Turning green streets red
February 2, 2020 7:30 AM   Subscribe

Berlin-based artist Simon Weckert went on a walk with 99 smartphones in a toy wagon. A video on YouTube, and some details, featuring a quote from Moritz Ahlert’s short essay on The Power of Virtual Maps.
posted by wachhundfisch (47 comments total) 28 users marked this as a favorite
 
There have been plenty of heist movies that have the nerd-genius member of the team hack the local traffic control system in order to clear one route or jam up another. I wonder if this kind of google-hacking-without-any-actual-hacking could serve the same cinematic purpose in a movie about a lower-resourced heist.
posted by jacquilynne at 7:41 AM on February 2 [5 favorites]


I absolutely love this
posted by nikaspark at 7:43 AM on February 2 [1 favorite]


I have a friend who regularly logs on to report a traffic jam on her street to try and redirect all of the cars cutting through her neighborhood. I kind of hope she doesn't see this.
posted by InkaLomax at 7:47 AM on February 2 [3 favorites]


This is very interesting, but are there really that many people RELYING on google maps for navigation? These streets are totally empty. It seems that there would surely be people who would be driving around without using any navigation. Or am I missing something?
posted by jonathanhughes at 8:00 AM on February 2




I'm guessing that he took his phones for a walk on a major holiday or some such time when there would be few cars on the street, specifically so he could know for sure that any traffic jam reports on google maps would be his own fault.
posted by moonmilk at 8:43 AM on February 2 [3 favorites]


Curious. But what is the actual connection between 99 phones and the red lines appearing on the streets in the map? This is not explained.
posted by njohnson23 at 8:51 AM on February 2


But what is the actual connection between 99 phones and the red lines appearing on the streets in the map? This is not explained.

Very generally, the phones are pinging local cell phone towers, with location tracking on, and contacting Google. Google then interprets a pile of phones moving slowly down a street, as an indication of congested traffic. I.e. it assumes that all the phones are in slow moving vehicles, and therefore there is a jam. This is then represented on Google maps in real time as a red line on the map in this particular location.
posted by carter at 8:59 AM on February 2 [5 favorites]


Google sees 99 phones on that particular street and thinks it's 99 cars. 99 cars on a street like that = traffic jam, according to Google's algorithm.
posted by Triplanetary at 8:59 AM on February 2


njohnson23: The phones are all running Google Maps. One thing Maps does is to send data about your location and speed to Google, which uses this information to make guesses about the state of the traffic. If there’s a bunch of Maps users in the middle of a street moving really slow or at a stop, it must have a traffic jam going on! So it tells everyone using Maps and that street turns red.

It’s prone to overestimating; it’s not at all uncommon to be out in the burbs and see a block-long “jam” caused by people stopping at the one intersection of a couple of highways in the middle of town.
posted by egypturnash at 9:00 AM on February 2


> But what is the actual connection between 99 phones and the red lines appearing on the streets in the map?

Waze and Google Maps monitor every app of theirs that's currently-running. If a phone running their app is moving significantly below the speed limit for the road their estimated to be on, Waze and Google Maps assumes that traffic is slow. Waze will follow up by asking users moving slowly to report on the density of traffic and cause.

Google Maps is going to register 99 phones all moving at walking speed as a pretty major backup.
posted by ardgedee at 9:01 AM on February 2


Why does Google assume smartphones are used by car-owners, not bus passengers?
posted by Monochrome at 9:06 AM on February 2 [8 favorites]


And here I am, in the Bay Area, thinking that road sensor data was being used to calculate traffic speeds. At least on the freeways. So now I can understand why my local streets here in San Francisco are also color coded. I guess the integrity of an algorithm is determined by the unlikeliness of someone filling a little red wagon full of phones and slowly walking down the street. Why doesn’t the obvious sameness of the location data on the phones trigger some other response than “lots of cars?”
posted by njohnson23 at 9:13 AM on February 2 [1 favorite]


Yeah, why are the phones assumed to be cars and not pedestrians? 99 people could walk over the bridge in a cluster and it wouldn't show red. Is it because the wagon is in the middle of the road?
posted by Pruitt-Igoe at 9:14 AM on February 2


I suppose a phone on wheels moves differently from a phone in a pocket, for one.
posted by Pruitt-Igoe at 9:18 AM on February 2


Is it because the wagon is in the middle of the road?

