Isn't the answer always going to be "brake, and do it faster than a human driver's reflexes could manage"?
I'm not sure philosophers could manage much more in terms of help here than to provide after the fact rationalizations for whatever the car companies decided in the first place. And the only advantage they'd have over PR people* is the "PhD" after (most of)their names. That said if anyone wants to start throwing money at me any time soon I'm not going to bring this up.
*Not that PR people don't have advantages over philosophers for this, like that they'd know how to do it in a way that would work.
Here's the base price for our latest model self-driving car.
The upgraded version that doesn't want to kill you will cost $15,000 extra.
Also as a side note the Trolley Problem is a hilarious comparison to make because the entire problem associated with those cases* is "holy hell all our basic moral intuitions are just flying all around on the map with no reasonable pattern behind them* in these cases and we have no idea how to make them consistent and/or come up with a way that lets us put enough of them into a coherent reasonable pattern that we can comfortably disregard the other ones".
*The original "kill one or allow five to die because of a trolley" case is actually half of the problem not the whole one. The other half looks something like "is it ok if you go in for a minor surgical procedure and your doctor, realizing that you have compatibility with like five other people each in need of a separate organ (heart;kidney;kidney;liver;liver, say) goes ahead and just kills you and takes the organs once you're under". People tend to react one way in the first case, but the other way in the second. And then there are eight hundred other cases due to the whole "why why why?" problem.
**Patterns, sure, but not ones most people would think are worth anything as moral principles.
1 is what I was thinking. Programming the car to crash into a building or a light poll to prevent a collision with humans sounds like a non-starter anyway. From a technical point of view, I don't understand how you program "things you can crash into without killing anybody but the occupant of the car" with any degree of accuracy.
Finally, a private sector opportunity for Dominique O'Brian.
The idea that self-driving cars are anywhere near functional enough that this kind of thing would be a real problem worth considering is nonsense. It's about the seventieth or so biggest problem with self-driving cars as they currently exist. Nicely done land-grab by ethicists, though. There was an article in Technology Review about challenges remaining in the development of self-driving cars; those problems included: operating in rain, identifying pedestrians, determining if road obstacles are crumpled paper or, like, rocks, and the fucking _trolley problem_.
We have on-street trolleys here, so there might be two trolley problems.
7: Except that, at least in the article, it wasn't even ethicists who were involved - it was psychologists and, apparently, people from the Toulouse School of Economics. And if that isn't terrifying...
So if we divide the fat man into tranches...
I mean, drivers are taught to brake (rather than swerve) in these situations anyway. Aren't they? Or is this a US/UK distinction (I have a vague memory of the subject coming up before in TFA)? The problem with swerving is that pedestrians will try to jump out of the way of the car and they will get it wrong some of the time. How often are you - is any driver - in the situation where you have no choice between "drive into pedestrian and definitely kill him" or "drive into wall and definitely kill self"? There will be a much larger bracket of speeds at which (car, pedestrian) might kill the pedestrian, but (car, wall) would be much less likely to kill the driver, because she's strapped in and has an airbag and a crumple zone and so on, and I suggest that this bracket includes most speeds at which unexpected pedestrians are likely to appear, ie the speeds you'll be driving at in towns.
DFT data:
Speed Chance of pedestrian death Of driver death
20 mph 5% less than 1%
30 mph 9% 3%
40 mph 50% 17%
Not to mention that a self-driving vehicle will start braking sooner - that halves your stopping distance from 30mph, so you may not hit the pedestrians at all, or only when doing 10 mph or so.
9: I would much rather have a psychologist involved in developing self-driving car algorithms than a philosopher, that's for sure.
How often are you - is any driver - in the situation where you have no choice between "drive into pedestrian and definitely kill him" or "drive into wall and definitely kill self"?
Assuming a self-driving car will actually follow urban speed limits, I'd bet that situation would be even more rare than it is now.
It might be possible, with some sort of radio tagging scheme, to program self-driving cars to crash into ethical philosophers, either in preference to other pedestrians, or just whenever the opportunity presents itself.
They could name the different models of cars after the philosophy guiding the brake/swerve/keep going decision.
"The Peter Singer model will always sacrifice the driver rather then hit a squirrel. It's popular with animal rights crowd."
"The David Hume model will just keep going, because you can't know that running into pedestrians at 50mph will kill them this time just because that's what happened in the past."
It's not so hard to think of a scenario where a car would do something different than a driver. Say you're on a road where the speed limit is 45--pretty fast, but still possible that there are sidewalks and pedestrians--and there's a wreck right in front of you and a truck that definitely isn't going to stop in time behind you. If you brake, you can avoid hitting the wreck, but will be hit by the truck. You can swerve to avoid harm to yourself, but with the result that you'll hit some pedestrians.
