Good$evening.$Thanks$for$having$me$here.$Today$I$want$to$give$you$a$bit$of$my$perspec;ve$on$user$experience$design$for$predic;ve$analy;cs$for$the$internet$of$things.$Ill$unpack$some$of$those$terms$later.$The$talk$is$focused$on$the$consumer$internet$of$things$and$it$will$be$in$two$main$parts:$a$general$overview$of$the$deep$change$Internet$of$Things$has$on$peoples$rela;onships$to$devices$and$companies$rela;onships$to$value,$and$some$thoughts$on$the$user$experience$design$of$predic;ve$analy;cs,$the$core$technology$driving$the$business$models$of$a$number$of$consumer$IoT$companies.$

0$

Let me begin by telling you a bit about my background. Im a user experience designer. I was one of the first professional Web designers. This year I am celebrating professionally designing Web sites for 20 years. This is the navigation for a hot sauce shopping site I designed in the spring of 1994.

1$

Ive$also$worked$on$the$user$experience$design$of$a$lot$of$consumer$electronics$products$from$companies$youve$probably$heard$of.$

2$

Ive$wriDen$a$couple$of$books$based$on$my$experience$as$a$designer.$One$is$a$cookbook$of$user$research$methods,$and$the$second$describes$what$I$think$are$some$of$the$core$concerns$when$designing$networked$computa;onal$devices.$

3$

I$also$started$a$couple$of$companies.$The$rst,$Adap;ve$Path,$was$primarily$focused$on$the$web,$and$with$the$second$one,$ThingM,$I$got$deep$into$developing$hardware.$

4$

Today$I$work$for$PARC,$the$famous$research$lab,$as$a$principal$in$its$Innova;on$Services$group.$We$help$companies$reduce$the$risk$of$adop;ng$novel$technologies$using$a$mix$of$ethnographic$research,$user$experience$design$and$innova;on$strategy.$We$do$everything$from$coaching$teams$inside$companies$to$developing$novel$products.$

5$

The$implica;ons$of$this$technological$change$are$broad,$but$today$I$want$to$focus$on$two$aspects:$how$consumers$rela;onships$to$devices$change,$and$how$companies$value$chains$change.$

6$

The$rst$thing$I$want$to$talk$about$is$how$peoples$rela;onship$to$their$devices$is$changing.$

7$

The core of this change is a shift from generic devices and software to specialized devices and software. When computing was expensive, you had general purpose devices and general purpose software that had deal with almost every situation. This necessitated design compromises that resulted in devices and software that could do almost everything, but did none of it well.!!Now that processing is so cheap, you can have a combination of 10, 20, or 30 computing devices and apps for the price of that one device, and you can acquire new functionality as needed. This means that every device and software package can have a narrower purpose.!!Adobe brush!Haiku deck!

8$

The$second$trend$is$that$much$of$the$value$of$compu;ng$is$now$remote.$Today,$most$people$understand$that$the$experience$you$see$on$one$device$is$oTen$a$part$of$something$thats$distributed$throughout$the$world.$Theres$no$longer$a$need$to$pack$everything$into$a$single$piece$of$soTware,$and$theres$no$expecta;on$that$everything$will$be$there.$$Foursquare$as$a$whole$doesnt$live$on$your$phone$in$any$meaningful$way,$and$people$know$that.$

9$

If we chart these two tends, two broad classes of digital products emerge.!$If$we$follow$the$local$to$remote$axis,$we$nd$generalXpurpose$devices$that$do$roughly$the$same$set$of$things,$and$dier$primarily$in$size.$They exist to provide access to online services, in a form factor thats appropriate to the context in which theyre used.!I$call$these$devices$terminals.!$If$we$follow$the$general$to$specic$axis,$we$see$a$shiT$is$to$more$narrowXfunc;on$devices$that$are$designed$to$do$a$small$set$of$things with specialized hardware. A parking meter can take quarters, which your phone cant do, and a digital SLR has a giant lens on it, which you probably dont want on your phone. These devices differ in their specialized hardware. I$call$these$devices$appliances.!

