Designing with Data

Here’s my notes from two back-to-back talks on how the UK gov­ern­ment uses de­sign and data sci­ence to make bet­ter pol­icy by Cat Drew and Laura Malan.

Data and Design in Policymaking

Cat Drew, Policy Lab

pearsall map Phyllis Pearsall made the first pocket-sized street map of London in 1935.

nightingale preventable deaths Florence Nightingale helped to re­duce mor­tal­ity in the Crimean War by tak­ing ex­ist­ing data (causes of death) and pre­sent­ing it in a new way. Eventually led to changes in nurs­ing pol­icy.

  1. Data can be used to cre­ate prac­ti­cal ap­pli­ca­tions to help cit­i­zens
  2. Data can be vi­su­al­ized in a dif­fer­ent way to revveal new in­sights which lead to pol­icy change

Policy lab is a 10 per­son team that sup­ports de­part­ments across gov­ern­ment to use de­sign, data and dig­i­tal tools to make pol­icy. They look for pro­jects at the in­ter­sec­tion of de­sign, data sci­ence and dig­i­tal me­dia.

Traditional data in pol­i­cy­mak­ing

Government gen­er­ates loads of data, but it’s usu­ally pre­sented in 200 page PDFs with a mil­lion bar charts, weird maps, bar charts, line charts all that crap

Traditional data, pre­sented bet­ter

IPPR: Brits Abroad (2006)

Used early form of data sci­ence to com­pile data in a mas­sive table Used a dis­torted map to show where British peo­ple were liv­ing in the world (which was new in 2006)

Guardian public spending graphic The Guardian: Public Spending Diagram (2011) Source

Billion Dollar-o-Ggram David McCandless, Information is Beautiful: The Billion Dollar-o-Gram (2013) Source uses data, puts data points next to ecach other t make a po­lit­i­cal point

London Fire Brigade Callout data London Fire Brigade, GDS: Fire Callout Data (2014) GDS Article on this pro­ject

ONS halloween data Office of National Statistics: Halloween Data (2016) think of ways to en­gage the pub­lic in data by do­ing top­i­cal de­sign pieces

an­other way to get the pub­lic en­gaged in data

New types of data used by gov­ern­ment

Tweets correlated to novovirus diagram The Food Standards Agency mon­i­tored Twitter for Tweets about ill­ness and used that to pre­dict Novovirus out­breaks

IBM used mo­bile phone data to re­design bus routes in Ivory Coast Used phone records to find out were peo­ple are ac­tu­ally walk­ing in­stead of re­ly­ing on as­sump­tions by the city.

Policy Lab pro­ject on re­duc­ing home­less­ness Data can also be qual­i­ta­tive. Policy lab does ethnog­ra­phy were you spend time with peo­ple and ob­serve them. THis can give you in­sights that aren’t eas­ily cap­tured by sta­tis­tics.

Sometimes just mak­ing data avail­able to peo­ple is ben­e­fi­cial

Data can be used to hold the gov­ern­ment ac­count­able

po­ open data Worked for months to get the the po­lice to pub­lish all of their ar­rest data. This gives peo­ple the ar­gu­ments they need to make change hap­pen

London air pol­lu­tion map The cur­rent de­bate about air pol­lu­tion in London could only re­ally hap­pen be­cause high-qual­ity data be­came avail­able for every­one to look at.

Anna Powell-Smith: What size am I? Interactive graphic that shows what dress sizes are of­fered by dif­fer­ent shops, can be used to put pres­sure on those com­pa­nies.

Cat Drew, Dalston Gentrification Final LCC pro­ject; plot­ted every shop on Kingston road and what year they opened up. Data has to be pre­sented in dif­fer­ent ways to dif­fer­ent com­mu­ni­ties.

Ways de­sign­ers are in­volved in data sci­ence pro­jects:

1. Data col­lec­tion

How do we get data from peo­ple in a way that cre­ates trust and helps peo­ple un­der­stand wht we’re try­ing to do? For a traf­fic pro­ject we had to put up huge posters on the m4 to tell peo­ple they were us­ing their phone data. Thnik about what sort of ques­tion you can ask peo­ple and how you ask it.

