSE to keep the price high and this guy was trying to get their kid to the hospital!". And for once people complaining will be right about the "chose" part.) I don't get why it's so hard to understand that it's a bad idea to put a ton of decisions every day into the hands of frontline employees where any single "bad" decision could turn into a PR nightmare.
  - dllthomas 12y ago
    It doesn't mean inspecting every surge in real time; it means monitoring for unusual surges and keeping an eye on the news. It's not outside the reach of a large company. I also think it's odd that you think "it's not our fault, it's the algorithm we designed and chose to deploy" gets rid of responsibility for the decisions.

dllthomas 12y ago

Further, the "refund you after the fact" opens people up to the risk that they believe an event will motivate free rides when in fact it fails to. Definitely tough. That said, "at the very least it would require them to keep tabs on every situation that might be considered an emergency and devote time to classifying those: everything from strong weather events to more anomalous things like the current hostage situation" You mean a 50B company with the ability to massively shift allocation of resources needs to pay attention to what's going on in the world, to deploy such resources effectively and ethically?

mikeash 12y ago
There's a tendency among programmers to reject any solution that requires O(n) human work. Either automate it or don't do it at all. This seems to drive Google's approach (or non-approach) to customer service, as an example, and also leads to a lot of "why didn't I think of that?" moments when some company figures out they can make a lot of money without automation. (For a good example of the latter, look at all the attempts at electronic bill pay that tried and failed because of network effects before somebody finally realized that they could bridge the gap by mailing checks to anyone who wasn't yet in the system.) Here, there's an inherent resistance to anything that would require Uber to deploy humans in a way that scales linearly with their business. Surge pricing can be automated, so it's OK. Modifying surge pricing based on the nature of current events requires a number of people that increases linearly with the number of locales that Uber serves, so it gets rejected.
- wutbrodo 12y ago
  Just because you evidently don't understand the matter at hand and the infeasibility of pushing certain decisions down to a huge employee base, doesn't mean it's due to some irrational mental block on the part of "programmers". Honestly that's the laziest form of ad hominem: "This conversation is hopelessly beyond my comprehension so I'm going to fall back on a lazy stereotype of a category that I assume the other person falls into". Hint: it's not just about labor costs. Imagine Google ranked their search results by hand, and had to defend themselves against every website owner who thinks there's foul play going on? Being able to point to a ranking algorithm which they swear they don't bias has saved them countless PR headaches (and it's not a coincidence that when other Google properties starting showing up prominently is when they started having to deal with these kind of complaints from govts and others). Uber's case is even simpler, since the surge factor algorithm is much simpler and much less secret that Google' ranking logic. If you hire a huge base of employees and task them with making tons of potentially PR nightmare decisions day in and day out, you'll be dealing with explosion after explosion and in every case, you can't point to an algorithm, but instead "Uber CHOSE not to count this weather event as an emergency and now some guy got charged a crazy amount of money to get his sick kid to the hospital!!!!".
wutbrodo 12y ago
That's cute, but I still maintain that it's not feasible. It doesn't matter how much manpower you have, determining whether every single surge is worthy of a price increase or an emergency that people should be protected from is literally impossible, given that there will be disagreements on every case that you choose to ignore ("there was heavy rains and uber CHOSE to keep the price high and this guy was trying to get their kid to the hospital!". And for once people complaining will be right about the "chose" part.) I don't get why it's so hard to understand that it's a bad idea to put a ton of decisions every day into the hands of frontline employees where any single "bad" decision could turn into a PR nightmare.
- dllthomas 12y ago
  It doesn't mean inspecting every surge in real time; it means monitoring for unusual surges and keeping an eye on the news. It's not outside the reach of a large company. I also think it's odd that you think "it's not our fault, it's the algorithm we designed and chose to deploy" gets rid of responsibility for the decisions.