10$

The most interesting thing, however, to me is that these shifts are part of an even larger transition, one where devices are simultaneously specific AND deeply tied to online services. In this model, the service provides the majority of the value, and can be represented either as a dedicated appliance, an app running on a terminal, or anything in between. !!I call these devices service avatars. And this is where much of the Internet of Things currently lies, at least for consumer products.!

11$

As value shifts to services, the devices, software applications and websites used to access itits avatarsbecome secondary. A camera becomes a really good appliance for taking photos for Flickr, while a TV becomes a nice Flickr display that you dont have to log into every time, and a phone becomes a convenient way to take your Flickr pictures on the road.!!Hardware becomes simultaneously more specialized and devalued as users see through each device to the service it represents. The hardware exists to get better value out of the service.!

12$

Amazon$really$gets$this.$Heres$a$telling$older$ad$from$Amazon$for$the$Kindle.$Its$saying$Look,$use$whatever$device$you$want.$We$dont$care,$as$long$you$stay$loyal$to$our$service.$You$can$buy$our$specialized$devices,$but$you$dont$have$to.$

13!

When$Fire$was$released$3$years$ago,$Je$Bezos$even$called$it$a$service.$

14$

Smart$Things$clearly$states$its$service$oering$right$up$front$on$their$site.$The$rst$thing$they$say$about$their$product$line$is$not$what$the$func;onality$is,$but$what$eect$their$service$will$achieve$for$their$customers.$Their$hardware$products$func;onality,$how$they$will$technically$sa;sfy$the$service$promise,$is$almost$an$aTerthought.$

15$

Compare$that$to$X10,$their$spiritual$predecessor$thats$been$in$the$business$for$more$than$20$years.$All$that$X10$tells$is$you$is$what$the$devices$are,$not$what$the$service$will$accomplish$for$you.$I$dont$even$know$if$there$IS$a$service.$Why$should$I$care$that$they$have$modules?$I$shouldnt,$and$I$dont.$

16$

The$environment$automa;cally$predicts$our$behavior$and$adjusts$to$it,$an;cipa;ng$problems$and$intercep;ng$them$before$they$occur.$The$no;on$is$seduc;ve$and$almost$magical:$an$automa;c$espresso$machine$that$starts$a$fresh$laDe$as$youre$thinking$its$a$good$;me$for$coee;$oce$lights$that$dim$when$its$sunny$and$electricity$is$expensive;$a$taco$truck$that$arrives$just$as$the$crowd$in$the$park$is$gehng$peckish.$

17$

This$is$a$richer$and$more$nuanced$promise$than$just$connec;ng$stu$to$the$Internet$and$making$an$app$for$it.$Merely$adding$sensing$and$internet$connec;vity$to$gives$you$a$gadget,$which$is$the$kind$of$thing$that$may$be$briey$interes;ng,$but$it$will$soon$end$up$liDering$your$closet.$Gadgets$are$not$things$you$use$every$day$and$that$you$love$and$associate$with$your$own$happiness.$

18$

The$real$value$proposi;on$is$a$service$that$integrates$deeply$into$your$life.$If$you$look$at$the$value$proposi;on$of$many$of$the$most$compelling$products,$you$see$that$predic;ve$analy;cs$is$at$the$core.$The$Nest$says$that$it$knows$you.$How$does$it$know$you?$It$predicts$what$youre$going$to$want$based$on$your$past$behavior.$

19$

Amazons$Echo$speaker,$which$was$just$announced$says$its$con;nually$learning.$How$is$that?$Predic;ve$analy;cs.$

20$

The$Birdi$smart$smoke$alarm$says$it$will$learn$over$;me,$which$is$again$the$same$thing.$

