s3e35: Interesting Times

by danhon

0.0 Station Ident


12:09pm on Monday 28th November 2016 on a puddle-jumper from PDX down to Sacramento for the day. I am now the kind of person who irregularly commutes by plane for the day. It takes around 90 minutes to fly down to Sacramento on a prop (turbo-prop?) plane, so it’s a *bit* like, I dunno, going from Brighton to London. Of course, it costs around $150-300 depending on when I get the tickets, but it means that I can get down to Sac, have a meeting and get home and either be around for breakfast with the pre-schooler or bedtime with the pre-schooler.


1.0 Interesting Times


I haven’t written any newsletter episodes since the Event because, well, there’s been a lot to deal with and process, never mind the overwhelming feeling that there’s so much to *do*. And there’s been lots of doing going on. I’m on my way down to Sacramento to present a digital government strategy and, fingers crossed, will be helping to put in place some big, substantive changes that will build upon what I’ve been doing with the Child Welfare Digital Services demonstrator.


I guess there’s a couple big things that I’ll write about in this episode: i) something to do with the digital transformation work I’ve been doing in California and ii) something to do with some recreational whiteboarding I did in response to wondering out loud “what would an ACLU digital services team make?” So, let’s just see what happens when I start typing…


1.1 Fixing Government, Starting in California


(a.k.a. everything I ever learned about Digital Transformation I’m getting to do right now, and I need help)


I’ve a draft Medium (I know) post that I’ve been working on for a while about how a) I’m doubling down on getting things done in California and b) what that means in terms of real opportunities for people who want to build digital services that meet user needs (ie: come work in California and work on stuff that matters for everyone).


The thing about draft posts is that they feel like they require a lot of thinking and then they invariably get bogged down in a sort of how-can-I-make-this-perfect, *despite* me knowing that it’s better that something exists than not at all. So, I’ll try in this medium, the medium of stream-of-consciousness newsletter, to at least get some thoughts out to people who might already be interested. And for everyone who’s new, this is my Short Version of why there’s a big opportunity in California to do good work, right now.


Bullet-point back-story:

  • a long time ago, we (as in: society) decided that it’d be a good idea if we made sure that people who couldn’t afford food had help so that they could afford food. Enough people agreed that it was a good idea that we turned it from an idea into policy into legislation. Thus: food stamps, also known as SNAP, the supplemental nutrition assistance program.
  • these days, if you want to get food stamps, you need to apply for them. The process for applying for food stamps is, more or less, a clusterfuck everywhere. It is not user-friendly. In fact, it’s so user-hostile that you’d be forgiven for thinking, while you were applying, that *maybe the government is actually trying to make sure you don’t get it*
  • enter Code for America: one year, a fellowship team took a look at the process for applying for food stamps in California and saw that it was… not ideal? So they built a better process. Much better.
  • This better process impressed California’s Department for Social Services, who administer food stamps in California, (but the program is itself delivered by counties) so much that they wanted to find out a way to work with Code for America more.
  • As a way of testing that relationship, California’s Department for Social Services asked Code for America to do a review of a request for proposal of their new Child Welfare System, an estimated $500m, 7-year project to replace a 25 year old legacy system.
  • While at Code for America, I reviewed the request for proposal and saw that it was a fairly traditional monolithic (ie: one big project), waterfall (no iteration, spend a lot of time coming up with a spec/requirements, then spend a lot of time building exactly to that spec, then testing it, then… everything works?) project that, given all other monolithic waterfall projects, was most likely doomed to a) not delivering what was promised (not “working”), b) being over budget and c) being late.
  • My attitude in writing the review was one of a take-no-prisoners, “I have no fucks to give”, and following the advice and exhortation of Ed Catmull in his book [Pixar], I decided to experiment with being bluntly candid.
  • In the end, the State decided to throw out the $500m 7-year monolithic waterfall project and go with a modular, iterative, focussed-on-user-needs approach.
  • That project is now the Child Welfare Digital Service and it’s a demonstrator project for state technology projects focusing on meeting user-needs and delivering iteratively.


This new user-needs focused, iterative delivery approach? It’s working. The services have working code. They are working in the open, because it’s better.


In the previous approach, the Child Welfare project wouldn’t even have started working with a vendor until around mid-2017. The new approach has had working code since August 2016.


The timeline that Child Welfare Digital Services have been working to is unprecedented in State government, never mind potentially even Federal government.


There’s more: California now has the nattily named ADPQ, the agile developer pre-qualified pool. It’s a pool of agile software development companies that the State has pre-qualified by getting them to complete a technical exercise. The developers were given a problem and a public data set through a California Health and Human Services Agency API, a deadline, and asked to show their work on Github. Then, the State did a technical evaluation of the quality of their work *and* got users to try each of the services to see if, well, they would meet user needs. Many development companies submitted work that, bluntly, didn’t pass the user test.


