Entertainment And Data

Speaker A: 00:00:00

Hello and welcome to the Cognitive Engineering podcast produced by Tell Me Studios for Aleph Insights. In this series of podcasts we take a look at interesting topics and discuss what we think they tell us about analysis and decision making. I'm Fraser McGruer and I'm here with Peter Coghill and Chris Wragg of Aleph Insights and this week we're discussing the connection between data and entertainment. So as you heard there we've joined by Chris Wragg, you've just joined Aleph Insights so tell us a bit about yourself.

Speaker B: 00:00:36

Yeah so my background really is sort of over a decade working in government in various roles supporting decision making from everything from sort of intelligence analysis and security and counterterrorism work through to speech writing. So the full range of kind of civil servant type roles prior to that a sort of academic background but joined Aleph Insights recently by virtue of

Speaker A: 00:01:08

my interest in the stories of Borges. Well of course I think that's an important requisite

Speaker B: 00:01:15

for working at Aleph Insights yes. Did you hear our Borges? I did I did that was part of the

Speaker A: 00:01:22

recruitment drives. There we go excellent. Okay well look so as I said we're talking about data entertainment so from what you've just said there you're eminently qualified to talk about this. So Chris if you can lead us in on this can you tell me how data is being used to measure and modify how we consume entertainment and I suppose we're talking what we see most about here we're talking about film and TV consumption I think. Yeah no that's right I mean I suppose I first

Speaker B: 00:01:47

started thinking about this based on my recent experience of watching the Netflix series Stranger Things and I started watching it and thoroughly enjoyed it but as I went through the show I started to get this sense that it was specifically tailored for me it kind of had all the things that I would look for all the ingredients in a show it was set in the kind of early 80s there were lots of geeky references to things like Dungeons and Dragons it was science fiction the mix of characters was right from the middle-aged through to teenagers through to young children and you know representing the various stages of my my life so I just I just got this sort of eerie sense that here was something that was that was perfect for me and you know it almost been produced for me specifically and it turns out that you know based on Netflix collection and analysis of data that it that it may well have been if not tailored specifically for me for people like me. So one thing I was going to ask is how that

Speaker A: 00:02:56

made you feel and maybe you answered it because you said this eerie sense. Yeah well I actually

Speaker B: 00:03:01

I mean it's interesting because although the fact that I'd been modelled I suppose didn't actually reduce my enjoyment of the series anymore you know I suppose it's a little bit like when you have an attentive host who you know brings you that cup of tea at the at the right moment they've simply been observing what it is you like and are supplying that at the point at which you want it and you don't feel you don't feel a sense of having been duped in some way by

Speaker C: 00:03:36

that and so it doesn't feel creepy it just feels kind of right. Yeah so it's got this benevolent

Speaker A: 00:03:42

omniscient omnipotent presence great nothing can go wrong there. So okay well look you started talking about the data there I want to bring in Peter so what are what are Netflix doing to to help us or feel Chris and others feel all fluffy and looked after like this? So what I think

Speaker C: 00:03:58

they're doing is that they are making use of the huge amounts of data they can collect about their their customers so online on-demand suppliers like Netflix, Amazon, Google, even BT and other sort of more traditional suppliers they can collect by virtue of the platform that they deliver their content through a huge amount of information about you as individuals so your age, where you live, they can do some sort of basic analysis and work out what your sort of family life is like, your movements etc and they can and also be also your watching habits so whether or not you've watched something to the completion how often you've watched it whether or not you repeat watched it if you've and and if you've recommended and shared it with people all of this stuff tells them a great deal about how much you sort of like a particular show. I mean

Speaker A: 00:05:00

obvious things like the date that you watch something where you're watching it as in your your postcode but there was no things like when you pause it and when you leave it when you come back to it you're scrolling and search browsing habits through the through the interface. I'm sure there's other sort of also they know what device you're watching on as well yeah so that's how they do it so what's the next what do they do next and how do they get from having that data to to molding the entertainment the in-house entertainment that they put out to us so with

Speaker C: 00:05:36

that with that data tells you a great deal about how successful any given show or any given episode or any given portion of any given program is because you also can you can detect whether or not people just watch a particular scene and share that because it's got some significance they you can build a model which correlates all of that data with certain facets or features of a show so you can you can analyze a show you can analyze a particular episode and say okay so in this episode there's a wedding and this thing happens and it's got these characters and it's got these actors it's directed by this person it's written by this person all these are sort of features which are present in all shows and there may be repetition of some of these specific features across shows and then you can and then you can you can you can model how popular a show is and see if there are any significant indicators in any of these features and early attempts include Netflix I think they had a quite a detailed genre system of different shows rather than just classics of comedy action etc they were much more much more much more granular tags that shows were given but beyond beyond the just simple classification of shows and features of shows you can even really go down into you know what is the color hue in particular parts of scenes what is you know what's the music to dialogue ratio what what's the content of the dialogue and I think

