MORITZ STEFANER - OpenVis Conf 2018

so good morning everybody hey so the title of my talk today is called maca holes and catacomb or catacombs

for reasons hopefully they've become later apparent thanks Lin for the super kind introduction and I'm super excited to be here and to open this conference because afterwards I can lay back and

listen to all the talks I want to hear so it's amazing yeah my name is Masha fauna I do data visualization occasionally under the name of truth and

beauty operator that's my job title I think I'm still the only one maybe there will be like a ripoff version at some point but I think I'm still gonna conduct regard I also run a podcast

together with Enrico Bertini called data stories where we discuss the role data plays in our lives we have often conversations interviews may be that

some listeners out there yeah so um yeah great to be in Paris a quick show of hands who's an original Parisian or

lives here at least yes yes a few folks but as we have seen a lot of international guests as well right so let me give you a quick tour of the city

of course Paris is the most beautiful city of the world I think everybody knows that I could take pictures of lesson all day long basically you know just sit there see see the boats go by

or maybe also for weeks you know as enough photos to take there although if you know me you know I'm here mostly and for the food so that's that's a very

good motivation to come here as well I hope they also have a painting here so

maybe two of them so art is of obviously super amazing as well you won't see any guards in these pictures here right so because that actually can take a

photograph as you can see but you don't have to go to museums there's lots of really amazing street art you can find I mean the city in itself is an artwork

right so all the little shops the store is the streets buildings yeah like some famous Tower you need to bring a red umbrella that's rule number one

yeah like think of that yeah yeah if you go to the Louvre do me a favor don't do this right just a hint it's been done

yeah I checked the stats and I think it's been done of course and then there's the fashion you know say the city of style city of fashion hairstyles

are big as well you know so of the quarter yeah it's quite amazing lately cool - two stars as well right and of

course it's the city of love so if you come here you need to take a selfie at the wall of love where all the couples take a shot and so yeah and so I find

all these montages of these images super fascinating to look at in a way is that thing like do we all take the same pictures you know it's like how original are we really do we need to take another

crappy shot on our cell phone at the same time they are all unique right and there's something much more happening than documenting like a photographic like trying to do a good shot of the

Eiffel Tower you say like I was there and I enjoyed it it's stuff like this so and I think that's really a unique quality in this redundancy and multiplicity and this is why I made this project also around these images of

Paris it's called multiplicity and it was commissioned to you by phone a CEO at EF and it's on show as well here so there's a database exhibition called

one two three data or data I guess at the fondazione and you should take a look so and I'll be the unpin see as well so we can take a look together but

the project again is about can we actually draw a portrait of a city as seen through the lenses of thousands of photographers I think that's just a fascinating question to think about and

so what I was very inspired by were Johanna she knows a Japanese artists diorama maps and similar to David Hockney in the 70s who made all these

photo collages like almost cubistic like reconstructions of a place he creates these honors city scales of basically goes through the city and takes photos everywhere he finds

interesting and later assembles these photos into this big rich ecology city map right I was super fascinated with that and I felt like maybe we can do that now with social media and like have

other people take these photos and we just curate and assemble so I collected a lot of images taken in Paris from social media sites so here you basically see actually metadata for millions of

images on like tens of thousands maybe hundreds of thousands of locations right and in the beginning I actually made these maps to cousins always so nice yeah you know it's like you're flying

over the city everything's full of stars particle that's amazing and then I was thinking oh maybe I can bring this together with the photography can I bring in some of the qualities from these images not just the dots and the

data right but really what they are about so I was trying to mix like I don't know if you can see that well but here mixing like here's the Java it's the park so I tried to bring in the imagery from the park into the map

background or thinking about can we mix it somehow with the satellites then I was thinking how surely we can draw the city map out of photos right that must be it so he would see la saying there

and the mountain like the Eiffel Tower and so on but it ends up as a big mess and it's really super visually super hard to parse and I was very like at this point pretty stuck because I knew

this ones I don't know this one we've seen a few times before right this one is doesn't quite go together and this was just a big mess so I was like okay well what do we do and so then