Big bang data balloons Signal Noise: Big Bang Data Asked peo­ple to an­swer ques­tions on a card, then at­tach that to a bal­loon with a length of string ac­cord­ing to their age.

2. Data Magic

Not much (I think de­sign­ers should know about this)

3. Using the Data

GDS: Future of age­ing Collected loads of aca­d­e­mic ev­i­dence about the ag­ing pop­u­la­tion; 200.000 words in to­tal that no pub­lic ser­vant could ever read.

Policy Lab broke the in­for­ma­tion down into cards that made much more di­gestible chunks Worked with an ill­stra­tor to cre­ate per­sonas, mak­ing the data more hu­man al­lowed civil ser­vants to think more clearly about pol­icy changes that needed to hap­pen.

Citymapper built a fea­ture that lets you choose a rout were you’re in the rain for the least amount of time. Again, think­ing about what sort of peo­ple need at what point in their day.

Christopher Murray: Global Burden of Disease Comparison Led to an ef­fort by the NHS to pre­vent dis­eases they weren’t aware of be­ing a prob­lem

Sometimes you need to do things that are counter-in­tu­itive to get peo­ple to trust your data: This phone book com­pany in the 70s launched a ser­vice were you could call an a com­puter would give you any phone num­ber you needed. However peo­ple did­n’t trust the ma­chine be­cause the re­sult came pretty much in­stantly, so the phone com­pany had to play a recordign of some­one flip­ping through a phone book for a cou­ple of sec­onds so peo­ple would be­lieve the re­sult. Today Skyscanner is do­ing the same thing.

Designing with data

Laura Malan, uscre­ates


Here’s a pro­ject Laura and Cat have been work­ing on for two years

Designing data is dif­fer­ent to de­sign­ing with data, and both can be done at the same time. It’s good to use quan­ti­ta­tive as well as qual­i­ta­tive data. One gives you facts and num­bers, the other helps you un­der­stand rea­sons be­hind those num­bers.

How can we sup­port peo­ple to man­age their health con­di­tions and stay in work?

1. Use data to see pat­terns

Made a graphic that shows the flow of peo­ple from em­poy­ment into dif­fer­ent stages of ben­e­fits and back, looks like the user flow di­a­gram on Google Analytics

2. Data =/= Insights

had to use de­sign tech­niques to un­der­stand com­plex data sets iden­ti­fied clus­ters of peo­ple in ben­e­fits, pulled those out and vi­su­al­ized them

  1. hu­man­ise data data sci­ence is good but hu­man info has a big im­pact went out and talked to peo­ple to get their sto­ries, in depth in­ter­views, give peo­ple cam­eras to cap­ture di­aries

rich ethnog­ra­phy data and sto­ries and names and pho­tos helped them com­mu­ni­cate data to pol­icy mak­ers.

  1. data vi­su­al­iza­tion as a tool can be a pow­er­ful tool to com­mu­ni­cate ev­i­dence cre­ated user jour­neys for peo­ple who flowed in a nd out of work shows that peo­ple have wuite com­plex jour­neys, come in con­tact with many agen­cies

  2. use data to solve prob­lems use data to solvve de­sign prob­lems work out prob­lems you hav to solve


Can you give an ex­am­ple where you could show the ben­e­fits of a de­sign pro­ject us­ing quan­ti­ta­tive data?

po­lice data, us­ing data to im­prove crime vic­tims ex­pe­ri­ence on­line crime record­ing tool got qual­i­ta­tive feed­back

How is data-dri­ven de­sign go­ing to change in the fu­ture?

gonna get a lot more data from the nter­net of things doc­tors will be able to re­motely di­ag­nose ro­bots dri­ving cars data ethics will be­come more im­por­tant

How do you avoid bi­ases in your de­sign process

nyc street bump app bi­ased to­wards rich peo­ple his­toric data might have bi­ases (racial bi­ases, gen­der bi­ases) you risk just con­form­ing, per­pet­u­at­ing bi­ases

live data and data around that are there peo­ple usign live data per­for­mance plat­form, page an­a­lyt­ics gives us a sense of what peo­ple are search­ing for from gov­ern­ment perdic­i­tive al­go­rithms al­low us to pre­dict what peo­ple might need at dif­fer­ent times in the year