21$

Jaguar$comes$right$out$with$it.$They$even$obliquely$reference$the$40$years$of$ar;cial$intelligence$research$that$powers$predic;ve$analy;cs$by$calling$their$car$not$just$learning,$but$learning$AND$intelligent.$

22$

The$Edyn$plant$watering$system$adapts$to$every$change.$What$is$that$adapta;on?$Predic;ve$analy;cs.$

23$

So$whats$Predic;ve$Analy;cs?$Its$a$component$of$what$used$to$be$called$ar;cial$intelligence,$and$is$now$associated$with$the$cluster$of$stu$under$the$Big$Data$umbrella.$From$my$perspec;ve$its$the$combina;on$of$two$things:$machine$learning$of$sensor$data$and$sta;s;cal$predic;ons$of$possible$future$states$based$on$a$large$number$of$examples$of$past$states.$Machine$learning$creates$models$of$sensor$datawhich$includes$peopleand$thus,$using$past$behavior,$can$create$probabili;es$for$future$behavior.$Given$enough$data,$these$sta;s;cal$predic;ons$can$be$very$nuanced$and$highly$accurate,$becausebroadly$speakingpeople$and$their$machines$are$preDy$predictable.$$

24$

For$example,$the$Nest$thermostat$starts$by$recording$readings$from$its$onboard$sensors$and$its$users$behavior,$to$which$it$probably$adds$the$weather,$the$;me,$the$day$of$the$week,$and$secret$sauce$data$(the$spot$price$of$energy?$How$many$people$are$in$the$house$right$now?$How$neighbors$have$set$their$Nests?).$It$(and$by$it$I$mean$the$Nest$service$in$the$cloud)$builds$a$model$of$how$people$prefer$hea;ng$and$air$condi;oning$to$work,$subtly$adjusted$for$every$household.$Then$the$Nest$acts$on$its$users$behalf.$It$creates$a$schedule$that$changes$the$temperature,$but$it$doesnt$just$play$back$past$behavior$or$preset$preferences.$Instead,$it$acts$on$what$it$believes$users$would$have$wanted$if$they$knew$more$than$just$Im$hot$or$Im$cold.$$$This$is$a$very$new$kind$of$product,$since$our$things$didnt$make$models$of$us$before.$And$the$core$principle$is$not$newXXmachine$learning$has$existed$for$decadesXXits$that$it$has$tradi;onally$been$used$in$places$such$as$nancial$services$where$only$experts$ever$interacted$with$the$systems$directly.$Now$products$based$on$machine$learning$are$quickly$becoming$parts$of$our$everyday$lives.$Which$means$that$their$design$is$now$a$user$experience$problem.$

25$

Let$me$step$back$for$a$second$and$dene$user$experience$design.$

26$

First,$Id$like$to$dene$user$experience$design.$UX$design$is$not$interface$design,$its$not$industrial$design,$its$not$product$design,$but$it$includes$all$of$those$things.$UX$design$is$a$userXcentered$problem$solving$approach$that$brings$together$business$needs$and$user$needs$to$create$products$that$are$valuable$for$both$groups.$The$eld$is$about$20$years$old$and$heres$a$diagram$by$Jess$McMullin$that$lays$out$the$basic$idea$of$the$goal$of$the$prac;ce.$

27$

Heres$a$somewhat$more$elaborate$illustra;on$done$this$year$by$Corey$Stern.$I$wont$go$into$the$details,$but$you$can$see$that$the$pillars$are$what$users$want,$what$businesses$need$and$how$the$two$interact.$

28$

Now$Id$like$to$oer$some$ini;al$thoughts$on$user$experience$design$for$predic;ve$analy;cs$for$the$internet$of$things.$Were$s;ll$very$early$in$the$evolu;on$of$this$kind$of$interac;on,$so$these$are$very$much$a$work$in$progress.$I$want$to$outline$three$problems$with$predic;ve$analy;cs$UX,$and$then$oer$four$paDerns$that$I$believe$address$some$of$the$issues$raised$by$the$problems.$