Yes I'm assuming that the GPS data being sent over the cell network to Google, is of sufficient resolution to identify this.
posted by carter at 9:19 AM on February 2


Android definitely comes with an Activity Recognition feature that, for example, allows apps to “switch to car mode when it detects that the user has started driving” (presumably using sensor data from the inertial measurement unit). But I suppose in this case, it might be as simple as switching the navigation mode to driving before putting it into the handcart.
posted by wachhundfisch at 9:21 AM on February 2


gps.gov says max resolution is about 16 feet. Maybe it's also possible to triangulate with cell tower data? And there must some good analytics at the back end to stop false positives emerging.
posted by carter at 9:22 AM on February 2


Google knows what lane you are in. This is why when he crosses the median, it flashes red for a bit on the opposite side. When he is sufficiently able to cross multiple times, it averages to the phones reporting traffic in both directions. Ergo, both lanes begin to report standstill traffic.
posted by Nanukthedog at 9:23 AM on February 2


Why does Google assume smartphones are used by car-owners, not bus passengers?

I find no level of government or any corporation assumes or even imagines people could be going somewhere by anything other than private car.
posted by ricochet biscuit at 9:26 AM on February 2 [13 favorites]


A 'Cautionary Tales' episode about blind allegiance to algorithms, with mobile routing as a principle example.
posted by j_curiouser at 9:29 AM on February 2 [6 favorites]


BRB, rewriting Neunundneunzig Luftballons to a more inspiring Neunundneunzig Telefons.
posted by ricochet biscuit at 9:31 AM on February 2 [19 favorites]


If GPS has that kind of resolution, wouldn't 99 phones in the same spot not look like 99 cars, but 99 passengers in a single vehicle? If so, maybe it's only showing a jam because his wagon is the only vehicle on the road and going quite slow.
posted by Pruitt-Igoe at 9:56 AM on February 2


As he walked down the street, I wonder if the cars that drove around him (or the cellphones in those cars, more specifically) could be used to make his wagon data into false positive hits?
posted by They sucked his brains out! at 10:26 AM on February 2 [1 favorite]


Why does Google assume smartphones are used by car-owners, not bus passengers?

I'm fairly confident that (for this purpose anyway) the traffic-monitoring is triggered by people actively using their GPS apps, which bus passengers are less likely to be doing. In other words, people using their phones for non-GPS activities like browsing Internet, streaming music, etc. aren't being used to estimate traffic.
posted by Greg_Ace at 10:28 AM on February 2 [3 favorites]


For the same reason, 99 people could walk over the bridge in a cluster and it wouldn't show red because a cluster of pedestrians is also a less likely group to all be using GPS.
posted by Greg_Ace at 10:30 AM on February 2


Performance art being what it is, I think there's a good chance this was at least partly faked. Someone on Twitter pointed out that the map displays on the phone screens look suspiciously uniform (none of them auto-rotated into landscape or getting inaccurate location data).

However, the phones do all appear to be in driving navigation mode, so Google Maps would naturally assume that they are in car traffic.
posted by mbrubeck at 10:48 AM on February 2 [3 favorites]


I love this. I'm going to choose to believe this works until someone replicates it and shows it doesn't work.
posted by 23skidoo at 11:16 AM on February 2 [2 favorites]


Why does Google assume smartphones are used by car-owners, not bus passengers?

When I use Google Maps on public transit, I select the public transit option so that it gives me bus routes. Same goes for being a pedestrian. When you select a non-car option, Google gives you directions customized for your mode of travel, and behind the scenes, probably takes that into account when estimating traffic probabilities.

I like this piece as art! I liked being confused by what was going on and then figuring it out. I’m definitely forwarding it to some people, who I hope will have a similar experience
posted by LEGO Damashii at 11:20 AM on February 2 [1 favorite]


I have a friend who regularly logs on to report a traffic jam on her street to try and redirect all of the cars cutting through her neighborhood.

I think Google, etc., easily moderate out this lazy method of abuse. Really easy to pick up on. The post's more effortful bad faith, that might have gotten through, at least for a bit.
posted by save alive nothing that breatheth at 11:53 AM on February 2


Why does Google assume smartphones are used by car-owners, not bus passengers?

