s07e07: Safer Routes Home?
0.0 Sitrep
10:22pm on Sunday September 22, 2019. The plan was for me to go to bed a couple of hours ago, but I did not because I saw something on the internet. It wasn’t that someone was wrong, it was more that something turned out to be interesting and I’m afraid that kind of thing is fatal these days.
Oh. I treated myself to a Nintendo Switch. I have one simulation of a naughty goose that’s currently paused, sitting inside some sort of ARM environment, but while that one’s paused I’ve got two other actual live humans pretending to be naughty geese now, but they too are paused because they are sleeping. Unfortunately while I get to control the sleep/wake schedule of the naughty goose inside the ARM processor, the geese that are being simulated by two small humans are utterly beyond my control. The only saving grace is that I will not have to physically deal with them for the next few days, because I shall be in Sacramento. My partner will be goose-wrangler-in-chief and I shall advise her henceforth not to wear a tweed gardener’s cap or to have her house and car keys dangling from a chain on her back pocket.
While it is relatively cold and rainy here, it is not so in Sacramento. It’s jeans and long-sleeves and rainjacket here in Oregon, but when I land in Sacramento tomorrow, it is going to be Ridiculous Degrees (90f) on Monday and then Stupid (97f?) on Tuesday and Also Stupid (97f?) on Wednesday. I am the kind of person who wears shorts in the office, but in many significant ways, not like how someone like Cal Henderson wears shorts in the office.
Yesterday, we took the family to see the Oregon International Air Show where I tried very hard not to think about the procurement process behind the F-35 fighter, nor to think about the issues dogging its advanced supply chain management software. It is a very sleek plane and very loud and it totally did the trick in getting my children to think it was either Echo or Bravo from Planes. Well done both Lockheed Martin and Disney [in a previous version, this inaccurately said Pixar] for making the military industrial complex cute and endearing.
Oh, I also saw the Red Arrows which were also fun, they zoomed around a lot and had cute names for their maneuvers which is great until you realize what that precision flying is for when it’s not being used at the circus. (The pilots were very nice people, insert customary “you won’t believe how young pilots are these days!” exclamation to which the answer is: youngest pilot on the team was 33 years old).
Oh, if you’d like to, you can subscribe. This newsletter has gone paid now, so only one or two newsletters a week will be free and public. They’ll probably also be the best ones (don’t tell the subscribers), but I guess if you don’t subscribe you wouldn’t know. So complicated! Anyway, subscribe and contribute toward my Shoe Money or Help Me Feel Less Guilty And Practice Being Professional About Writing Fund:
On with the public display of thinking, I guess:
1.0 A thing
Only one thing today but it is long, so here is long thing:
1.1 Safer Routes Home?
[content warning: discussion of stalking]
Somewhere in my timeline the other day I saw something from Emily 艾米莉:
I wish Google Maps had a "I'm walking home by myself at night" mode that would show you a way with well lit paths and open spaces. It always seems to take you on narrow, dark pathways in cities 😕 [link]
And I saw another tweet about the idea from @theannalytical:
Someone commented saying that this is a harder problem than it seems. Google was built on solving hard problems, I hope the right people can see this problem and be inspired [link]
I feel like I’ve written about this before - can’t remember whether it was here or not, but when trillion-dollar companies complain that things are “hard”, what they really mean is that they “do not want to do them” or that they are “not that important”, or perhaps more tellingly, that “there is no [perceived] profit in doing it”. Because, lest we forget, Google is a company that went from zero to having a gigantic fleet of sensor-laden cars driving around every single road in as many countries as they could, just so you could look at shops and the less said about having more data to sell ads against, the better.
Anyway, I quipped that I’d totally trade the late, lamented and dearly missed Google Reader for some sort of “walking by myself at night” mode (and, I’m a man!), and that while we’re at it, it’d be nice if Apple Maps beat Google to the punch, just so we would have some competition in the mapping space.
All of which is to say, you could work backwards from a Safer Routes Home on Apple Watch product that could give you turn-by-turn directions without having to have a conspicuous phone out, provided you were outfitted with several hundred dollars worth of Apple kit. In which case you probably would want a safer route home, right?
Also if it came down to it, I’d probably prefer something like Safer Routes Home over Stand Reminders even though in the grand scheme of things, Stand Reminders are probably (he jokes) marginally more tractable and easier to do than something like Safer Routes Home.