I started up to think about okay what do we do actually when when we visualize how can we create sense out of such a complex thing and at the heart of it is in many ways I think the art of tidying

up right like how can we take something as complex and a big mess and sort of as a first step just look at what can be identified that seems to belong together seems to be similar it could form a

group and how do these groups relate to each other so I think that's something that's much more fundamental then counting is this idea of tidying up and also severely has this beautiful book the art of cleanup

his risk I so he's really into cleaning up and and he so he cleans up all these messy places like and we sleep all right or like a fruit salad

anyways this idea to spatially first arrange something and then think what's what and what do we need to count I think that actually much more important or importance that we often jump over and jump to measuring and Counting right

away so luckily today we have computers to help us clean up and and so in in my

case I used an algorithm called tease me to make a map that put similar images close together right so you take all these messy images and you put all the cars in one place all the parts of the

the food maybe and the big question is of course what is similar this is determined by a neural network so it took a pre trained your network coming from Google

it's called vgg sixteen and take basically the last layer of the network that encodes what is important for classification so it doesn't say this is a dog and this is a cat but the features

like the descriptions that go into that decision these I take to determine what is a similar image to another one and then I end up with this map as in like the big overview of twenty five thousand

images you can see already roughly our vigil s visually similar images seem to end up in the right spot right but you can't really see already like what is that category of images and for that you

need to zoom in a bit more I hope you can see it a bit so here you have that painting we've seen the Mona Lisa he is a clusterf of the round things here's all the doors and the beautiful thing is

you don't just get these hard categories but you also get these transitions in between right so it's a continuous space it's not a hard classification but you have this continuous map of that image

space and I've been doing these types of things for many years but now with these newer networks there's really a new quality to how these turn out so for instance I was quite surprised I had one

corner here with bikes in the map and I had one here with cars and what do you think that the algorithm put in between it was the bicycle and it was the

motorbikes right and this totally makes sense but it's nothing that the machine was taught to do it's just visually they end up between Smith's like a small car or a big bike you right

motorized bike so visually they are so like capable of capturing stuff that we as humans are also interested in that all these really meaningfully relations early merge which I find really quite

fascinating anyway so I turned this big map into an installation at an Exhibition and it's a big zoo mobile interface so we have three different projections but they all form one

projection space in the touchpad here and then you can on the touchpad navigate this big map and you have a little joystick actually for a little arcade action and I was reported like

kids use this now to be the space navigator for the other visitors and then you can zoom into this image space and sort of explore all the regions of

interest that you're interested in right so here's a bit how it works so down here is the control and I don't know how well you can see it can be turned on the

lights a bit yeah so you can tap a spot here in this control interface and then zoom into that image cluster or you tap another spot and then it fully zooms out

and zooms to the spot and you you're like why why did you use the cool check van Wyk t3 transition and I will tell you why because if you do that on a big

screen you end up nauseating your whole audience everybody will basically throw up if you start to do this big panning motions but what's files like this you know zoom in zoom out it's like if you

ride on a bus you want to look ahead not to the other side windows so and then you can zoom into here's the food

cluster and here's the other clusters and so on or you can also use the the

joystick and the zoom buttons to pan we'll see that in a minute yeah so it's a really interesting way I think to explore a city just by looking

at okay what are all the images people take and one thing I was working or struggling with really quite a bit is this idea of all of you in DFS so also this makes it obviously a data

visualization the idea that su means well and just in case you were wondering so but what I was struggling with so I knew I wanted to have this cloudy

overview because it looks so nice and the grid as an overview is very heavy and like very busy and I also knew I wanted to have this grid when you zoom in because otherwise it becomes quickly

a real big mess to Paris visually so this grid is really nice to see all these photos and be able to study them the promise what's in between like how do you how do you get from here to there right and again you might as a web

person think about yeah we animate all these little particles into place and we can easily move a few thousand things around but again it creates a big visual mess in that in-between step and you don't really want to have that on that scale

