If you’re like me and you’re building a social app and you don’t have a lot of marketing experience, one thing you may try to get traffic is reddit ads. And on reddit, people can comment on your promoted post.

The reddit community didn’t like my app idea and offered immediate and specific criticism. reddit is an unflushed toilet so look past the tone & language:

Why, so we can be more closed minded and literally f*cking in an echo chamber?

This user was talking about filter bubbles. I dashed off a reply using my profanity-free powers of debate, but have been thinking ever since – if large groups of people have the same information, is that a bad thing?

Bubble, bubble toil and trouble

The ‘filter bubble’ is the idea that ranking tools use popularity & demographics as inputs, and therefore will lead to polarization of opinions on demographic lines.

Filter bubble pushers fear polarization, not manipulation. When mass media happened during and after World War 2, irresistible persuasion was the concern. I’m lifting from a 2017 survey of web search behavior run by the Quello Center at Michigan State1:

  • Since WW2 the news media as shapers of opinion have been a focus of concern in democracies
  • Use of radio & TV for propaganda and persuasion has been a research question for a long time
  • “mass news media were found to be relatively un-influential in changing the opinions of individual voters (e.g., Berelson et al. 1954; Campbell et al. 1964)”
  • But there may be subtle ways in which mass media does manipulate opinions:
    • via agenda-setting (McCombs and Shaw 1972)
    • via the ‘two-step flow model’ (Katz and Lazersfeld 1955) in which media influences opinion leaders who in turn influence their community

The good news is that online news consumers are at least checking their references:

50 percent of users report they “often” or “very often” use search to check facts. The findings indicate that misinformation can fool some search engine users some of the time, but that most users (over 80%) are sufficiently skeptical of information to use search to check facts. Nevertheless, fact checking is limited by skills in the use of search. Many users could benefit from support and training in the use of search.1

Paul Baran, a network engineer from the 60s who dabbled in sociology, was concerned by his own version of the filter bubble:

government requires a measure of cohesion of the ruled … With the diversity of information channels available … nearly every ideological group, from the student underground to the John Birchers, now has its own newspapers. Imagine a world in which there is a sufficient number of TV channels to keep each group, and in particular the less literate and tolerant members of the groups, wholly occupied? … Will they ever obtain at least some information through the same filters so that their images of reality will overlap to some degree?2

You can criticize news in general instead of algorithmic rankings. Francis Crick, best known for photographing chromatin with x-rays, was anti-news:

He seldom read newspapers, because working in intelligence had convinced him that most stories never reached the press.3

Nassim Taleb, the economist who wrote The Black Swan, is ‘anti-idea’ in that he believes social science theories are summaries that are less useful than their raw data. He may also be anti-news4.

My take on Parisier vs Taleb (without having read much by either author):

  Parisier Taleb
News is: opinion observations summarized to draw false conclusions
News dooms society via: sowing discord & internal conflict justifying bad decisions, hiding actionable information
On diversity of ideas: polarization and balkanization are dangerous; we have too much diversity conclusions are less useful than raw data, we want diversity

I saw a review of Taleb saying he’s “annoying but not completely wrong”.

Writing the news is a process of digesting facts into a compact package. I wonder if, like most digestive processes, the nutrients go to the person doing the digesting and not to the people who study the final output.

My read of Parisier’s take on information, though I doubt he writes this explicity, is that it tells people what to do and it had better tell everyone to do the same thing. Taleb’s take on information is that it supports creative, non-binary interpretations of problems and searches for solutions.

I used ‘collaborative filtering’ in the title of this post instead of ‘filter bubble’ because collaborative filtering is a more neutral take and also describes a wider variety of media with two common properties:

  • winner-take-all – early success can pin an article or product to the top of a keyword’s results.
  • ‘personalized’ in that demographics or past behavior is an input to the model. I don’t love the word ‘personalized’ – ‘pigeonholed’ is a better description of how these systems are categorizing users.

Group decisionmaking in Flint

This post is about how reliance on summarized information can lead groups to disaster by delaying recognition of a problem and delaying corrective action.

What does it mean to say that groups make decisions?

While I was researching this post, news broke about a measles outbreak in a Somali immigrant community5. There are high rates of autism in Minneapolis and so ‘anti-vaxxers’ gained a foothold there. And the Somalis were receptive to their message, bringing vaccination rates down to 40% for Somalis in the city.

This is a group decision because:

  • Whether or not the group consulted each other, something about their shared culture or information made them susceptible to the message.
  • (I’m guessing) community ties between Somalis led them to discuss their decision. Once the first few families made the choice not to vaccinate, others probably followed more easily – i.e. the decision has momentum.
  • Disease impacts groups differently than individuals: If one person doesn’t get a vaccine, that’s their problem. If a whole community doesn’t get it, disease can spread.