If that happens, I think the government should step in and mandate the Kant model.
Perhaps you could strap a philosophy postgraduate to a boom on the front of each self-driving car, like Tom Hardy in "Mad Max: Fury Road". He could provide rigorous ethical oversight of the car's swerve/crash decisions in real time, and also act as an additional impact buffer.
11, 13: Right. Thinking of this issue as a practical problem is idiotic -- human drivers don't deal with this sort of situation "Save myself, or avoid that busload of nuns/children?" in any systematic way, and it's not a problem because it's not a situation that happens with any frequency at all. We're talking about an issue where perfect design on otherwise well-functioning autonomous cars (which isn't likely in itself) might save tens of lives in an average year. Not worth worrying about.
Assuming a self-driving car will actually follow urban speed limits,
But would they need to? Speed limits are designed with human drivers, and their associated reaction times, in mind. If all cars were self-driving, is there any reason why they couldn't drive faster (on average) than cars do now?
20: Self-driving cars won't make me any more of a sober pedestrian.
16: problems with that scenario are:
1. how do you know the truck _definitely_ isn't going to stop in time?
2. even if the truck doesn't stop, it will still have slowed down. So you have the choice between swerving and hitting pedestrians at 45 mph, or keeping your line, braking and stopping, after which you are rear-ended by a frantically braking truck going at 10 mph. That's assuming the truck does the same thing you do. If the truck driver knows she definitely won't stop in time to avoid the wreck, then (by your logic) she's going to swerve. So if you just brake, you'll stop short of the wreck, and watch safely from your stationary car as the truck swerves past you and onto the sidewalk, strewing pedestrians in its wake.
Philosophers' time would be much better spent trying to come up with a machine vision algorithm that can work in rain, if 7 is anything to go by. Get to it, chaps.
16 seems very contrived. I don't think there's any way a human or computer would reliably estimate all of that. And, unless the truck is deliberately trying to kill you, if you could stop before the accident, it would be able to slow down to well below non-lethal speeds.
A computer-assisted commenting system could stop me from being pwned.
13: So, possibly rare enough that they market it however they feel like and eat the inevitable but rare out of court settlements. Or I guess what 19 said--programmers do have to think about these things in somewhat systemic ways because the code has to do something, but it's probably not that important unless you're the one being mowed down.
If the car does value pedestrians' lives over drivers, I could see teenagers playing chicken on the highway take on a whole new level of destruction.
Since I don't believe in akrasia, clearly there's a sad, sick part of me that wants to argue with you.
1) You don't have to know for sure to decide not to take the chance.
2) Maybe the truck is braking, maybe it isn't. Also, the truck doesn't need to swerve because its driver isn't going to get hurt: you're the one in a vulnerable position here.
How many human lives could have been saved by the person-hours which have been devoted to writing about and discussing the trolley problem, assuming normal safety-management valuations for human life and reasonable estimates for the hourly cost of philosophers?
Assuming that philosophers react rationally to incentives, would it therefore be ethical to threaten to murder the next philosopher who writes about trolley problems in order to discourage others from doing so?
16 seems very contrived
People die in accidents like this (the truck behind them didn't see that traffic had stopped) with some regularity. The major contrivance is locating it on a street with pedestrians, but even that isn't much of a stretch.
If all the vehicles are autonomous and the truck is also guaranteed to be following safely and stopping in time, then sure, no problem.
27.2 assumes that the self-driving car knows something about the state of mind of the truck driver. That is an unwarranted assumption.
26.2: assuming self-driving cars get released in the near-term trolling them is instantly going to become a thing. See how long you can keep a self-driving car stuck at a stop sign by sticking your foot out into the street? Sure! Convince a self-driving car to stop short on the highway by tapping your brakes or swerving ever-so-briefly into its lane? Kinda fun, amirite? I assume people have already trolled Teslas just on the off-chance they're in autopilot mode.
Also have people seen this video? There are much more pressing issues ("don't randomly swerve into oncoming traffic") to hand.
27: but the rational thing is to assume that the truck is being driven safely and reasonably (as most trucks are most of the time), no? In which case, yes it will try to brake when its driver sees the wreck, and will probably stop in time.
29: if a self-driving car detects that a vehicle behind it is not keeping a safe distance for a given speed it will presumably either change lanes or slow down to a safe enough speed for the given following distance. (Another way to troll them, btw.)
12 was clearly written by someone who has not read enough moral psychology to know what it looks like when psychologists try to write about ethics...
33: I'm going to actually stop interacting with you because I think, in a larger sense, you're an idiot, so we can leave it here.
program self-driving cars to crash into ethical philosophers, either in preference to other pedestrians, or just whenever the opportunity presents itself.