so I was thinking ah maybe I'll do a more gritty cloud right so I can stretch this original cloud into a squarish shape right and then maybe get more and

more gritty as we zoom in and fill up the rest right but this one looked really strange it look bit like the skin of a cow or something that's you yeah so

then it was like okay maybe I'll start with the grid make it more cloudy the obvious next thought so I would align everything on a cloud like on the grid

but scale the images that are close to a cluster center like that a part of a group of very similar which is the one in the center I could scale up right so that a cloudy structure would emerge smart but again it's it's not great

because if you zoom in here you still have sort of an odd situation and you still have to think about how do I now scale up these images gradually or I don't know it didn't really work then I

was thinking maybe I'll make a disk so a roundish cloud maybe it's easier to you know looks like the moon as well so that might be a plus but then it has nothing

to do with the topic but yeah but then I also had this awkward here the the round corner didn't make sense when you zoom in like why would you then have I was a bit took me actually a few weeks

to figure out like I was at this point it was like there must be a solution right so how I solved it in the end this I do something like a stepwise ratification of that clouds from the

very cloudy to the bit more pixelated cloudy to the full grid yeah and the way it works is like this so here's the cloud here I do the same trick by the way I scaled the images that at the

center of a cluster and make them a bit bigger so it becomes more fluffy and you can easily see the groups of images that are similar then I take these and align them on a grid and you can also see here

do some photoshop tricks to bring out the colors a bit more and make more coherent colors here are the real images it gets a bit more busy immediately so this is a good overview this is the next closest grid I could find to that

overview then I fill in all the rest of that grid so here you can see they're still gaps I felt that with the images that were still left over and then I

still at the really ugly corners here damn it and then I just take other random pictures from a bigger pool of images to fill up the rest of the space

just because I didn't want to have these awkward you know but so so it might happen that actually an image that here here's more food coming in although food

should be over here I mean there's always some compromise you have to make this the best I could have come up with so yeah anyways and now if you look into

the look at the clouds ooming you can see first is the cloudy now it becomes more gritty but it's still very similar very similar and now we bring in the other images and nothing not many big changes because it's basically the same

images but we just fill in the gaps right and what seems yeah quite natural now it's actually kind of a complicated procedure to calculate all these

different base maps yeah Ami's technical details but just so you know how I spend my time anyways I like to frame this project as

a portrait yeah like a image portrait of the city because what I find interesting really about it is you don't really see much charts you don't any data in the project but it did use a

lot of data and calculations in the background to create this map right but what what I end up with is just the spatial arrangement this tidied up version of that mess of that image space

so but these measurements I take I've used them to create a really a subjective portrait of the city and I think it's kind of interesting to look at parallels to painting or traditional

portraying right so for instance in painting I think we all agree that this is a very subjective representation of reality right so somebody takes a look

at reality and craft an image of reality that is their personal expression of that reality right and we tend to think in photography of course that this is a

much more objective way of depicting reality because there's an apparatus involved that takes a snapshot of reality at a certain point in time right

it's like it's a physical process but of course as we all know there's decisions being made like when do I press the button how much do i zoom in and even if

there's not even a manual photography taken there's still always a selection process and this selection process is ultimately where all the authorship also happens so even if you had a webcam

taking like images 24/7 somebody will pull up that one image where you look really funny right and so also here with Margaret Thatcher this becomes as iconic image of the IRA lady but if you had they have taken this image of that you

know things might've played out differently and and so we have seen there's so much creativity and authorship in like how I design my layout how I do the interaction and so

on but the really such a big part is also how do we frame reality how much we zoom in zoom out and how do we pick the data we choose right and so the

authorship in Dana selection I think is not really discussed all that much but there's so much of decisions being made in yeah what is the best portrayal of a

phenomenon right and in my case I was like hmm how many photos do I take right so these are the runs like help you hurts like just a fraction what's the selection do I use a random sample do I

try to map the city area equally to take the most liked photos you know it's like what is the best way into this what's the best angle I can take on that subject but and in my case I started