The case I decided to write about instead is the Flint water crisis, because there are clear timelines showing when information led to inaction and, eventually, action. If you don’t know the details: a financially motivated changeover from Detroit water to local water supply was mismanaged, leading to several instances of bacterial contamination (‘boil water advisories’) as well as acid-related corrosion that caused lead to leach out of lead pipes that were still installed on some blocks.

Compressed timeline of Flint:

  • 2014:
    • April: Flint switched over to Flint River water
    • October: GM stopped using Flint River water because it was corroding metal parts
  • 2015:
    • before February: LeeAnne Walters, a Flint resident, did a ton of homework and contacted the EPA6
    • The EPA spars publicly for months with Flint’s city government and the MDEQ.
    • Virginia Tech published increasingly damning water lead tests, and a local hospital published blood tests of children with lead poisoning.
    • October: the state reconnected Flint to Detroit water.

My point is to show that the trigger for change was one person, LeeAnne Walters, going through public data and learning about water treatment. She contacted the EPA and the team at Virginia Tech that did most of the water testing in 2015.

A Pew study shows water-related searches spiking after a boil water advisory in 20147. Interestingly, search activity is silent. If I’m having issues with my water and my neighbor is too, I want to be connected. Privacy is counter-productive here.

Flint in fact had two group decision events:

  • Government planning during a budget crisis led to switching water supplies in 2014
  • In the crisis that followed, activists and experts created a loud minority. Expert attention identified the problem and, eventually, fixed it; though not soon enough to avoid serious health consequences in Flint.

Historian Barbara Tuchman said that once a policy has been adopted and implemented by a government, all subsequent activity of that government becomes focused on justification of that policy8. I prefer LeeAnne Walters’ grassroots leadership.

Information & infrastructure

When De Tocqueville visited a young America he saw:

here the people of a district are assembled to discuss the possibility of building a church; there they are busy choosing a representative; further on, the delegates of a district are hurrying to town to consult about some local improvements; elsewhere it’s the village farmers who have left their furrows to discuss the plan for a road or a school.9

But between then and now we offloaded building and planning schools and roads to government. ‘Self-government’ and ‘government by the people’ mean something different in 2017 than they did in 1831 when De Tocqueville visited.

We’ve created an unfortunate divide between the planning, the labor, and the traffic; in 1831 it wouldn’t be unusual for the same people to design build & then use a road or public building. It’s unfortunate because when designers and users are disjoint sets it takes longer to understand what’s going wrong when something goes wrong. This is part of the appeal of open source software – it’s easier to fix – and the reason people are using Ukrainian viruses to hack their John Deere tractors10.

I don’t know why the Flint water supply engineers got the corrosion treatments wrong and why they shouted down the EPA. I believe that if anyone living in the bad parts of Flint knew the ins and outs of municipal water systems, the problem would have been caught and fixed more quickly.

You can also consider this a consequence of the drawdown of local news. A recent Politico article claims that as print ad revenue goes away, local newsrooms in flyover country are the first to disappear11. LeeAnne Walters’ investigative work would have been someone’s job 30 years ago, instead of requiring an Erin Brockovich-style local hero.

We are John Adams’ grandchildren

I must study Politicks and War that my sons may have liberty to study Painting and Poetry Mathematicks and Philosophy. My sons ought to study Mathematicks and Philosophy, Geography, natural History, Naval Architecture, navigation, Commerce and Agriculture, in order to give their Children a right to study Painting, Poetry, Musick, Architecture, Statuary, Tapestry and Porcelaine.12 (video here)

Never mind that he was justifying to Abigail why he isn’t sending her descriptions of the “Temples and Palaces, Paintings, Sculptures, Tapestry, Porcelaine, &c. &c. &c.”. It’s a great line. Read more of his brags in appendix 1.

If watching 4 hours of TV a day counts as studying painting and poetry, then as Americans in 2017 we’re doing it.

Some news is actionable in our daily lives but some of what we read, even stories that are thematically ‘politics & war’, is entertainment: stories about far-away events which probably won’t affect us and may not be under our control. At best they can tell us we’re in the wrong place, to change cities or jobs in order to escape a crisis; but crises have a way of sneaking up on us without advance warning.

Pliny the Elder said ‘Ex africa semper aliquid novi’ (out of africa always something new. There’s also a Meryl Streep movie about this, I think). I have a citation saying he’s borrowing from an Aristotle line “Always something fresh in Libya”13 which was a subtle way for Aristotle to say he didn’t believe a traveler’s fantastic report.