Ever tried using public transit in the valley?
posted by mhoye at 12:44 PM on February 2


BRB, rewriting Neunundneunzig Luftballons to a more inspiring Neunundneunzig Telefons.

You and I in a little cell shop /
Buying dozens of sims with the money we brought / plugged them in at the break of dawn / till one by one they logged on.

Algorithms in the software / flashed the message / something’s parked there / it can’t tell it’s just one guy / as 99 telephones roll by.
posted by mhoye at 12:55 PM on February 2 [25 favorites]


We were driving through the middle of nowhere Nebraska on our way to Falls City for the 2017 Eclipse. 3 cars on the road. All of us going slowly because the town was hosting a Fun Run on the morning of eclipse day, and the road was full of runners. That was enough to make Google Maps go red for our section of rural highway. On those occasions when the runners cleared out a bit so we could hit 45 or 55 mph, the road turned orange almost immediately, then back to red once we had to slow for another pack of runners.

I was amused at how little activity was required to make the road go red, and how quickly Maps responded to any small increase in speed.
posted by Teegeeack AV Club Secretary at 1:46 PM on February 2 [5 favorites]


See also: crossroads (what to do), another phone-mapping performance art project.
posted by oulipian at 2:06 PM on February 2


The rough rule of thumb for GPS resolution is that it can pinpoint location to within 5 metres. So it can sort of work out which lane you're in (given most traffic lanes are 3-5m), but it does so by controlling for lateral jitter by averaging out values in your direction of travel.

Incidentally, this is why the fancy-pants hire scooters and bikes are often in places they shouldn't be, and the companies who technically are responsible for them don't actually know. I have an invisible property line where only a change in paving stones gives any hint that you might not be on the public footway. I have found dockless hire bikes on that spot, and contacted the operators to see if they could tell if bike number NNNNN was in a valid parking spot.

Since these bikes are at a halt, there's even less data to work out precisely where they are. And reflections from tall buildings can add even more jitter to the values.

GPS is pretty inaccurate, for individual readings. It takes a lot of work to estimate actual position from a stream of updates.
posted by rum-soaked space hobo at 2:11 PM on February 2 [1 favorite]


I like how he walks past Google Berlin.
posted by bendy at 2:31 PM on February 2


This is very interesting, but are there really that many people RELYING on google maps for navigation? These streets are totally empty. It seems that there would surely be people who would be driving around without using any navigation. Or am I missing something?

Google Maps does an amazing job finding backstreet routes that you would never, ever think of using when main roads are congested, because even with all the extra turns they're very slightly faster.*

And every time I bike past a queue of traffic, every phone on every dashboard I see is running
Google Maps. I have no problem believing that if you can jam Google Maps you can rid a street of traffic.

* This is of course a terrible thing if you happen to live on one of those streets, but why would not-being-a-dick be a consideration in the design of the algorithm?
posted by grahamparks at 2:56 PM on February 2 [1 favorite]


Google Maps does an amazing job finding backstreet routes that you would never, ever think of using when main roads are congested, because even with all the extra turns they're very slightly faster.*

I can tell when the main thoroughfare near my house is jammed up, because google sends dozens of people doing 70kph through the 30kph school zone on my street. Good times.
posted by mhoye at 3:00 PM on February 2 [3 favorites]


Moritz Ahlert’s short essay on The Power of Virtual Maps.

Is the last link possibly geo-blocked or something? I get a pdf-viewer scaffolding telling me I'm on page 0 of 0.
posted by PMdixon at 6:39 PM on February 2


Google Maps does an amazing job finding backstreet routes that you would never, ever think of using when main roads are congested, because even with all the extra turns they're very slightly faster.*

It does a piss poor job of it in Los Angeles. Google Maps insistently directs you to make multiple left turns against traffic to try to save 10 seconds here and there. You're risking sitting around for 3 minutes waiting for a light to change (since there are no green arrows at even many major intersections, given the age of the traffic grid) or for a teensy gap in oncoming traffic you can risk life and limb to gun through all in the hopes of shaving 15 seconds off your projected travel time.

No, google, I do not wish to turn left off Beverly against speeding oncoming traffic, travel one block down Oakwood, and then turning back to make ANOTHER left back on to Beverly at an intersection with no light because some asshole algorithm thinks it will save me 7 seconds. It will not. It will cost me minutes and possibly a limb or two.