One thing I’ve learned as I’ve become older and wiser and less of a jerk is to ask why things that seem at first blush obvious do not exist. Many times these ideas (“Why doesn’t the Department for Motor Vehicles Just Work?”) are borne out of a combination of naivety and sheer, brute, intentional ignorance. Other times, I think it’s a mark of the maturation of the Young Excited Technologist when they consider that there may be reasons or circumstances outside of technology (I know! What a shocker!) that affect implementation in the real world.
Example: if you use Google Maps for cycling directions in a place that supports it (I am on the west coast, but not in the Bay Area, so I have no idea what sort of coverage you’ve got where you are), you can get elevation information about the highlighted route. If you are like me, which is to say a person who is not as fit and healthy as he might want to be, as well as someone who has gazed upon e-bikes from afar but does not own one, then such elevation data is useful and a good predictor as to whether I’ll be needing a shower when I get to work, or when I get home. But! Google is quick to point out, in labeled text under the elevation diagram that looks a bit like a cute sparkline, complete with exclamation-mark-in-yellow-circle, “Use caution - bicycling directions may not always reflect real-world conditions”
In other words, there’s a liability question and, I guess, in this day and age, a horrific public relations question. Because accidents will happen and nobody, not even Google, with its infallible search engine and above-the-fold answers to questions that mean people don’t even click on search results anymore, not even them, not even Alphabet’s pride and joy can protect you and know the future with 100% certainty, no matter what they may tell you about their recent quantum supremacy.
So yes, they might be able to recommend you a route home that is predominantly in well-lit paths and open spaces, but you can bet there’s a lawyer or risk-adverse product manager saying: yeah, but what if something bad happens?
It’s complicated, right? As Lyzi Diamond points out, “Like all mapping problems, it is doable with the right data (and consensus on what "well lit" means) but you know... data.” My somewhat facetious reply (okay, actually facetious reply) would be to glare at the near trillion dollar company that, as mentioned above, has a fleet of sensor laden cars driving around, etc, etc.
And Lyzi’s right—of course she’s right, she’s a well-qualified professional software developer and GIS expert—it’s a data problem and it’s a defining what well-lit problem is, and a whole host of other things and those are… not necessarily harder problems than figuring out how to launch a network of balloons traveling on the edge of space that deliver connectivity to people in unserved and underserved communities around the world?
I mean, that’s pretty hard too. I think they might have even had to invent some technology for that.
(dirty secret: sometimes inventing technology is easier to do and used as an avoidance technique by people who find it difficult to face up to the really hard problems of negotiating human relationships and compromising between competing or complementary goals)
At which point the tweets collide, because here’s another Thing that caught my attention (sorry, embedded tweet alert because the image is important).
It’s a Strava heatmap visualization of runners in Chicago and as @MLKendrics points out, it’s a “vague, but fairly accurate map of White People in Chicago”.
Now, you might think that something like a runners route heatmap might be useful as a dataset for “safe route” because, well, people run along it. So you’d use something like that. But, oh sweet summer child of a product manager, did you consider what that dataset implies? One thing about data visualizations like these is that they don’t necessarily tell you the full picture about what’s not in the data. I can imagine that something like this is published and everyone figures hey, it’s implicit that this data is only from Strava users, and some might even argue that it’s explicitly from Strava users because duh it’s on the Strava website or there’s a giant honking Strava watermark on it, or it’s in an article on The Verge or The Atlantic or wherever and you can’t tell it’s quite sponsored content (sponcon!) but whatever it is, you can be sure they’re saying the word Strava so many times that it loses meaning, just like when you say the word Smock a lot.
Anyway, the problem is that it might be too easy to forget who Strava’s users aren’t. Or, if you’re Strava, you may have a vested interest in making sure that other people or your competitors don’t know who your users are. You may want your users to be a representative bunch, but it’s safe to assume that the userbase might be… trending a bit white? Professional? Middle-class-y? Maybe?
So now our product manager starts to freak out a little bit and say, well, is there a way to do this without , as the many Irenes say, “perpetuating racism and classism”? AND it’s 2019 so let’s just be generous and pretend there’s like a 10% chance our product manager has been paying attention and is thinking: ah, shit: Should I be thinking about how trolls or abusers might use this feature? How could this feature be used to hurt people or harm people?