He was right to be skeptical. We can all afford to be a little more skeptical, at least about topics that affect our health & wealth. LeeAnne Williams studied investigative reporting so her children & neighbors would be able to study painting & poetry (and drink clean water). My point is that good information matters to us as much as good anything else.

This can kill you

The USMC’s Advanced Situation Awareness program, originated by NATO’s own Jim Mattis, aims to make Marines “keenly aware of the details in their surroundings” and “alert to unusual environmental changes” by teaching skills that used to come in to the corps through Boots who grew up hunting14.

Hunter-gatherers at the edge of the neolithic revolution must have all been data scientists; paying attention to threats and where to gather food, moving from summer to winter range to cope with the seasons. Depending on where you grew up, you may not have the concept ‘this can kill you’, but people who have to catch what they eat, preserve it for the hard months, and live outside with dangerous animals are always thinking that.

There are consequences to screwing up.

For my job I’m around data and once in a while get the chance to open a dataset and explore with no agenda. When I do:

  1. I usually learn something interesting (albeit not useful) when I open a new kind of dataset (a new financial market, a new kind of user-data from web); even if it’s just about time of day structure of activity, confirming a behavior I expect, or a simple correlation.
  2. Whatever property I think I learned, if I’m foolish enough to remember it, when I come back to the data 6 months later, it’s changed.

Both experiences are humbling and useful. Combining the ‘this can kill you’ perception skills of situational awareness training with the ability to understand datasets and remark on their changes is a way to free communities from the tyrrany of indifferent local news coverage by wire services.

I’m proposing:

We should add data dumps to the daily ritual of reading the news. Instead of tweets like ‘TIL what TIL means’, we’ll see ‘TIL commuting 15 minutes later is safer and faster – from data’.

How much more quickly would lead testing have started in Flint if everyone was in the habit of using data to answer questions instead of pre-digested news? ‘Daily data’ is an experiment we should try as a society.

Appendix 1: Other John Adams-isms

Source for all of these is the Jefferson-Adams letters. Start reading pg290 when they get back in touch as old men.

The thing to know about these guys is Adams wrote Jefferson twice as often and much of the exchange was Jefferson trying to keep up with Adams’ stilted metaphors and often vague requests for information about books and articles mentioned years ago in passing. The conversation was made more difficult by the inconsistency of the mail, so that letters could arrive out of order and letter+parcel might not arrive at the same time.

  • pg294 oh that I had devoted to Newton and his Fellows that time which I fear has been wasted on Plato and Aristotle, Bacon … with twenty others upon Subjects which Mankind is determined never to Understand, and those who do Understand them are resolved never to practice, or countenance.
  • pg295 I have read Thucidides and Tacitus, so often, and at such distance Periods of my Life, that elegant, profound and enchanting as is their Style, I am weary of them. When I read them I seem to be only reading the History of my own Times and my own Life.
  • pg296 I have a Complaint that Nothing but the Ground can cure, that is the Palsy


  1. Dutton &al, Search and Politics  2

  2. copied in Matt Novak, TV will tear us apart 

  3. Nicholas Wade, A Peek Into the Remarkable Mind Behind the Genetic Code 

  4. Dobelli, News Diet. “My good friend Nassim Nicholas Taleb was one of the first people to recognize news consumption as a serious problem.” 

  5. Megan Molteni, Anti-vaxxers brought their war to minnesota 

  6. Julia Lurie, Meet the mom who helped expose Flint’s toxic water nightmare. “Walters, who is trained as a medical assistant, began staying up late at night to go through reams of Flint water quality reports. She learned that Flint River water is more corrosive than Detroit tap water, and she wondered why Flint hadn’t applied standard chemicals—known as corrosion controls—to prevent the leaching of metal from its aging pipes into the water supply. This treatment is critical in a city such as Flint, where half of households are connected to a lead water line. She also didn’t understand why the city employee who tested her water ran the tap for several minutes before taking a sample. If something were building up in her pipes, wouldn’t flushing it out understate the results?” 

  7. Pew, Searching for news: The Flint water crisis. “Even before the water crisis had started to unfold, residents of Flint had begun searching for information about their water at increased rates – largely preceding both government notice and heightened regional news coverage. In mid-July 2014, searches in the News and Media category related to water issues began to increase both within Flint and across the state of Michigan.” 

  8. Nickerson 1998, Confirmation bias is ubiquitous 

  9. Alexis De Tocqueville, Democracy in America. (via Bartleby) 

  10. Jason Koebler, American farmers are hacking their tractors with Ukrainian firmware 

  11. Shafer & Doherty, Media bubble is worse than you think 

  12. Massachusetts Historical Society, Letter from John Adams to Abigail Adams, post 12 May 1780 

  13. Wikipedia Talk:Latin_proverbs 

  14. Maj V Tubbs, Advanced Situational Awareness