Safecast at Stanford’s Medicine X conference

I gave a presentation about Safecast at the Medicine X conference at Stanford today. There was a mix up with some of my slides and I’m not sure if I hit all the points I wanted to, and several people asked me afterwards if the presentation would be online so I thought I’d post it for reference. The slides are largely photo-based (truth is I was going to go no slides until about 72 hours before the talk) so on their own they might not mean much. I’m including the raw notes/text for my speech below as well – forewarning, the text was written just for my own use and meant to be ad libbed a bit so take it with a grain of salt, and please excuse sentences that end weird or misspellings.


full speech text after the jump.

Last year, shortly after the Fukushima Daiichi nuclear plant meltdowns, when the public didn’t know the severity of the situation or the extent of the contamination, we were at a school in Fukushima with a bag full of geiger counters. Our intent was to measure the playground and surrounding areas so that people – parents, teachers – would be able to decide if it was safe for the kids. We had just gotten our hands on these devices and didn’t know how long we’d be able to do this, so thought if nothing else we should measure the schools. We also anticipated a little bit of political blowback and assumed going to schools first would give us a little immunity. Who would object to looking out for the kids? Turns out, the schools themselves. Instead of taking measurements we were being told to leave. We asked why and were told “because what happens if the levels are high, then we’ll have to do something about it.”

A few months later, when we’d developed a bit of a rapport with the locals, we were standing on the border of the evacuation zone speaking with the police officers guarding the line. This was the point where just past the barricade was considered unsafe to live. The officers standing there had no exposure protection at all except some paper face masks. The radiation levels at that spot were equivalent to 6 chest x-rays a day. These guys had been there for months. We asked them if they felt safe standing here all day no protection and they responded “our bosses tell us it’s safe so we have to trust that.”

In the early days of this situation ignorance was bliss. Since you can’t see radiation, many people hoped if they just pretended it wasn’t happening it would go away. Of course it doesn’t really work that way, and when people started realizing that they wanted answers. Answers that no official source would comment on. So people had to look elsewhere.

Today I want to talk to you about why speed is sometimes more important than perfection, and the importance of getting it wrong the first time. Let me put this in a little bit of context for you.

Immediately after the earthquake people were scared. They didn’t know what was going on, they couldn’t get accurate information… it was kind of a mess. Some friends and I who had other friends and family in Japan were in touch with those people and saw this frustration and confusion right away. We assumed that stress and general rattled nerves after experiencing a 9.0 earthquake and then being told that nearby nuclear plants were “having trouble” had hindered people’s ability to use google. We thought we could help find the information that was out there and maybe put it into a single, simple location for people to reference. Seemed easy enough.

Turns out, it was a little bit more involved than we’d thought. It seems the reason people couldn’t find accurate current information wasn’t because they didn’t know where to look, but rather because it didn’t exist. No one was paying attention to this stuff. That’s sounds crazy right? Yeah, that’s what we thought too.

OK that’s not totally true. We did find a little bit of data that we were able to put on a map that was live online about a week after the earthquake. But the data was only in tokyo, hitachi.. a few places only, and the reports where we found it had been designed to measure cosmic rays so the readings were taken 10-20 meters above ground and reported in this giant averages/ It was like trying to find out the weather and being told the average temperature in California is 75. so the info was sort of useless to begin with.

OK, so, Plan B. Maybe rather than just aggregating this data we should see if we could help collect it. We thought, we’ll buy a bunch of geiger counters, pass them out and get people to send us back the readings, then publish those so people would have something to consult. It was a little bit more tricky, but seemed doable.

We did a kickstarter and raised some cash to buy these geiger counters only to find out they didn’t exist either. Or rather, the entire whole world supply of geiger counters had sold out within a week. Mostly to people in Ohio. Hrm. This was turning out to be a little more involved than we’d planed.

Somewhere along this line what began as a email with a few people CCd, and then a few more people CC’d, and then a few more… got moved to Skype and became 24 hour group chat that we kept pulling new people into. Oh! What’s the difference between Alpha, Beta and Gamma radiation? I know a guy who can answer that, let me pull him in. Oh! Where in Tokyo can we hook up with people who know how to solder and build stuff, I know a dude, I’ll get him in here. Oh! How do find volunteers in Fukushima, my friend is dialed into some groups there, let’s get her in here too. My friend Joi Ito called it “a cross between govt situation room and news room” And that’s pretty accurate. this lasted for 2 or 3 months. Yeah, lots of coffee at work here.