We’ve been able to do this in California because right now, there’s unprecedented political will to do things better. Cabinet-level political will. The Secretary for Government Operations, Marybel Batjer, talked about this at the Code for America summit[x].


As an aside: the Harvard Kennedy School is putting together a case study on California’s Child Welfare Digital Service. As part of my interviews with them, they kept saying: hang on, this sounds too easy. If all you had to do to move from the monolithic waterfall model to the user-need focused iterative model was to point out the problems with the former, how come everyone else hasn’t done it? To which, I figured out in conversation – yes, part of this is to do with the fantastic political will to do things better and the recognition that the current way just isn’t working and isn’t delivering upon the social promise and compact of government. But, then I remembered another thing that I haven’t really talked about in public: my give-no-fucks candid review of the request for proposal indicted the entire structure of technology delivery in the State of California. A good way of putting it is this: the RFP that I reviewed was a genuine best-effort attempt by everyone involved *under the constraints they were in*. And ultimately, those constraints were imposed by state technology policy. My final set of recommendations addressed not necessarily the RFP but pointed out that the commissioning department – social services – actually had no power to substantively improve the outcome, even if they implemented all of my recommendations that they did have the power to implement. My final set of recommendations was, more or less, a revolution – not an evolution – of state technology policy. The thing is, I was pretty sure that I wouldn’t have a productive (ie: get anything done) conversation about the recommendations with the State’s CIO. So I didn’t. I had a conversation with his boss, instead. Again: no fucks to give. And his boss was someone that I did know was actively interested in finding a better way of delivering technology. In the end, after the decision was made to go ahead with the demonstrator project – the user-need-focused, iteratively delivered Child Welfare Digital Services – that State CIO ended up taking early retirement. California has a new State CIO now.


All of this is to say:


  1. How government works matters
  2. Policy and legislation doesn’t matter if delivery is terrible. That is, when government is bad at technology – horrible user experiences for applying for food stamps, antiquated technology for child welfare case workers, making it difficult to start a business – then for you and me, for citizens and residents, the *intent*, the policy or legislation doesn’t matter. If *in practice* it takes 150 screens worth of forms to apply for food stamps when it *doesn’t need to* then government has failed you and not met its promise in the social compact
  3. In this way, government fails everyone and people lose trust in government
  4. This isn’t just about cheaper technology delivered on time. Yes, government wastes billions of dollars on tech that is late, over budget and doesn’t work.
  5. Government technology that doesn’t meet user needs is an absolute failure of government *for* the people


So. I need help. I’m in Sacramento today for a huge digital services strategy presentation. Part of that presentation is covering how the state fills its talent gap, and making the case that the state people with the critical skills needed to deliver digital services that meet user needs. And yes, delivering those digital services will always involve working in partnership with the private sector.


If all goes well, California will be needing people very, very soon. We have a lot of work to do, and we’ve got a window to do it in. And if California can do it – and they’ll do it open source (most likely viral open source too, if they accept my recommendation) – then all America’s other states have no excuse.


I can’t promise a market rate salary. But, I’m recommending that the State is able to pay something close, I think


You’ll even have to spend some time in Sacramento. This isn’t the kind of work that can be done remote – yet, but it will be. Enough of this job is culture change as much as it is having the right skills and experience. Culture change means you’ve got to be there.


I’m collecting names. If you’re interested and you want to make a difference somewhere where you *can* make a difference, then get in touch. Who are the kind of people I’m looking for? Digital product or service managers. Designers. User researchers. Content designers (aka: writers, content strategists). Technical / architecture leads. People who know about legacy systems and look forward with relish to updating and replacing them. Is that you?


1.2 Recreational Whiteboarding


Ugh, I spent what feels like so long writing that stuff about California above that I’m going to do this one as bullet points as well.


  • I started off by thinking: what if organizations like the ACLU built digital services? What would they look like? What user needs would be in their remit to meet? What do digital services designed to (understand and meet the needs of users who want) protect civil liberties look like?
  • And then: why did I think of the ACLU? Three (in the end, four) reasons:
    • Reach (also known as “brand recognition”): the ACLU are well known and have large earned reach – when they talk, lots of people listen. Lots of people also want to hear what the ACLU have to say.
    • Results: the ACLU are known for getting things done. They take court cases and then take them all the way up to the supreme court to preserve civil liberties. People know that the ACLU are zealous in fighting for civil liberties in court.
    • Focus: they don’t try to do everything. For example, people know that the ACLU cares about first amendment rights and profiling, for example.
  • There’s lots more organizations like the ACLU, then, that you can apply these “interestingness” criteria against: the Southern Poverty Law Centre, Planned Parenthood, the Sunlight Foundation and so on.