Perhaps you could strap a philosophy postgraduate to a boom on the front of each self-driving car
would it therefore be ethical to threaten to murder the next philosopher
I think the real lesson here is that anything that keeps ajay away from the wheel of a car is probably a good idea.
32: as indeed any safe driver should do, human or not.
(This reminds me of that Asimov story, "Evidence", about this nutcase trying to prove that the liberal lawyer running for president was secretly a Kenyan Muslim robot, and Susan Calvin points out that you can't prove someone's a robot from their behaviour alone, because the Three Laws apply to ethical human beings as well as robots.)
This may just be my bias against perfectionism, but it seems pretty implausible to me that "all cars think about trolley problems before deciding what to do" is going to save more lives than "all cars brake as quickly as possible without overthinking it." You just want everyone to do something good enough and do it quickly.
ogged, I'll grant that it's not a contrived accident. There were seriously like three within the span of one week that basically fit that description not five miles from my house! I've been hit when someone didn't see a light change -- I was hit by the car that was hit actually, but I was in my car. And the situation is worrisome enough that I know that when I have to stop suddenly on a highway I always check the rearview mirror to see if the idiot behind me has noticed the slowdown.
Where I'm not sure is that the question is ever going to require a "kill driver/kill pedestrian" algorithm, because I'm reasonably sure I don't have that algorithm, and I manage to drive a car reasonably well, and the self-driving car just has to be better than me in ordinary situations. So I think it will be fun for people to think about, and gold for PR people who are having to deal with trying to talk up excitement about things that can't yet tell rocks from paper bags, but practically, by the time you have braking suddenly in traffic sorted, "how much risk the car itself takes on" has to be sorted.
the self-driving car just has to be better than me in ordinary situations
Why?
Because people are bad at driving cars? Because it's hard to stay vigilant in a monotonous lengthy task?
The thing I look forward to with self driving cars is the inevitable hacking (as recently happened with Jeep). You could hold someone hostage in their vehicle going until they paid a certain sum of money, for example. Or threaten to take them somewhere undesirable and turn the car off. Assassination will become easier, too. The more automated cars become the more attractive a bicycle seems.
At the moment there is no system on the planet that is remotely as good at driving cars as people are. The thing with self-driving cars is that they are bad at different things than people are. The conversation about self-driving cars always revolves around things that people are bad at, but the list of things that are essential for driving that people are extremely good at and that computers are hopeless at is much, much longer.
The more automated cars become the more attractive a bicycle seems.
A bicycle on a road full of self-driving cars? Good luck with that.
39: I'm just not setting the bar at perfection, or on whether it understands the seven billion variations of the trolley problem. It just has to be better than the average person, no?
It just has to be better than the average person, no?
I definitely see no reason why that has to be the case.
30: The only thing that will be more popular than trolling will be my-friend's-attempt-to-troll-a-self-driving-car-didn't-work Vines.
45: Well, I suppose you could reason that an automated car which is allowed to drive itself on public roads will have to be better than the average person or they won't let it drive there.
42: but no one is suggesting that self-driving cars are ready for mainstream rollout. They're still getting better, but they're getting better quickly. (Unlike human drivers.)
42: I'm very tired, and not explaining myself well, but I grant that they're not even close to having a good self-driving car. I'm just saying that I suspect that by the time cars are as good as the average driver, trolley problem situations aren't going to look like a big deal.
Sifu -- I think you've got crossed wires here. Cala is saying that an autonomous car that's as good as or slightly better than a person in all regards is good enough to be usable, which I'm pretty sure you'd agree with. It's just that, as I understand you, you're very skeptical about autonomous cars getting anywhere near that point.
47: I don't think that's a particularly warranted assumption, but even taking it as given, better on what scale? Better at a driving test? Better at not getting in at-fault accidents? Neither of those strikes me as remotely sufficient.
49, 50: oh, I see. Yeah, we're on the same page.
If bank robbers use a self-driving car as their get away vehicle, can the car be charged as an accomplice if they get caught?
31: I was told to drive defensively, not rationally. I'll save the principle of charity for when thousands of pounds of fast-moving metal aren't involved.
Self-driving cars are also starting with a significant disadvantage--roadways, signage and traffic laws and patterns that were built around human drivers. If you result everything from scratch in a way that was optimized for computer drivers instead of human drivers, the computers would probably suddenly look a lot better (and people would look a lot worse). And yet still, the computers are quickly getting better at driving on the real-world existing roadways. They've got a long way to go, but technology is improving quickly.
thousands of pounds of fast-moving metal
One or two nights ago, I considered doing a post about how people who use this formulation are insufferable. Thank goodness I didn't do it.