with very simple analytical rules like top popular best geographic coverage and so on but then then it was always like yeah this is still super subjective in a way you know it's it's still social

media it's never an objective portray of reality so I said like okay if this is the case I can also make a custom blend of data right like like you would they do with an iced coffee or a curvy write

and so me as the sort of the the designer of that data said I took a very conscious approach to actually design it like a recipe so I take one part stoplight photos one part part uniform

spatial sample one circle one part hand selected clusters it's a give-and-take yeah and two parts from most recent divisions from most active locations and I just found that end result to be the

best representation of reality but that's just me right but I also make that very clear by stating the data selection here as a cooking recipe basically the other big

parts come back to cooking is like yeah how much do you turn up the oven right and here we get into this area of parameter selections of when you work with all these algorithms it is kind of

important what these numbers are right and this is the art of tweaking and the art of finding the right settings then it's also a thing not discussed much it's just something some people are good at and have the tenacity to go through

and then they end up with something that works well but that tends to be forgotten I think in many cases how much role that can play lastly I was like okay so we have this all view map you

don't really see all that much I think we should somehow give visitors a bit of guidance of how to navigate in that space and either that here with handwritten annotations again to

reinforce that point okay there is actually an awesome loft here yeah and that's me as a person and I make these decisions right it's like there's nothing there's no absolute reason the

food corn is here or that is exactly the size it's just something I found right so I documented in this handwritten style just to make clear again it's not a measurement it's an opinion right and so these are little cues to avoid

this perception that okay here's what machine learning can tell us about parrots because it's not true machine learning that was super interesting and I've really enjoyed this

collaboration with the machine using all these new algorithms because it can be actually quite inspiring and it can be quite enlightening also and so I got into this dialogue so I had this huge

and we said I did this analysis I looked at the maps then I thought like oh what's happening here I went back to the day they found more similar images so I went into this constant iteration dialogue with the machine and the

algorithms and I was really also sometimes really surprised by the results so I think this is as a date of this person I really like how here all the circle packing options are laid out

for me in the form of coffee tables or desserts and so on and all these layouts are really inspiring and overall like the color palettes then it's quite something and or you end up in these

strange corners of the data which is like things on the floor like what things on the floor do we have right and so it's it's not really a meaningful

photo category you would necessarily come up with when you start with sort of unique images like this lanterns she end up with this really interesting mood

board so you know this is like if somebody told me yeah this is the mood board for the new campaign for Brand X and it cost like millions I wouldn't be surprised it's a really good composition

here right it's super interesting and so yeah it's quite intriguing and how I generated these images is always find one like in this case I think it was this one and just take the most similar

ones and then put them in this grid right now anyway the biggest surprise in here we come to the title of my talk is

when I found one quarry funded a corner of the map where there's mackerels here lovely macarons and there's the catacombs and all the skulls that's that's kind of dark this escalated

quickly and the thing is it actually makes sense you can see how why you would think to put these and did in the same space

that's always this done here it's it's confusing so I've been thinking what our machines are telling us here

the new machine overlords I think the message is eat enough mackerels in any in the shortest amount of time you can

[Laughter] so anyways and so it's on this play the project I could just show you a small slice of it this it's 25,000 images so

obviously there's a lot to explore and if you still down Wednesday you can meet me there between 12:00 and 2:00 so it's right after your workshop or right before your workshop maybe

hopefully I don't know if it works out with the travel time but sure and yeah I'd be happy to show you the exhibition it's a lot of really cool data stuff and

this project anyways moving on for something completely different that's the second project I want to show you before I wrap it up and this one's

called peak spotting and it's a super applied project for the German railway the deutsche bahn in their design they developed it together with studio anand

and christian lessor so it's a bigger team that I'm speaking for here it's a really interesting one so it's a sort of a tough challenge we're trying to solve

the the deutsche bahn has this system where you don't have to book a specific place like if you take a TV or a talents or so you have to book one place it's like an airplane and then it's gone and when the train is full it's full in