29$

The$rst$problem$is$that$most$predic;ve$models$are$preDy$bad$when$theyre$new.$Because$theyre$sta;s;cal,$they$need$lots$of$data$to$be$accurate,$and$ini;ally$they$dont$have$that$data,$so$they$wont$be$accurate.$60%$accuracy$is$preDy$typical$for$a$rst$pass,$but$even$90%$accuracy$isnt$enough$for$a$predic;ve$system$to$feel$right,$since$its$going$to$be$making$mistakes$all$the$;me.$And$sooner$than$we$think$were$going$to$have$100$connected$devices$simultaneously$ac;ng$on$predic;ons$about$us.$If$each$is$99%$accurate,$then$one$is$always$wrong.$So$the$problem$is:$How$can$you$design$a$user$experience$to$make$a$device$s;ll$func;onal,$s;ll$valuable,$s;ll$fun,$even$when$its$spewing$junk$behavior?$

30$

The$next$big$class$of$issues$in$predic;ve$IoT$experience$design$is$around$expecta;ons.$As$more$devices$adapt$their$behavior,$how$will$they$communicate$that$theyre$doing$so?$Do$we$s;ck$a$sign$on$them$that$says$adap;ng,$like$the$light$on$a$video$camera$says$recording?$Should$my$chair$vibrate$when$adjus;ng$to$my$posture?$How$will$users,$or$just$passersXby,$know$which$things$adapt$and$which$merely$behave?$I$could$end$up$sihng$uncomfortable$for$a$long$;me,$wai;ng$for$my$chair$to$change,$before$realizing$it$doesnt$adapt$on$its$own.$How$should$smart$devices$set$the$expecta;on$that$they$may$behave$dierently$in$what$appears$to$people$as$an$iden;cal$set$of$circumstances?$$$People$are$already$prone$to$project$more$intelligence$on$devices$than$those$devices$actually$have,$so$a$couple$accurate$predic;ons$may$imply$a$much$more$accurate$model$than$actually$exists.$Robo;cists$talk$about$an$uncanny$valley$where$robots$exhibit$just$enough$humanXlike$behavior$to$ini;ally$fool$us,$but$become$creepy$when$theyre$revealed$to$be$anything$but$human.$$$Chair$by$Raaello$D'Andrea,$MaD$Donovan$and$Max$Dean.$

31$

The$third$issue$is$control.$How$can$we$maintain$some$level$of$control$over$these$devices,$when$their$behavior$is$by$deni;on$sta;s;cal.$When$I$visited$Nest$once$they$told$me$that$none$of$the$Nests$in$their$oce$actually$work$very$well$because$theyre$constantly$ddling$with$them.$In$machine$learning$this$is$called$overtraining.$If$I$cant$actually$train$my$device,$and$I$have$no$direct$way$to$control$it$other$than$through$my$own$behavior,$how$do$I$adjust$it?$Amazon$and$Neslixs$recommenda;on$systems,$which$is$a$kind$of$predic;ve$analy;cs$system,$give$you$some$context$about$why$they$recommended$something,$but$what$do$I$do$when$my$only$interface$is$a$garden$hose?$

32$

Here$are$4$paDerns$Ive$observed$in$developing$predic;ve$systems$that$I$think$map$to$the$IoT.$For$most$of$these$Im$going$to$be$using$nonXIoT$examples,$since$theyre$preDy$rate,$and$most$of$my$paDerns$come$from$recommender$systems$like$Amazons,$Googles$and$Neslixs.$These$systems$are$based$on$similar$predic;ve$technologies$and$theyve$been$around$for$much$longer.$