Sometimes, I feel, there is not an answer to these questions, or rather, there is no right answer to these questions. There’s just a continuum and the attempt to pull out of easy black/white either/or thinking. One way of looking at this is: sure, a stalker could use a mapping service right now to figure out the route home of their victim. They could use this Safer Routes Home feature to figure out a well-lit, open space route home, too. But, you know, this is why we have user researchers and anthropologists and designers who go out in the real world and talk to people and find out about what people want and do not want and what solves problems for them and look at what they pay attention to and what they’re afraid of. Maybe this is better? Maybe it wouldn’t be too much better and it would give too much to the trolls? Maybe it just increase choice?
But a very real thing, a very real effect of this product would be a potential reinforcement or reintroduction of redlining. What’s safe for you might not be safe for me. I’m reminded of a whole bunch of well-intentioned but incredibly badly and insensitively (never mind in some cases overtly racist) attempts at apps that would crowdsource “safe” neighborhood maps.
I am reminded of Mikki Kendall’s talk from this year’s XOXO, and comment in the middle about moving back to Chicago and not being worried about any of the people threatening her on the internet: they were welcome to come to her front door because (and I have to admit I am paraphrasing badly here), it’s not like some random white guy on the internet is going to feel safe heading to her neighborhood of Chicago, even though she’s perfectly home.
So the product ends up either being a horribly racist and classist feature at launch if none of this is thought about—much like, say, launching health tracking features but not ones that are particular to just over half the population.
At this point our product manager is despairing. It sounded like such a good idea! That chap on the internet said he’d trade Google Reader for it, so it must be a good idea! But that thought about trolls and abusers keeps coming back - how do you control for terrible humans? This product manager is starting to have a bit of an existential crisis. Has software or technology ever solved the problem of terrible humans to a sufficient degree? No?
Well, like Michael Johnson and Jesper Andersen say, maybe it could work in a private community? And then we’re back to square one but it seems like we understand the problem a little bit better. Maybe the key is the private in private community. There is a bit of weird thing going on in the internet now about everything needing to be public. And in some places, what now feels like a useful examination of when concepts like whisper networks outlive their effectiveness or their purpose and their intended benefit either atrophies or stalls, hits some sort of peak. Because the whisper network that doesn’t break out and doesn’t result in action or curtailment of behavior doesn’t change anything — and this isn’t necessarily the fault of the whisper network itself. It can easily be the environment in which the network exists. Why bother breaking out if nobody will believe the claims of the network, right?
But then if the data weren’t public. If the data were small groups of people annotating OS-provided maps with their own metadata, their own safe-not-safe areas… would that be okay? Would that let you choose where you want to go, and at least introduce some more agency into how mapping software is used, or how you use mapping software? Is there a view where this just comes down to customization, or community contributed block lists, for example? If there’s an area I want Maps to consistently avoid when giving me directions, can I geofence that out? Can I import my friends’ geofences? Should I be allowed to?
These are big questions and initially they don’t seem like they’re big because they’re looked at abstractly through the lens of software and data, and it’s easy to forget that (and I’m now becoming a lot more familiar and comfortable with this metaphor) the map is not the territory. The map in this case—ha ha, the actual maps and the related map data—don’t tell us enough about real people, the environment they live and move in, and their needs, wants, emotions and, well, I guess, the psycho-emotional geography. What places feel safe to you? God forbid there be some sort of uber Safe Score per square meter of land… I mean… nobody would actually do that, right?
I mean, nobody would come out and say that there should be a utility function for each square meter of the planet’s surface, right? I mean, it would be entirely reasonable for the founder of Microsoft’s AI for Earth program and first Chief Environmental Officer to say something like that. I mean, yes, I’ll stipulate that “every square meter of land and water on Earth has an infinite number of possible utility functions” and may I be the first to chime in at this point that even though it’s a pretty good quote, a bit like saying “exploring the phase space of this problem”, even me with my lack of knowledge of the natural sciences and, say, marine science is grimacing a bit and going: square meter? Like, not volume? Like, not… individually addressable cubic volumes of seawater? That’s most of our planet? I mean, we probably have enough IPv6 space to give each cubic meter of water an IP address so I don’t see why we need to be thinking in a flat plane. Or whatever, geometry wasn’t ever really my thing.
But, really! Lucas Joppa, is quoted as saying his program is:
“a plea for people to think about the Earth in the same way they think about the technoloogies they’re developing. You start with an objective. So what’s our objective function for Earth? (In computer science, an objective function describes the parameter or parameters you are trying to maximize or minimize for optimal results.) [link]
Ok, so the good news is that Joppa doesn’t “think we’re close at all to being able to do this” where “this” is “[having computers] to produce optimization results that are aligned with the human-defined objective” for “every square meter of land and water’s utility function”.