Germany they always needs to be like half of the ticket say to be open in terms of you can get on any train if you have the right ticket right so it seems to be a political really important thing

so the challenge is then how do you design the pricing and how do you give cues that the trains are full but not too full it's a very tricky soft management problem because you have only

a rough idea of how many people will be there and you have only very soft measures to affect what people will do there's no way to force people out of a train unless it's to fall and then you really have to get cops which also

sometimes happens in the interesting situation here is that's basically the good news lots of AI yeah so there are pretty decent predictions in house like how full the trains will be so they have

a lot of data that could help them make them good decisions the bad news is there's no real tools for humans to work with that data so you know it's mostly

table based lots of like big specific tools but they still do look a lot of like Excel with lots of visual basic madness and that's a general like observation I don't know it's especially

in German industry or worldwide but in my experience if you see often either very simplistic ways of dealing with data in business context so you see these super simple KPI is you know

dashboards for physics executives and then you just have the KPI but it looks a bit dull so you can make a circle around it right it's like it usually

helps or and I can or both right and so and this approach I like to call number decoration because this is all that's happening if we're everyone Australians you can be quite good take it quite far but there's always so and so far you can

get it's a bit like you know people would be interested in the forest and all the yeah all the details but they just get these nice little pre-chopped pieces of wood and yeah just dressing up numbers basically and then you complain

that again need to know the underlying data this is too simplistic and then then you get the data Explorer there you have like a drop-down for every middle

management executive for heads like a special requests and you have all the views and you have access to all the data so no more complaints right yeah so

this is the here's all the pain a good luck approach and my feeling is there must be something in between maybe like do we always need to like you know go to

these extremes and in Thor Japan it was actually quite similar as I said lots of tables lots of like number based stuff and the other thing is lots of tools in

parallel right so if you this is a picture of real working space like 2030 people having like 6-7 screen setups with five six seven individual tools

from different decades of the computer age yeah and they need to make the connections in their mind between these tools and also memorize all the numbers and all the connections and all the

things that were interested in right so there's still a lot of manual working on to bridge these did they even resorted to making printing out stuff and then you know doing text

markers in like for the Christmas which is always the worst like because everybody's traveling so they just went back to printing stuff out and then discussing it on paper right so yeah our

pro two years okay we I think we can craft really a data experience that is much more tied to what people actually want to do and need to do but this is also opinionated in a sense that we say

okay we analyze your situation and we try and make a solution that really really works well for you right and still allows these open experiences active learning active exploration really helping be able to make sense of

that data and here's quickly an impression how this Tool Works so basically you have a calendar of a hundred days looking ahead and you can first find out what are the worst days like what are the fullest the biggest

bookings and you can see that these are the more red ones so these are little thumbnails over time number of bookings and then you can go into these days individually and have a lot of different

views to look at okay what are the train tracks and the the routes that cause these situations where are bottlenecks and what can we do right and so here we

have these lists that have these little elements for each train again time running from left to right you have different aggregated view so these are groups of similar trains like basically

on the same big route through Germany and again timeline running left right so you can see there's a lot here in the afternoon these are these path time diagrams if you're into trains you will

encounter these quite often and here the stations run left to right at time top to bottom and you have small multiples of course because there was a

visualization designer involved and you can filter in different ways so it's a very interactive way to make sense of that actually quite large data sets what we also put a lot of emphasis on is

actually actions so you can do something on these trains and of course we have an animated map because it's trains yes

we have to actually it is important but if because if you just draw a normal Network you don't see the directions right so you can drive from the A to B or B to a this is always very hard to

model so animation can actually help there so yeah how do you design solutions like they say in my experience is really about understanding both the

data and the user but if you just concentrate on one part you're somehow gonna fall short right and in our case we had really good access to users and really and smart and understanding

client and we had data to work with right from the start and this is a really big recommendation if you do such an applied and practical project get data right from the start don't really start properly before you have data

because you need to dive into that data and get a feeling of how it looks like what it affords what it what you can do with it what not and for us Jupiter and notebooks and tableau have been super