We computer scientists have a program for everything except premature ejaculation, but I hear that's coming quickly.
If you result everything from scratch in a way that was optimized for computer drivers instead of human drivers, the computers would probably suddenly look a lot better (and people would look a lot worse)
There are discussions about how to do this. If it happens, pedestrians, bicyclists and other road users would be shit out of luck, in basically the same kinds of ways that they are currently shit out of luck on freeways.
And yet still, the computers are quickly getting better at driving on the real-world existing roadways.
Are they? How do you know?
I mean, what was the point of that question. Are you implying that technology for self-driving cars isn't rapidly improving? And if so, is this a quibble over the word "rapidly", or something else?
59: from reading news articles where the sources are the companies -- and other researchers -- that are working on self-driving cars, and have a huge vested interest in making the case that things are moving briskly along, yes.
I'm not saying there's not progress; that would be disingenuous, and the past five or six years have in fact been insanely productive for computer vision and machine learning in general. But articles that imply a relatively clear straight-line course from where self-driving cars are now to vehicles that can operate autonomously on urban streets in places with weather are... wrong, not to put too fine a point on it.
I'm incredibly bad at closing my italics today.
Anyhow, 61 cont'd: it is entirely possible -- in fact, per the history of AI and ML, it is likely -- for the technology necessary for self-driving cars to advance incredibly quickly along some vectors and yet remain hopeless along other vectors, so the actual end product will remain elusive even as it seems like things are progressing impossibly quickly towards that goal.
This article understates the problem, but is the kind of thing I mean.
Speaking of advancing quickly, do the existing collision-avoidance systems work enough to be worthwhile?
Probably? On the highway? They presumably do false-alarm (and are probably hackable)
I keep having to unfollow people on Twitter because their enthusiasm about self-driving cars sends me into a pit of panic and despair.
Sifu is reassuring.
61: A buddy of mine in grad school worked on AI and vision, and my conclusions from talking to him is that it's just a giant cluster. How you'd get a car to recognize a pedestrian, let alone recognize the body language of a pedestrian who is distracted and going to step out into the street, seems like a nearly impossible problem.
I'm basically a voice in the wilderness. I think the most likely outcome is that self-driving cars get deployed despite not actually working very well. (This is already happening, with Tesla.)
69: Will the inadequate self-driving lead to the consequences that everyone is excited about, like millions of people becoming unemployed, and public transport being replaced by something similar to public transport but more expensive and far less efficient? Or will they just be curios and status symbols forever?
68: that's about right. Pedestrian detection is making strides, which in this context means they get picked up, oh 70% of the time? Pedestrian intent (and theory of mind for driving generally) essentially nobody (but me! for all the good that does) is thinking about because it's so hard/so far from being practical.
I'm slightly more optimistic about self-driving cars than I was, mainly because of the advances in machine learning. It's not just that they've made advances, but that they have made advances in a way that I would have confidently predicted wouldn't work: general black box algorithms that work just by taking in massive amounts of data, rather than through lots of careful programming by humans.
56: because you've now realized that it isn't true (because I'm not insufferable) or because you're glad you prevented me from shedding an insufferable phrase?
73: yeeeeah but that only buys you some things. For object recognition, it's (relatively) fabulous. For other things it's harder to figure out how it applies.
It's too bad they can't just run on tracks. Tracks that run throughout your city, and everyone could share.
68 is the impression I've had from reading along with other problems mentioned the in article Sifu linked like correctly identifying dangerous objects so that the car doesn't do something drastic over a big ball of butcher paper or some leaves blowing across the road.
76: I know. We've emailed!
77: right?
because you've now realized...or because
Actually, I was using praeteritio to insult you.
car doesn't do something drastic over a big ball of butcher paper or some leaves blowing across the road.
It's worth noting that people aren't always great about this, either. To put it mildly.
i thought the medium-term semi plausible future for autonomous cars was to have dual-mode cars woth a human override, so you drive onto certain freeway lanes or something and the car switches into autonomous mode, you get off again and you drive yourself. Basically the equivalent of hitching your car onto a very well designed, personalized car-train that manages traffic and allows you to get off of at certain spots. I don't know how close even that is but it seems much less likely to raise ethical trolley problems than fully autonomous cars. Even for those cars they seem like an exotic rarity.
Obviously you need some kind of liability system whatever new technology you choose, but like any kind of liability problem they don't seem to be particularly usefully illuminated by philosophical ethics. I mean there are liability issues for currently-existing trolleys, too.
Self driving cars just need warning signs on top, like those "student driver" signs they put on the driver's ed cars. Then other people will know they need to watch out for the unexpected.