So. Plan C.

Who am I kidding? Plan? We were firmly just making it up as we went by this point.

So we finally got our hands on some radiation detection devices but not the hundreds we’d hope for and passing out 5 or 10 geiger counters didn’t seem like an idea that would really scale. My friend Ray Ozzie had this brilliant idea.

What if we attached them to cars?

This could work! If we drove around taking readings we could cover way more ground than just taking single readings by hand. OK. So how do we do this. Got it! We take one, duct tape to the car window and take a photo of it with an iphone every 5 minutes and then used exif geo tagged data to know where the photo was taken.

This totally worked.

It also totally sucked.

Really, can you imagine anything more obnoxious?

OK, so we’ve got a proof of concept here, but it needs some serious refinement.

By the end of the week we’d come a long way. We took that geiger counter, stuck it in a weather proof box with a GPS module and an arduino – attached that to the outside of the car, and then ran a line into a laptop that would display and log the readings and pair them with a GPS coordinate. One every 5 seconds as we drove. We called this the “bgeigie” because it kind of looked like a bento box. We did the first drive with that system that weekend.

The people we met on these drives had no info, and were angry and scared. Like the people at the school I mentioned earlier. They didn’t know who to trust and were skeptical of everyone. But we were doing something different. We weren’t just taking readings and leaving, we were giving people back the data we collected. And in sharing that data the people who were initially apprehensive became out evangelists, and some of them have gone on to be some of our most dedicated volunteers

We started making more of these bGeigies, but every one we made had improvements from the one we made previously. We were able to scrap the laptop and put everything in a self contained box outside the car. We made smaller, more portable devices, attached to them to cars, to bikes, to like a neckless/harness/walking rig thing.. University support got us access to evacuated areas, right up to the daiichi gates.

People donated cars, volunteers would drive these things around, we partnered with a shipping company to attach our devices to their delivery vehicles. We kept revising the devices to make them simpler, which increased volunteer options. The first devices we made were the size of a suitcase and took a full weekend to build, and cost close to $2,000 each, all things considered.

This (hold up nano) is the latest iteration, it’s under $400 in readily available parts and can be built in a few hours. The designs for these are open source so you can build your own if you want, and we’re about to release a kit with all the parts to make that even easier for you.

We now have over 100 bGeigies out in the world collecting data. We’ve also installed over 300 static sensors monitoring a single location all day long – these are kind of an early warning system to detect if something changes. Keep in mind that the reactors at Fukushima Daiichi still aren’t stable.

6 months into this we published our 1 millionth data point.

This month we’ll surpass 4 million.

Now all of this data is published completely openly using a creative commons zero license, basically public domain. Anyone can use it for anything, with no licensing restrictions at all. On our website people can see a map of the data we’ve collected, download the full data set, or the more technical out there can interface with our API to upload their own data or query our dataset for specific parameters. And because we map to an individual GPS coordinate, you can look up your address and if we’ve mapped that area you can see the reading right outside your door step.

We publish data immediately after it’s collected. We were the first to publish the data showing that the wrong areas were evacuated. That some areas within the 20km mandatory evacuation zone were not very contaminated at all, and some areas far outside of the voluntary 30k evacuation zone were quite hot. And likely some people were moved from places close to the plant with lower levels, to places further away from the plant with much higher levels.

I saw we were the first to publish this, but we weren’t the first to know it. The day after we published our maps showing this, the Japanese government released maps showing the same thing, and have recently received a lot of criticism as evidence has come to light showing that they had this data within weeks of the event and sat on it.

Yahoo Japan has been using our data since last year to power their radiation maps, which is cool, but a few weeks ago that was completely surpassed in all levels of awesomeness when Fukushima Prefecture Government launched official their maps driven by Safecast data.

Now the “official” dataset from the Japanese Government is growing too. In fact they DOUBLED the data they have released just last month, that dataset it now around 30k data point. And it’s provided in handy PDF format. And in 1km mesh. And no info about how readings were taken or with what devices, so you’ll just have to take their word for it. Oh, and that data is copyrightten, so if you want to do anything more than just look at it you’re gonna need to ask permission and get some approval first.

The more of these devices that get out there, the more data we’ll have coming back in. It turns out that the data that wasn’t available in Japan isn’t available anywhere else – and the readings we’re taking now are so much more precise than anything taken previously they are hard to compare anyway.