helpful to explore solutions real quick in the beginning we also had even an interesting interplay in that we would roll out little prototypes based on tableau dashboards that were actually

put into daily use before we even designed our tool so we could quickly test out ideas and some of our design ideas basically were directly implemented in tableau just to see how

it would play out on a concrete day and this really helps quite a bit then you always hit a wall with these standard tools so then I go into d3 and just quickly try out or with react quickly

try out very simple ugly quick explorations of we could do this we could do that just to have a basis for discussion and refinement but I also really like to think about is like what

is the basic metaphor we take what is how do we use visual language to communicate what we talked about sorry this is a German slide it's a German client so a lot of the internal documentation isn't German but the idea

here is ok the Train is actually a vessel so it's it has a fixed size right so we can work with that in terms of language and then it has different fill levels and we have opinions about these film levels right

so we have an attitude towards them and just thinking that through immediately suggests certain encodings make sense and others don't make so much sense right and this will also inform what types of designs are come

right so in these first phases obviously find the right angle again like how much do we zoom in and is it a black and white for our colors and yeah just take a lot of shots and see where you can end

up with always doubt the data never trust the data it's never complete it's never the thing you thought it would be it's usually also not measuring the right thing identify that early it's really super important because then you

know can we extend on that or can we make these deficiencies clear also in the design right it lasts if you and especially if you're in a managing position right give the project please give it a chance to change directions so

just because it was called the X dashboard or the Y globe doesn't mean it literally has to be a globe or a dashboard right so let these explorations inform what the format is

of what you roll out right and then when you have found a good combination of okay this view would make sense here and this view would make sense here then you can go more into refinement and

information architecture like how do we combine all these views into a meaningful whole right and in this case we had this drill down left to right so here you have hundreds of days a single

day some trains from that day and the single train from these right so you have a clear drill down left to right and always two of these panels fit onto

one screen if you go full HD so it's a very simple way to structure this application and present a lot of detail and still always provide orientation and to me this works best when you have your views later to cut them out and arrange

them in different ways and things like what do people need at the start what do they need on click what do these people need would do that people need it right yeah the other thing is of course more

iterative refinement as I said it's a much weaker so I like to just start and see where it takes me so here for these daily overviews we had a lot of like much more detailed charts in the beginning you will see these are the

same also so it just took screenshots and pasted them in just to see how it looks like from the bigger views and then in the refinement we did much more like okay we can get this much more on

point right but it was still helpful to have these rough ideas first even if they were too detailed same with the map here you know the rough version of the map proof-of-concept this is the refined one

where you have more like borders you have trails because it was hard to follow all these individual elements so these are all things that you realize when you just build a rough prototype

what you can still like get out of in the refinement phase also if you design by coding you have the big advantage that you can actually test individual alternatives so for these lists we were

really not quite sure which one's work best or this is the contrast here between bookings is the black part and the predictions so it's a bit like what we know for sure already versus what we think is gonna

happen and the prediction predictions are actually a bit more important and so we ended up with this solution but it was good to see all these different

alternatives yeah and a lot is really classic user interface design so at some point we're not thinking that much about is it a bubble or is it like you know

nobody really cares to be honest but how can you interact with this data and can you navigate easily and this is where the whole classic user interface design becomes interesting

so now luckily we're now we'd rolled out this tool it's been in use for months we have dozens of active users at hundreds of occasional users now we can actually track what people are doing and this is

the most exciting part for me so we built our own dashboard now we have KPIs where KPI back basically and we even we

built a little session Explorer tool where we have like a little like almost like a musical notation where we can sort of analyze how people navigate through that tool which decisions they make also which mistakes they make or

which art behavior the application shows and which different types of users we have so here we have somebody who just uses search so all the trains they select they just navigate through by search and all the navigation options

are unused so that's super interesting so maybe we should put the search much more to the front and not just the top right things like this and yeah just just quickly to see what we were able to

draw out of this so these lists we introduced very late but they're really the workhorses of the whole tool like 90 percent of activity is probably on these lists and the search right and this