81: Most of that stupidity with people revolves a desire to not hit a small animal with their car, not a failure to correctly identify what objects are non dangerous to a vehicle.
Actually, I was using praeteritio to insult you.
:(
I brake for objects that aren't dangerous to a vehicle that go into the road. I always figure there's a strong chance that a kid is chasing whatever it is.
86: Sure, but I doubt you have a lot of trouble correctly identifying the possible pending danger of a tumbleweed blowing onto the highway vs a basketball rolling out from between some parked cars. My impression is that getting machines to do it is a fairly herculean task.
Tumbleweeds are very uncommon in these parts.
16 is not dissimilar from the circumstance of the only serious* accident I've ever been in: driving in the right lane of a four lane road in the rain. Spot slow-moving car ahead of me, start to slow, realize car is parked, because the road is abruptly becoming a two lane road with on-street parking, jam on brakes and try to get into the left lane. Probably fortuitously, the slick road and pre-ABS brakes prevented me from changing lanes, and I totaled both cars, but I was completely unhurt.
So in this case, no pedestrians, but I risked hitting someone in the left lane in hopes of avoiding the crash altogether. To the hypothetical in 16, I've absolutely been in situations where I was aware that someone behind me was following too closely, and started to calculate how to respond if e.g. the car in front of me abruptly braked to enter a driveway.
Oh hey, and when I was 4 years old, I was in the car with my mom (front seat, of course, but buckled; no booster). She stopped at a light, but the sun was low and the driver(s?) behind her didn't see the light in time, and a truck hit a Cadillac hit us. No pedestrians, but point being, trucks rear ending people is pretty fucking common, because their braking distances are ridiculous.
*I was also in a fender bender in which I attempted to pass on the right a car with its left turn signal on. The car turned right, into me.
Dammit, 89 was me.
I should clarify to the fender bender story, obviously I should have just waited for the turn to happen instead of passing on the right. But still, putting on your turn signal the wrong way: Jesus.
I've absolutely been in situations where I was aware that someone behind me was following too closely
My solution here is to troll the tailgaters by slowing the fuck down, then speeding up. It will take some tricky AI to pull that one off, particularly the part that evaluates the make and model of the tailgating vehicle to determine the likelihood of its driver being a gun-toting road rager, vs. standard idiot teen driver.
I've absolutely been in situations where I was aware that someone behind me was following too closely
My solution here is to troll the tailgaters by slowing the fuck down, then speeding up. It will take some tricky AI to pull that one off, particularly the part that evaluates the make and model of the tailgating vehicle to determine the likelihood of its driver being a gun-toting road rager, vs. standard idiot teen driver.
How the fuck did I manage to double post there?
Intent aside, ogged has in fact written (approximately) 56 before, in response IIRC to an LB post.
55 and 77 seem to me to be converging on what is basically the advantage of self driving cars, namely that they're as close to having really great public transit systems without actually having to give up a car centered culture.
I'm not sure if that's a bad thing or not - it's not so hard to imagine a kind of gentle progression along the lines of "how about we build some infrastructure - dedicated tracks, special signage, etc. in their own lanes on roads to make them work better" going to "it's inefficient when people are using self driving cars with different accelerations, braking abilities, etc. so how about we standardize them" going to "it's annoying to have to find places to park/wait to merge onto roads/etc. so let's just make them public and you can get in whatever one is closest", at which point you just have an awkward but way more comprehensive train system. It would be a ridiculous kludge but all the things that are appealing about self driving cars seem to me to just be things that are appealing about taking the bus only with a (way) better funded infrastructure and without having to sit near poor people.
Sort of on topic, I just passed a couple of kids rolling coal in a huge pickup.
Is kale more powerful than arugula in a situation like this?
Kale is better, because its got that hippy vibe that burns like holy water when it touches coal rollers. Arugula is more of an elitist thing.
I suspect that the real benefit of striving for self-driving cars will be the resultant technology that helps make regular driving safer. But man, I think it would be nice: just thinking of all these aging Boomers who won't ride public transit and won't stop driving is kind of terrifying.
It'll be interesting in the future when blue states start to require your car to be computer-operated, but red states decide that millions of deaths are worth it for FREEDOM and then everyone fights over whether red staters can drive their cars into blue states.
90. "But still, putting on your turn signal the wrong way: Jesus."
In MA we avoid this problem by never turning on our turn signals.
55 and 77 seem to me to be converging on what is basically the advantage of self driving cars, namely that they're as close to having really great public transit systems without actually having to give up a car centered culture.
It's the only advantage over when there's literally no alternative, but I'm not particularly clear on the supposed advantage when taxis exist.
Taxis don't exist everywhere. And very few places have taxis that are reliably available at anything close to all hours of the day.