So we’ve increased our scope and are now trying to map the entire planet, so we’ll have a global baseline that can be referenced in the future should that be needed.

And the data is already being used by more than just people. There are doctors in Japan comparing our data to the data they have about where symptoms where reported and when to see if there are overlaps. There are researchers at Harvard who collected a huge amount of data about stress levels by data mining social media for certain keywords – they are contrasting our data to theirs to see if fears and reality match up.

Across the board the feedback we’re getting is that researchers are ecstatic to get access to raw data like this. Nothing like it has ever been available to them.

There’s a number of things worth noting in this

Most obviously, we saw how the lack of information and lack of transparency from official sources caused people to look to alternatives and in that create an even better system than what was in place before hand.

We saw how the simple desire to get some info for friends and family has spawned developments in hardware, software, data formats, visualizations.

We saw how openness and willing to make mistakes and fail publicly helped us move incredibly fast. We’ll often build a device, send it out with volunteers and publish the docs right away, which gives us immediate feedback on usage and allows us to revise and release an updated design a week or two later.

By comparison, the closed approach moves much slower. I was in Tokyo a few days ago attending Radiex, a trade show for the industry that surrounds radiation monitoring, analysis, cleanup and decontamination. The later being mostly giant trashbags for storing the top level of dirt – a solution that works fine if you are talking about a localized situation, but doesn’t really scale when you are talking about a whole country.

The “monitoring” section was an eye opener to what happens when the design process takes place behind closed doors. There were no shortage of hand held devices for consumers that only detect gamma radiation – being marketed to people living in the aftermath of an event where the most widespread fallout was a beta emitter.

The thing that stood out to me was a trailer the size of a mini cooper that you could attach to the back of a truck and tow around a city. It had a sensor on the bottom to measure radiation and onboard GPS to map out readings as you took them. Sound familiar? Also, it wasn’t yet available, but they were taking orders. Or rather, hoped to be taking orders. Many of the customers they were hoping would be lining up had been volunteering with Safecast for more than a year and already had better data and the trailer could collect.

The people at that booth thought this trailer was incredibly innovative.

In looking at what we’ve been through in the last 18 months, we saw how those things we created for radiation could work to help monitor other environmental factors. We looked around and saw that air quality suffered from a lot of the same problems as radiation. Few people monitoring it, easy to find reports but hard to find original data collected, published data is often reported in huge averages – we thought perhaps what we learned mapping out radiation could be a benefit here as well.

so we’re now working on a design for a sensor that you could put at your house and it would give you a real time breakout of what you are breathing, and how that changed on an minute to minute basis. if enough other people help us distribute these devices we think it’ll paint a very interesting picture.

What we learned from Fukushima is that knowledge relieves stress. When people don’t know the facts they assume the worst. And assuming the worst is stressful. People get depressed, they get sick, they give up, they kill themselves. It’s ugly. Giving people actual information allows them to make educated decisions based on facts rather than speculation. It empowers them to have some control over their fate, and we’ve seen even when the actual info is bad news, at least it’s something concrete and this makes people feel better because now they know.

To loop back to the school I mentioned in the beginning, people in Japan are no longer happy to ignore the situation. They aren’t content to just sit silently by and be told “don’t worry about it.” They have questions and are demanding answers. I commute to Tokyo about once ever 5 weeks and the change I’ve seen has been night and day. We now get invited to schools, not only to take measurements but to give seminars about radiation and tutorials on how people can take their own readings. People no longer depend on official sources for information, and now actively participate in collecting and sharing the data with each other. The government wouldn’t help, so they routed around them. That mindset shift will never be undone, now people know there is another option.

The main point here is that all of this was possible because we moved fast and weren’t afraid to fail, so we did it often. Our failures and mistakes helped us learn faster and let us get to the next step long before we could have if we weren’t field testing things hours after they’d been initially conceived. We welcomed the feedback about what didn’t work and fixed it for the version we sent out the next day. Working beat perfection. Sharing all this will the community every step of the way gave us insights we never would have gotten if we’d been behind closed doors. Open beat closed. A handful of motivated people with the desire to do something and willingness to learn along the way beat huge industries, vast networks and mountains of cash.

The lesson here is don’t worry so much about getting it right the first time, but rather taking action and being persistent.

Thank you.