shows us okay not everyone has like this huge need for overview and spotting patterns and people just want to go into individual trends real quick and make a decision on them right then

we also had this these two views so this is the main corridors like groups of trains through Germany and this is the same so these two show very similar information in different ways we

introduced this one and added those on request of the client yeah what do you think which one works better right inside right whose it was for right-hand

side left-hand side pretty even split so these are higher decisions right I mean they both have their pros and cons as it turns out this is used much more the left hand side is used much more

although it might look more complicated more intricate more maybe even more fancy you know but it's one that gives them exactly what they need and one actually chart they're very used to so

this is the novel one in this context and had a hard time to to compete but that's fine too so you know it's whatever works yeah and these maps like the multiples and the animated map you

know we love them we also show them every time fact is they don't get used all that much so you know it's like they are good for some very specific tasks and so we would never throw them out but

like one of the users actually said like yeah we like to bring them up when there's visitors around so but then they have a function too right I wouldn't discard that as then like just useless

because they help us in communicating the tool and the value of the tool and also the content because if you look at that view it could be flights as well it could be aquariums or whatnot you know

and here it becomes clear case logistics or trains here anyways and now we're in this lucky situation where we can actually iterate so we have a few iterations planned where we take input

from the user tracking we also do qualitative user interviews I haven't really talked about that but that's equally important to us take that into planning a sprint setting up goals and metrics for these new features so when

we improve in an area we immediately decide how it will be measure if we were successful implement that and then have a few months where it's in use and we can go back into into the loop and yeah

for us this is a big big part of it the other big part is what's actually happening to existing job profiles right and so this is something we also watch very closely like how does the introduction of automation and such a

tool actually change the jobs people do in some cases we had to think about already okay this part of the job is now totally obsolete what could they do in the meantime or how could they use the

tool now for new tasks right so this is this happen and has to be part of the process too so yeah just to sum this part up I think it's an interesting case

forward we can enable really complex human decisions in a really complex data situation so it's sort of at the sweet spot for visual analytics right you have enough data but still the data is not

enough to make the call and and to automate fully so we need to find these sweet spots where data visualization really works and again I try to advocate for a really data-driven but also user

centric design and so we as you have seen maybe you are as you have noticed we used visualization at the beginning to come up with ideas and to test ideas during the design use real data and then

even after all out use data visualization there are analysis to understand if our data visualization works it's sort of a bootstrapping approach and using really data visualization through the whole

lifecycle basically of the product of the design process and I think then it becomes clear oh wow is this universal tool right you can do a lot with it not just animated maps maybe and what's

interesting I don't know if you Millia with that video but John Tukey makes a very similar point in his like summary of the I think the first interactive data visualization system ever if I'm

correct it's it's pretty early at least in 1973 so I'll see if the sound works do we have sound and from its development beyond a recent conviction

the pictorial systems to be effective must go through many stages of trial and error learning as it did what have we

learned that will be useful at other locations we now understand that the details of control can make or break

such a system we now recognize that these details of control must be adapted to what is available and to the people of Hawaii huge

decision so it's kind of interesting so it talks about trial and error learning right which is iterative development yeah talks about how the details of control can make a break such a system that's

interaction design basically and must be adapted to what's available yeah like the data and the the technology and the people who use the system use a centric design so I think that that's really interesting how he summarized that

already in this successful of this very first system anyways so you've seen to Mary I think different projects right so as everyone's very user Center very

practical applied what is totally a technique vision review no I did not do any user testing on the Paris map I just did what I thought was right and so you

might ask well Moritz what's it's gonna be what's the right path what what are you telling us here right and I think the main thing I wanna tell you is why choose you know why would you want to make a decision here because this is the

unique quality we can really bring to the table it's this idea that depending on the setting you might draw more from the artistic and things inspiration from

the data and the technology but you also need to combine that with what people actually interested in even in the exhibition setting I need to think in a way use the center in what will the audience want to interact with in that

case right so it's much more subtle and maybe between the lines but it's still happening and so if you are more on the engineering side maybe or more on the practical side please my advice would