I know that's true, but it has to be some kind of a regulatory problem. People want taxis, or taxi equivalents, badly enough that, e.g., Uber, is a big exciting deal. There's something broken about how most US places regulate taxis, I think has to be the answer, although I don't know what it is.
It's not entirely a regulatory problem. Some places are simply too rural to support any except the most causal service where you are totally at the whim of whatever the one guy who does that type of stuff is willing to do. And where the deadhead portion of the trip is likely to be very large in relation to the fare paying portion.
Given that I have not seen any stories about how Uber killed the jitneys, I think it's safe to say that what Uber has done here is improve the service for those who already had some service* and not that it has expanded service to those previously excluded from legal service.
* To be clear, this is not nothing. It was a big improvement.
Yeah, I admit that I wasn't thinking about real country, I was thinking about how people in lots of actual cities seem to think of cabs as useless and impractical.
Well, it costs $70 from my house to the airport. Fuck that.
Here I think of them as useless and impractical kinda creepy sharks that drive more recklessly than most.
112: Demand is low enough in a lot of smaller cities that the model doesn't really make sense for more than a trivial number of full time cab drivers, who operate almost entirely either out of the airport or on an on-call basis. When demand is higher, e.g. SAturday nights, etc., there aren't nearly enough cabs for everyone who in theory might want one (so most people don't even bother and make other arrangements). And the barriers to entry in terms o f a dedicated cab and licensing requirements means that a less than full time cab driver doesn't make sense. Uber solves this by having most of their drivers be people who aren't doing it full time. And there's almost no barrier to entry to being a driver. In that way it's possible for there to suddenly be tons more rides available during peak times.
Uber requires at car that is less than ten years old. That's a pretty big barrier to the jitney drivers.
A week ago Sunday we were taking the 24 to a concert and this older guy got on with his double bass and amp. Then got off about a mile later. He gets like quadruple transit rider points.
101: I for one couldn't resist responding to rolling coal with railing and kale.
Thanks, I'll see myself out.
I'll mention something tangentially related that pissed me off today. Local elections tomorrow, one city council member is a "vehicular cyclist" which means he thinks bikes should use car infrastructure with no special accommodations for bikes. He proudly mentions that he, his wife, and son have all been hit while biking. Apparently he's poisoned the views of other council members because he's considered the resident expert on biking in the city so they follow his lead on e.g. Does street x need a protected bike line (of course not!)
I've decided the best strategy for local elections is not to support any particular person but to try to take out the ones you hate.
We have a prominent local bicycle advocate (not an elected official) who takes the same view-- believes that bike lanes done incorrectly are worse than just riding in the road and sharing lanes with cars. Not everything he says on this topic sounds crazy. I do think taking measures to slow down traffic to safer speeds may generally be better than half assed bike lanes.
He believes that any bike lane is a bad bike lane.
122: god, the vehicular cycling people around here had so much influence for so long. It took forever to get bike lanes because of them. I mean, I'm a little sympathetic except for all the ways that it definitely means that fewer people ride and thus cycling is much more dangerous, which is to say I am not at all sympathetic.
Telling people not to bother doing something is always a good way to get influence.
Yup. "As a vocal cycling advocate, I'm here to tell you that the best thing you can do for me is totally ignore all other cycling advocates!"
Traffic calming vs bad bike lanes is a legit conversation for some streets -- a lot of the streets where I live are very aggressively calmed and that's probably a much better outcome than some half-assed painted lanes on otherwise awful streets. But sometimes - like arterials or where there's proper separated options on the table - that's not quite true and also why not just do both?
30: assuming self-driving cars get released in the near-term trolling them is instantly going to become a thing. See how long you can keep a self-driving car stuck at a stop sign by sticking your foot out into the street? Sure! Convince a self-driving car to stop short on the highway by tapping your brakes or swerving ever-so-briefly into its lane?
Take this a bit further: this is why the "hard" version of trolley problem programming could never work. Two or more people could assassinate any driver by running out in front of them.
69 & 83: I think the most likely outcome is that self-driving cars get deployed despite not actually working very well. (This is already happening, with Tesla.)
Self driving cars just need warning signs on top, like those "student driver" signs they put on the driver's ed cars. Then other people will know they need to watch out for the unexpected.
Yeah I think that the workaround will be to put the onus on other, human road users to use their in-some-ways-superior cognitive abilities to recognise the vehicles and stay out of the way. For instance, every designated route that the cars are allowed on will become the equivalent of train tracks - it's very much on you if you decide not to cross at a level crossing (or in this case, the traffic lights - which are presumably straightforward for the cars to recognise and use - or newly provided underpasses and bridges).