Speaker B: 00:07:07

that's that's really relevant for a show like Stranger Things where the tone of the show was very very particular so everything from the opening credits created a particular type of atmosphere about about the show which was very redolent of the types of things that people like me like so I don't know whether it is true but it felt like that had been based on some kind

Speaker A: 00:07:36

of analysis okay so I've got a couple of things I want to say I'm not going to say them yet but so that kind of that's our foundation here right so through all the marvelous ways that they're able to collect data they're able to commission types of shows that they'll know that certain audiences will like okay so where do we go from there so well all I think it's

Speaker B: 00:07:58

interesting because at the moment so I think you know what computers can do is or rather what you know machine learning can do is to tell you what the ingredients for success are but what they can't do at the moment is create that success or you know write write the screenplay for you or act in it or tell you know or direct it tell you what you know what emotion should be in each each scene so that they're not at the stage to be able to to do that but I can see and you know people listening may be thinking well it's you know there's no substitute for good acting for example and it's preposterous to suppose that that could be done by algorithm but there may be you know the next stage may be more of a fusion between the randomness of current creativity and a bit more structure around that as to actually saying okay so you know across a 12-part series or something we we feel you know there are these general emotional ups and downs and that you know individual scene patterns you know here's the most successful pattern of scene emotions

Speaker A: 00:09:18

for example yeah okay so I think that's that's interesting that's something I wanted to come on to actually one which is to do we were talking about Sunspring this yeah this complements what we talked about a few weeks ago was it Sunspring if I got that right yeah the film that was compute there was screen written by a by a robot let's all call it and it was bloody awful so let's but but one of the things I wanted to that you mentioned there was it can improve or help the randomness of creativity which is an element of me immediately doesn't like that but actually that already happens a lot and you know probably since antiquity ever since stories have been told there's a certain formula that stories tend to have but also even now probably some of my favorite authors um there's there's there's there's book writing software out there to sort of help um the the structure of a book but anyway the bit I want to explore there that Chris started talking about was the the Sunspring elements let's say and Peter do you want to come in on there yeah I think so

Speaker C: 00:10:15

so I think that some your reaction to I don't like this in this incursion into creativity uh you could ask does it does it take some of the joy out of creativity does it take some of the sort of humanity out of creativity um sorry to interrupt you but isn't randomness one of the beautiful things about creativity but anyway yeah possibly but I think it is is a feature the defining feature of creativity but you know if if if you're if you're able to use all this data modeling to provide you with a stimulus or a seed into your creativity that is more likely to give people what they want and and by extension make your make you allow you to make more money be more successful in creating shows there and and so you're you're able to provide a great amount of joy and enjoyment for for a greater number of people at a higher level of quality with a lower level of failure then you know potentially it's there is joy there is there is increased sort of goodness in this situation I mean I think there's a there's a human uh human trait to romanticize failure about you know the failed artist the van gogh um he was he was underappreciated in his lifetime and then only met only met fame posthumously now um that's the there's a romantic notion to that but had he had something similar so he could have been a jobbing artist making enough money uh by painting what people wanted one day a week say and then the rest of time he could have just done this taken the safeties off and just created the amazing art he did he would have probably lived longer been help healthier and happier

Speaker A: 00:11:59

and we would have had more van goghs well I disagree actually because I think first of all he might have found his creative um spark deadened by having to do I'm I work in a creative field myself and I know what it is to have sort of to do uh commercial work that I'm really not keen on

Speaker B: 00:12:17

doing um I mean I I think there are two sort of things I would uh I would um say there what one is that um sometimes constraint actually creates creativity you know it's the fact that you haven't got a particular thing uh which which enables you to um to be more creative so by setting uh boundaries and constraints around what it is you're doing that can sometimes uh increase um creativity I would also say the current model of um creating something that is generically appealing in entertainment tends to lead to lowest common denominators uh I would hope something like this would enable you to take more of a gamble on niche independent type projects so that you could say something like stranger stranger things may not be the biggest ratings success there is but they know um with greater confidence that the niche that they're going for will like this thing there's a defined niche they know what its requirements are and they meet those requirements and so actually it may lead to greater diversification and less generic scattergun approach to to trying to make yeah successful I think that's a good point because