really take some time to just play and then just try out shapes and colors and take some photos see if you can make funny collages but also if you more on the artistic side like don't just spend

hundred percent of time there but also really try to build something small that's super useful like be to yourself being to your colleagues or you know on a job but really try to do something it's super applied not artistic just to

get a sense of okay what else can I bring in from this switch off of perspectives right and so as you might have seen I'm also much more becoming no interest of what happens actually before we visualize

like all the tidying up all the data analysis like what happens before we can even have the spreadsheet that we can turn - a bubble chart but also what happens afterwards right it's like well how does

this what we do change people a society what what what's the actual effect that we have and I'm almost up with the time but just quickly let me connect this to

a wider point so John Maeda always publishes this designing tech report and now if you're familiar with it and he talks about three kinds of design and reacted to talk about this quite a bit

and so he says there's classical design there's a right way to make what's perfect crafted and complete that's like very idealistic you know old notion of industrial design maybe you build a

perfect chair and that's produced a million times right in that way that's it that's the chair all right then maybe 10-15 years ago something came into play

that's more like do we need a chair let's think outside the box here and so that's the basically the design thinking approach like yeah stepping a bit out of okay what do people actually need let's maybe not just think out of work

processes or production processes but we're thinking about what do people need how does how can this change maybe what we come up with and now he says we're in the age of computational design design

for billions of individual people and in real time and scale intimidate all done what Stevie did forget to look that up but okay and the question is becomes much wider than like how does actually

the act of sitting you know transform our society maybe and you know do we need to sit all the time maybe standing desks are a good solution to and just take it a bit further in terms of impact

and in what we think about right and I think this is really interesting and he says there's like a lot of issues facing design like the challenges inside this eye but there's also also these

technological drivers that way now like part of and what struck me is we are really and the yellow ones are the ones where he says computational design is sort of can play a crucial role this

affects computational design your way and what struck me is this is also like affecting us quite a bit right so we talked a lot about diversity algorithm by stark your experience these are all areas where data visualization can play

a crucial role but also where we need to position ourselves and also these technologies you know all of them will change radically how we how we deal with digital information

I think data visualization as a general approach can can play a huge role in shaping all these areas and we should and what I mean with that is not just building tools in that context but

really bringing that idea to the table of being inquisitive being working evidence-based right and literally working actively with data but also bringing in the human implications and

bringing in the human capabilities and being playful and you know being like just I mean not not just become part of this bigger machine and yeah so I think

I really hope that we as the divisional visualizers can lead the way there and I think the same counts for it like from an engineering perspective right but I

do think under the condition that we Furr the work to on the one hand unify and professionalize the field I think yeah a lot of it is well first of all

I'm super happy that we have been a data visualization community that's super accepting of many different styles and ways of working and perspectives on the topic but we really need to make sure we

keep that and maybe open up even more and not be like too dismissive of people who just use standard tools or things like this but we need to make sure everybody who works in their devastation

is part of that community at the same time I think we need the thing about standards and shared ethics and values that we all have and that we all work on because if you just always say like yeah

but he's working this fancy micro genre so it doesn't really apply yeah or you to quickly say yeah but it's alright so I don't have to think about users you know then it can also stop a certain

progress if we make it too easy in their big experience for me is now connected to all these neighboring areas and rediscovering who's a UX designer or user interface insight which are you know really established disciplines who

know pretty well by now what they're doing so let's look into that same of course with data science and in business intelligence and I think that the last big step is all we really need to get

our our own chronic impostor syndrome because I think we all here in the room have experienced that feeling of oh actually I don't think I can do anything really well little bit of code and a little bit of

design and there's all these other people whose other feels they have like proper acronyms for their methods that I have nothing right and so and I think it's always this tension I think we do

need to get more professional and sort of have established method but at the same that is something so unique we can bring to the table that no one else can write because if it works out this

unique combination of art and science that I think only we can do that well we can really create magical information experiences that nobody else can bring

and with this I say Maxie and enjoy the conference

[Applause]