Note that this is pretty much what already happens in Beijing where, as a trial run for of badly driven self-driving cars, we have badly driven cars by regular humans who are effectively self-taught and spend a lot of time looking at phone screens instead of the road. All more vulnerable road users stay out of the way, and large barriers in the middle of (even minor) streets discourage crossing in unpredictable places. In future, the same dynamic will result in allowances being made for the self-driving cars: rich people want them, for convenience and status, and only poor people walk.
Not quite sure how I messed up the tags there, but both:
I think the most likely outcome is that self-driving cars get deployed despite not actually working very well. (This is already happening, with Tesla.)
and
Self driving cars just need warning signs on top, like those "student driver" signs they put on the driver's ed cars. Then other people will know they need to watch out for the unexpected.
s/b italicised.
122: god, the vehicular cycling people around here had so much influence for so long. It took forever to get bike lanes because of them. I mean, I'm a little sympathetic except for all the ways that it definitely means that fewer people ride and thus cycling is much more dangerous, which is to say I am not at all sympathetic.
I always wonder if the hardcore bike fanatics who savor the thrill of weaving in and out of traffic would actually stop biking altogether if we turned into Copenhagen and they were outnumbered 20-to-1 by random bike moms and bike geezers. It's not fun anymore!
In the UK, I'm sympathetic to the idea that bike lanes cause bad habits. In the absence of a lane, less-assertive [which probably corresponds to "less testosterone-fueled"] then cyclists tend to hug the side of the street, which means that they're more likely to get doored by parked cars, and more likely to have drivers try to squeeze past.
Segregated lanes everywhere would solve this, of course, but I prefer the [equally unrealistic] scenario that drivers in urban areas are brainwashed, Clockwork Orange style, with the understanding that other road users exist, have a right to be in the road and need to be respected.
Oh, and the above comment is entirely based on my experience of cycling in London for several years - at that time, bike lanes were often poorly planned, less-well-surfaced than the rest of the road and often used as one enormous gutter.
My thinking was something like: "Since, averaging over the course of two or three traffic lights, I'm going to be a lot faster than the cars, why shouldn't I cycle in the middle of the carriageway, where I'm safer?"
I can also provide Strong Opinions on helmets, red light jumping and lorries
I could have sworn we discussed this exact self-driving car scenario but I can't find the old thread to see if I'd say the exact same thing again.
I liked the way Berkeley did things. Other than the main arteries, roads would be blocked by planters every so often, where cars had to turn right, but bikes could go straight. So most roads were only for local traffic and parking, and so were very lovely to bike on. I think a less extreme version could also work, where say one out of every three or four roads was a bike road made impassable to through-traffic.
In Manhattan, one avenue on each side should be closed to cars with bike lanes and expanded pedestrian space plus allowing delivery truck access at very low speeds.
Ah, I think I'd say the same thing but maybe try for more clarity of expression.
Also repeating stuff I said in a different thread last year. Basically, I agree with Sifu, except I do think they're aware of and thinking about the pedestrian/cyclist/other driver state of mind questions, but without resolving them. The Google employees in the car I rode in talked about trying to model intent, like having the car slow down if it anticipated a car in the neighboring lane might be about to cut it off too closely to be safe. But I bet real world conditions don't present that situation often and unambiguously enough to machine learn it through large-scale data collection.
Also, based on my casual observation of self-driving cars on the road where I see them regularly near my work, it's not clear to me they've gotten much better since last year. They don't seem to drive differently than they did before but do still seem to drive differently than humans. Mind you, I can't see who's driving, so Google may really be employing some weirdly-behaved human drivers, but I see self-driving cars stop at times and distances humans generally wouldn't and not that long ago saw a bunch of human drivers pull out into the opposite-direction lane on a two-lane residential street to pass a self-driving car that wouldn't pull out to pass a stopped mail truck.
131: If it forced all the 'expensive thrill lookitmego' cyclists to switch to pogo stick commuting I think it would be a strong enough benefit to society as a whole that it would be worth all kinds of infrastructure spending, even in the absence of other benefits.
In some ways I get the argument about bad bike lanes - I've seen them and they really are counterproductive because they're hilariously unsafe/worthless and simultaneously make it look to drivers like anyone biking even near to the actual lane is some kind of selfish scofflaw asshole biking in a place they're not allowed. But this is the worst argument for not having lots of bike infrastructure, because it's not like it's hard to make bike lanes that work*, and even small steps - possibly even ones that are a bit counterproductive in practice - increase the number of people who feel safe biking around a city. And that's where the value of bike infrastructure comes from. More bikes on the road makes bikes visible to drivers who would otherwise ignore them, it makes the "you don't belong here I'm going to buzz you" drivers less and less frequent, and so on. The '08 spike in gas prices around here created a whole bunch of (temporary) cyclists, and it the roads almost immediately started feeling a lot more secure than they had a few months before. A lot of those people went back to driving when the prices fell back down, but even a limited bump like that had a longish term effect on driver behavior.