I have feelings.
posted by Justinian at 12:25 AM on February 3 [3 favorites]


My #1 piece of advice for anyone driving in Pittsburgh for the first time is to not trust Google Maps at all. It will send you driving down a flight of pedestrian-only stairs with impunity.
posted by soren_lorensen at 5:56 AM on February 3


Yeah, Google Maps really needs a "stick to the main roads, no crazy taxi driver back roads routes please" option. One unexamined part of the sat nav age is that they've all been designed primarily to minimise journey time above all else.

(The opposite is true of bicycle route planners, which love to send you miles out of your way to use some muddy scenic path that's probably congested with dogwalkers. I don't think I've ever seen one with a "I have a train to catch, what's the fastest route on which it's technically legal to cycle and I'll take my chances" option)
posted by grahamparks at 7:37 AM on February 3 [2 favorites]


> Google Maps really needs a "stick to the main roads, no crazy taxi driver back roads routes please" option.

Of Waze, Google Maps, and Apple Maps, the Apple Maps app does the best job of keeping you off sideroads. I know Apple Maps' routing has been a punchline for years, but it's good these days, at least in major urban and suburban areas.
posted by ardgedee at 7:58 AM on February 3


This is of course a terrible thing if you happen to live on one of those streets, but why would not-being-a-dick be a consideration in the design of the algorithm?

Remember when Microsoft put their Tay AI chatbot on Twitter, and it was parroting racist and fascist slogans within four hours? If your algorithms aren’t designed to be robust in the face of abuse, they’re going to be turned into tools for abusers in sort order.
posted by mhoye at 1:50 PM on February 3


It does a piss poor job of it in Los Angeles. Google Maps insistently directs you to make multiple left turns against traffic to try to save 10 seconds here and there.

Nothing will match the horror I felt when I was trying to get to the 5 South ramp off Los Feliz Blvd coming from south of the street, and it sent me through a light cutting straight across los feliz, then had me go ONE BLOCK over and try to do an unprotected left turn through 4 lanes of traffic BACK onto Los Feliz blvd. I literally sat there waiting and cursing for half an hour.

I know this is so very LA specific, but if you know YOU KNOW.
posted by sprezzy at 4:43 PM on February 3 [1 favorite]


Along with the roughly 95% of samples within 9m accuracy for uncorrected* GPS, there are all sorts of issues with reflections, so in an urban environment the GPS signal is often bouncing off multiple buildings on the way to your phone, causing large skips. A few years ago I was working for a company that analyzed trucking company telemetry data, and we were doing all sorts of attempts to figure out which block a given GPS derived location was coming from (try to match vehicle direction, give benefit of the doubt for speed differences and one way streets, etc).

And the artist's description also says that he set the phones into car navigation mode. If every passenger in a car has their phone set to car navigation mode, it probably also screws up the data. The point for most system design, especially something like traffic in Google Maps where failure is a little bit of inconvenience that users chalk up to glitches in the system, is to not try to solve problems before they actually occur. Sure,there are ways to try to work around deliberate hacks like this, but thus far it hasn't been a threat model worth addressing, and probably still isn't.

About a decade ago I was working with some folks on some Federal research into vehicle to vehicle and infrastructure communications, and what the implications of that was, and it was fun to shoot holes into people's "we're gonna use radio signals for self-driving cars!!1!" enthusiasm: Yeah, if you use this for life-critical systems, a couple kids hit up a junk yard and play catch on a freeway overpass and people die. Don't do that. But, sure, feel free to do a little traffic suggestion that only a fraction of the drivers are gonna pay attention to.

* there are several ways to get quite a bit better accuracy from GPS,on a relatively broad scale, WAAS is an additional correction signal that broadcasts correction data for widely spaced terrestrial receivers, that helps some with adjusting for ionosphere conditions, and the new frequencies also let receivers do differential calculations on the refraction between two signals to try to get better. If you're building systems to do rough calculations from cheap cell phones, ya probably only implement that stuff when you have to.
posted by straw at 7:12 AM on February 4 [1 favorite]


I have heard from reliable first-hand sources that it's common practice among highway and other types of police to litter Waze with speed trap markers, as a traffic-calming/slowing device.
posted by not_on_display at 11:00 AM on February 6


« Older Thank you all for the gifts and the flowers   |   Mary Higgins Clark, 1927-2020 Newer »


This thread has been archived and is closed to new comments