33$

Set$expecta;ons$by$making$people$aware$of$the$predic;ve$nature$of$the$device$by$describing$that$it$is$trying$to$predict$the$users$behavior,$why$it$made$some$decision,$and$perhaps$an$idea$of$how$the$model$is$constructed:$which$of$their$behaviors$are$being$used,$how$long$it$take?$What$happens$when$its$wrong,$etc.$$Recommender$systems,$such$as$Google$Now,$describe$why$a$certain$kind$of$content$was$selected,$and$that$sets$the$expecta;on$that$in$the$future$the$system$will$recommend$other$things$based$on$other$kinds$of$content$youve$requested.$Nests$FAQ$kind$of$buries$the$informa;on,$but$it$does$explain$that$you$shouldnt$expect$your$thermostat$to$make$a$model$of$when$youre$home$or$not$un;l$its$been$opera;ng$for$a$$week$or$so.$

34$

Knowing$that$a$device$has$acted$on$your$behalf,$and$that$its$going$to$actand$HOW$its$going$to$actin$the$future$is$important$to$giving$people$a$model$of$how$its$working.$Nest,$for$example,$has$a$calendar$of$its$expected$behavior,$and$it$shows$that$its$ac;ng$on$your$behalf$to$change$the$temperature,$and$when$you$can$expect$that$temperature$will$be$reached.$

35$

Remember$the$Wall$Street$Journal$story$about$a$guy$(Basil$Iwanyk)$realized$his$TiVo$thought$he$was$gay$because$of$the$shows$he$recorded$(he$wasnt)?$Thats$funny,$but$the$second$part$is$frustra;ng:$he$tried$to$correct$it$the$only$way$he$could$think$of,$by$watching$war$movies,$and$his$TiVo$then$decided$he$was$a$Nazi$(which$he$also$wasnt).$You$have$to$give$people$a$clear$way$to$teach$the$system$and$tell$it$when$its$model$is$wrong.$Google$Now$asks$whether$I$want$more$informa;on$from$a$site$I$visited,$Amazon$shows$a$explana;on$of$why$it$gave$me$a$sugges;on.$

36$

Finally,$dont$automate.$These$systems$shouldnt$try$to$replace$people,$but$to$support$them,$to$augment$and$extent$their$capabili;es,$not$to$replace$them.$$One$of$my$favorite$current$examples$of$this$kind$of$system$is$from$Meshre,$which$is$a$social$media$management$tool$that$has$a$machine$learning$assistant.$Its$machine$learning$assistant,$called$Ember,$doesnt$try$to$replace$the$social$media$manager.$Instead$it$manages$the$media$managers$todo$list.$It$adds$things$that$it$thinks$are$going$to$be$interes;ng,$deletes$old$things,$and$repriori;zes$the$managers$list$based$on$what$it$thinks$is$important.$I$think$this$is$a$good$model$for$how$such$systems$can$add$value$to$a$persons$experience$without$crea;ng$a$situa;on$where$random,$unexplained$behaviors$confuse$people,$frustrate$them$and$make$them$feel$powerless.$Ember$is$an$augmenta;on$to$the$social$media$manager,$it$helps$that$person$focus$on$whats$important$so$that$they$can$be$smarter$about$their$decisions.$It$doesnt$try$to$be$smarter$than$they$are.$

37$

Finally,$for$me$the$IoT$is$not$about$the$things,$but$the$experience$created$by$the$services$for$which$the$things$are$avatars.$

38$

Ul;mately$we$are$using$these$tools$to$extend$our$capabili;es,$to$use$the$digital$world$as$an$extension$of$our$minds.$To$do$that$well$we$have$to$respect$that$as$interes;ng$and$powerful$as$these$technologies$are,$they$are$s;ll$in$their$infancy,$and$our$job$as$entrepreneurs,$developers$and$designers$will$be$to$create$systems,$services,$that$help$people,$rather$than$adding$extra$work$in$the$name$of$simplis;c$automa;on.$What$we$want$to$create$is$a$symbio;c$rela;onship$where$we,$and$our$predic;ve$systems,$work$together$to$create$a$world$that$provides$the$most$value,$for$the$least$cost,$for$the$most$people,$for$the$longest$;me.$

39$

Thank$you.$

40$