I could go on and dear reader I will because here’s the next few quotes from Joppa:
For any location on earth, you should be able to go and ask: What’s there, how much is there, and how is it changing? And more importantly: What should be there?
I feel a little like I want to talk to Joppa like I talk to my kids. I get that you feel like you should be able to ask each bit of the Earth or Water, “Hey bit of earth or water! How’s things going? Becoming a bit more earthy? A bit more watery? Oh, the latter? Huh, any idea why? Oh, you think that’s us? You used to be earthy before but now it’s really watery? Oh, great. Thanks for letting us know!”
And sure, you can want to know that in the same way my kids want to be a naughty goose and eat McDonalds all day but I will not let them do that because down that route bad things lie. And it’s not like having a greater awareness of the interconnectedness of nature might not be a bad thing but lastly, that most important question. Gee! “What should be there?” I honestly have no idea how we’d figure that out! First we’d have to agree what we want and I think we can agree that, historically speaking, humans have not been that great at agreeing at what we want. It always turns out that that what some people want isn’t the same as what other people want and then there come the arguments with nuclear weapons or other assorted military instruments. Rarely, may I say, are such arguments concluded via non-lethal means such as the Football World Cup (arguably lethal from the construction industry point of view) or Eurovision (arguably lethal from… oh, you get it).
This idea that there is one answer feels like such a weird engineering/software engineering STEM-y type answer that any liberal arts grad would take one look at and purse their lips and say they’ll get back to you after they’ve finished this incredibly stressful doctorate of philosophy. “What should be there?” Is it me? Is this reasonable or am I the only one who sees just a tiny bit of weird hubris there or am I reading too much into this? I mean, what’s a good answer supposed to be? Could an answer be: “Well, pretty much what’s there right now, only a lot fewer humans and with the flora and fauna doing whatever they want?”
I have a feeling that’s not an acceptable answer.
On land, the data is really only interesting for the first few hundred feet. Whereas in the ocean, the depth dimension is really important.
OK. Nice to see we’ve acknoweldged the planet has… depth.
We need a planet with sensors, with roving agents, with remote sensing. Otherwise our decisions aren’t going to be any good.
Look. I might be going out on a limb here. It may well be, and I’d need to look outside and check, but there might be a chance that the planet is covered in sensors right now, and also with the roving agent kind. Some of them may even have remote sensing and some of the others might even has some sort of communication network, although we may yet not be smart enough to crack their network protocols.
We’re asking ourselves: What are the fundamental missing layers in the tech stack that would allow people to build a global optimization engine? Some of them are clear, some are still opaque to me.
I am LEGITIMATELY TERRIFIED, people! Why are we talking about a global optimization engine? What is being optimized? Has nobody ever seen that bit where that mouse with the funny hat on is given a global optimizing engine by a senior individual contributor before they go to burning man and is left with said global optimizing engine and before you know it, every single square meter of the earth is, I don’t know, optimally something?
Look. Joppa is most likely a really smart person who is not at all like the Farm in a Box lead at MIT’s Media Lab who did not know anything about farming and may, but I cannot find the reference at the moment, just may have been illegally dumping polluting efluviant into a river. So even after all of this which I admit is somewhat weaksauce, I will give Joppa the benefit of the doubt and assume that he has been horribly misquoted and this is what you get by having an interview in the uh checks notes highly esteemed industry association periodical IEEE Spectrum.
This isn’t to say that instrumenting the planet isn’t a bad idea. It’s probably a great idea. At the very least, NASA’s fleet of earth sensing satellites are providing us with a bunch of data that we’d probably rather not know about and stick our fingers in our ears about, but I guess now we get the equivalent of seeing Mai Lai again, but as the planet itself burns and melts.
The bit about the objective function and the—in my personal opinion—hubristic dream of being able to derive not only a utility function for each square volume of our planet, and also to compute the outcome of that utility function? I just cannot. There are no gifs.
Sure. Bring it with the sensors.
But the utility function part? Pure human problem. Good luck. Please don’t try to solve it without adequate consultation otherwise we’re all fucked.
This was supposed to be a quick twenty minutes but I think it turned out to be about an hour. I was going to include some links with this one, but hey, we all get a long essay smashed into your email client instead.
I hope you’re all doing well and that this week works out at the least passably for you and at the best like, well, all the good things. May no geese cross your path.
Best,
Dan
PS. I suppose I am opening myself up for notes on this one.