'expensive thrill lookitmego' cyclists
I'm not sure who this is. The guys wearing $300 worth of kit and riding $2000 bikes are not IME the ones weaving through traffic; those are the weekend warriors out on semi-rural lanes at 8 am on Sundays.
Anyway, vehicular cyclists are a critical part of the equation (in the US, at least), because A. they're usually the first movers on non-recreational cycling, B. they do, in fact, generally teach drivers that cyclists will sometimes be in automotive traffic lanes, which is important because C. there is no future in which protected bike lanes will take everybody everywhere they want to go, so you'd damn well better have a driving culture that recognizes that bikes will be on the street sometimes, whether those riding them are vehicular cyclists or just, you know, cyclists using their bikes as vehicles.
That said, vehicular cycling advocates tend to be arrogant, dumb, and self-defeating. Even an ounce of awareness would let them recognize that robust cycling infrastructure benefits them directly and indirectly, even if they choose not to use lanes.
131: Euro-style slow biking is, indeed, a future I fear. I often bike places because doing so is, literally, as fast as driving+parking would be. If I'm expected to be in this slow-moving herd of fashionably-dressed, helmetless people riding 25 pound bikes at 8 mph, I'll just take the fucking car.
In practice, I don't see this coming to pass, at least not in my city--the hills will always be a barrier to entry for that degree of placid cycling. But I've spoken with people who live in places where that's the norm, and there is definite hostility towards cyclists who don't ride slowly in designated lanes. Society-wide, it's a worthwhile tradeoff, but it would be miserable for me personally.
The expensive thrill cyclists don't necessarily look like the bike-race-LARPers, although those are annoying as anything as well. It was a response to the people described in 131.
The majority of people using bicycles as basic transportation fit into neither group, and tend to just be relatively low income cyclists quietly biking at moderate, not especially fast speeds along the sides of roads. I have seen bike-race-LARPers trying to commute too, though, which is always hilarious because they're still dressed up like 1960's Batman on two thousand dollar racing bikes only they're just kind of meandering around going slowly enough that half the time I end up passing them on a 3-speed.
[shrug] As I say in 142.1, 131 roughly describes me, and I average 15 mph on commute-type rides; nobody on a 3-speed is matching me. I'm on a decent bike with clipless pedals and, mostly, some sort of jersey for obvious comfort reasons.
Clipless pedals being pedals with clips but not completely clip-like clips.
There are times when I've absolutely been able to reach and maintain 15 mph on a 3-speed, and not a 25lb one either. I couldn't do it now for any serious length of time, but if you gave me a few months of biking a lot and I lost some weight I'd probably be able to again. They aren't that slow, they just have more stability than most other bikes and so you can ride along at lower speeds than most other bikes without feeling like you're in danger of losing balance.
I'm not sure when you describe yourself that you're really in the hardcore bike fanatics who savor the thrill of weaving in and out of traffic category though. Those guys are a pretty clear niche, and riding a road bike with clipless pedals isn't enough to get you there. (Also have you tried a sweater? For anything other than long distance or super-racing-speed* biking specialized bike pants/jerseys/etc. aren't really much better than normal clothes.)
*which is way, way, waaay faster than normal city speed limits.
I'm less much surprised that I screwed up the html there than that it didn't just go on forever. Anyway, the italicized part was supposed to be limited to the quote from 131, not the entire paragraph.
that it didn't just go on forever
I believe that this is Mr. nosflow's doing.
Also have you tried a sweater?
Huh? I'm talking about wearing light, breathable, wicking fabrics because for all but the briefest of rides in mild weather, I'm going to break a sweat. 15 mph is the average because I'm climbing big hills at ~10 mph; on level ground, I cruise at nearly 20, but more importantly, those hills would be 6 mph on a 3-speed, and I wouldn't be able to start as aggressively from stoplights without a fairly light, multi-geared bike.
Also, wearing regular pants on my bike tends to wear out the crotch at an alarming rate, so I wear whenever practical a pair of Levi's Commuters with reinforcement. Which is, obviously, not the sort of gear you were describing, but I've ridden too many miles to buy the "slacks and a t-shirt are just as good as specialized clothes" line. I used to wear exclusively boxers, and those suck on a (non-slow) bike as well.
Wicking fabrics are really great. I don't even own a bike*, but running and hiking have really taught me the value of all sorts of fabrics that I used to think were ridiculous.
* I do, but I got it for free and I've never ridden it.