Here’s my gut-reckon non-researched user need that I’d go out and get validated through qual research and, you know, building something. Let’s assume that I’m a user who cares about womens’ reproductive rights and making sure that they don’t get eroded. As a user who… cares about women’s reproductive rights and wants to make sure that they don’t get eroded, I want to:

  • understand *what* is going on a) right now, and b) at what level (eg county, state, federal)
  • understand *why* that’s bad or important or how it affects womens’ reproductive rights
  • be able to *do* something about it

For example: right now, Mike Pence may have *said* something about womens’ reproductive rights. What he said is bad or important because he’s going to be VP. Given that I care about womens’ reproductive rights, what can I do to prevent the bad thing that might happen? Is there something better to do than donating some money?


At this point, I defer to Farrah Bostic, whom I hope I’m able to now refer to as a good friend. Farrah and I had a chat about this, about a) building stuff that helps people *do* things and get things *done*, and in particular, b) the role that news and media plays in this. Farrah reminded me about That John Oliver Medical Debt Episode, the one where he a) explained what medical debt was, b) showed how many people it affected, c) explained why it was bad, d) did something about it, e) showed the rest of us how to do something about it.


Here’s a framework of how I think “media” can work better in the environment that we’re in. I’m not necessarily interested in an argument or discussion about whether this environment is a good one or not, merely about being effective in it.


John Oliver’s Medical Debt Episode is a great example of “news” reporting doing three new things packaged together when it only used to do one or two at best. Here’s the three things:


The What: John Oliver explained *what* medical debt is. News media thinks that its job is to do just this. Whatever, it’s a product of when it evolved. But I’d argue that in the end, a way to look at news in satisfying user needs is “I need to understand the world so that I can make good decisions”. Sure, you can drive a truck through “good decisions”. With my never-ever-an-actual-journalist naive reckoning, thinking out loud hat on, this is what happens when the news media sees its job to report facts and, as they say, “all the news that’s fit to print”. But! Facts are only useful to the extent that they inform decisions, right? Those facts, in the end, are useful to people when they help you *make decisions*.


The Why: John Oliver explained – using rhetoric, emotion and so on, as well as facts, why medical debt – as it is right now – is bad from a societal point of view. He did this in a seamless move. He moved from what straight into why. Or, even, he mixed what in with why all along. Traditional news doesn’t do this! It’s why we got a whole bunch of headlines saying Donald Trump has said some things, but none of the papers were able to report on page A1 that those things were Bad and he Shouldn’t Have Said Them. Newspapers are only able to have an opinion about these sorts of things – for historical, structural reasons? – in the opinion section. So you get a headline about a fact, that Donald Trump Said A Thing. And then, much later inside the paper, you get an op/ed that says, Hey, That Thing Donald Trump Said Was Horrible, We Shouldn’t Stand For It.


(An aside: do you know who’s good at the mixing the facts (and, in some cases, not-facts) with the editorial? I’ll give you a clue – it’s the kind of organization whose publisher is now the chief of strategy in the White House.)


(A double aside: the atomization of news and its insertion into feeds means that you lose the context and the brand values when a New York Times op/ed appears in your feed vs the actual story. When you mix news and editorial together in one story, then you retain your “message” when your message appears in the feed. This is Important, for Later.)


The Do: As soon as you’re able to admit that you can a) report facts, and b) take a value position on them (this is good/bad/complicated), then you may as well double down and go through to c) which is Do Something About It. Organizations like Breitbart and Fox News will exhort letter-writing campaigns and so-on, but traditional (liberal?) media will go about as far as having an annual campaign to, I don’t know, raise money for orphans or something I’M LOOKING AT YOU, FINANCIAL TIMES. But! Digital is great at *doing* stuff! Digital makes it easier to *do* things than ever before! So, what’s the do? Each news story can and should become a campaign and a call to action because, hey, you wanted engagement, right? John Oliver, again: the do was him *buying up debt*, forgiving it on TV and *showing you how to do it* too. He could’ve gone even further and made it *really easy for you to do, too*.


So, the open question: what infrastructure is needed to build new journalism and new civil liberties protection services that package up the what, the why and the do, complete with metrics and measurement to measure performance?


That’s what we put on the second whiteboard Jesper Andersen and I recreationally whiteboarded the other day. Maybe time for next episode.


2.0 Grab bag


  • Researchers at MIT have published a paper about a deep-learning artificial neural network implemented in photonic circuitry instead of traditional Von-Neuman/Turing silicon, although there’s a comment on Hacker News that there’s one part of the entire stack that hasn’t been implemented in optics yet. If I read their paper right, their experiments have shown a 2 order-of-magnitude (100x) increase in speed and a 3x order-of-magnitude (1000x) decrease in power.[0]
I was going to put more stuff in the grab bag, but ran out of time.

[0] [1610.02365] Deep Learning with Coherent Nanophotonic Circuits

Um. So. Got to go to a meeting. Send notes! Mwah! Bye!