Speaker A: 00:13:37

I was actually going to talk about this sort of generic and lowest common denominator but actually the point you make is is good because you can have greater faith that yes your audience is there and they will go for this and it can be niche and I mean what I was going to say is that you know before Netflix was a thing people you know say five ten years ago people were saying look this recommendations you get on Amazon for example is just really actually um um suppressing creativity suppressing us from exploring genuinely different stuff and I still think that's the case so for example on Netflix I challenge any first of all I look at the recommendations it gives me and I think why the hell are you recommending that to me because I have zero interest in that and it just looks quite blunt the the algorithms that they're using but the second thing is I challenge you to find more than about three interesting documentaries on Netflix okay they're just most of them are just rubbish and they tend to be quite American oriented um and I think the reason why is this is to do with um I think it's for commercial reasons that it could be quite difficult to offer something that's genuinely different and esoteric some slightly bizarre documentary that um I think even that I don't know I think it looks even that looks too dangerous but surely even if it's not produced in-house it can't cost them that much I don't I don't know the answer there but nonetheless I think you're right because something like um what was the American show about um producing crack cocaine what was that one oh no about meth rather breaking bad breaking bad I mean fantastic that's a great example that's something genuinely different and I was hooked I absolutely loved it and I think that is quite a different kind of show so that's a long that was the last three minutes of me more or less saying I agree with you um so where are we what what what next

Speaker C: 00:15:18

um yeah so yeah so I think yeah the um historically uh when production houses have said like we need we we want to appeal to a market of particular types of people uh we want to make this movie you know the the the chick flick kind of movie we're going to appeal to a certain type of people I think the analysis they have you know previous have done analysis but it's been very basic and demographic so they've said right we want to appeal to teenage girls between the age of 13 and 18 with this movie and we think we've got an idea of what movies they appeals to them based on historical analysis but what this this highly data-driven approach allows you to do is say well we're not going to presuppose the classifications of people we're going to allow the machine to tell us what these unique people groups look like and what their intersections are so the the output of a machine learning algorithm might not make any or logical sense to us so there might be groups of people that have common features but nothing sort of physical or real world commonality that we would recognize if we if we met them but for whatever reason their viewing habits and their viewing tastes are similar and that's and that that may be why some of the recommendations that things like Netflix uh present to you are odd because part of your part of your um part of your collection of your part of your class of people happens to overlap with a class of people that probably like something else but you're on the fringe and you're

Speaker A: 00:16:50

not within that overlap I think you're maybe right but I actually think the the reason is probably more prosaic than that which I just don't think they've got enough content that's what I think

Speaker C: 00:16:59

the issue is yeah but I think they do they do incur a cost whenever they buy a license to to host and there's a minimal very small but still a cost involved in hosting and distributing um things so they have to make a business call as to what it's like but I just wanted to sort

Speaker B: 00:17:16

of talk um you know on Peter's point about uh demographics there and the evolution of demographics based on machine learning uh algorithms I think this is really sort of profound change that's going on at the moment uh that could have impacts you know more broadly on on society and this is the notion that the traditional um demographic classifications that we've used to work to predict things may be starting to break down as we get more valuable more insightful information about people so that actually you know age gender ethnic background employment those kinds of things they may still have some predictive power but there might be other things if we're trying to predict your viewing habits that may be much much better and that we can't conceive of you know they're they're um sort of multiple variables collapsed together and we as humans don't understand them but the machine machine learning algorithms do and they um make us offerings on that on that basis and I think this this could be something that um will will start to change

Speaker A: 00:18:26

us at a societal level potentially and do you think I mean so I mean what we're talking about here I mean we need to wrap up but what we're talking about here is the application of big data right and in this case the application of big data within entertainment when you talk about changing our lives and changing society I presume you're not talking just about within entertainment

Speaker B: 00:18:48

you're talking about within all aspects right yes exactly I mean um you know I I can see I can see a situation where at the moment you know we perceive ourselves predominantly through visible characteristics I can see a time or I can see the potential that we shift away from that way of uh deciding you know what social uh and ethnic group we are we are in towards viewing ourselves more based on um classifiers that aren't aren't visible but are only visible to computers um and and and

Speaker C: 00:19:25

potentially much more behavior based so um at the moment it's easy to measure stuff like your age and your your sex and where you live and things which is all um recorded by various uh institutions but your but what you actually do day to day what you watch where you go who you talk to um all this stuff is largely unrecorded but more and more of it is being recorded by by um things that we subscribe to our social media platforms etc so the the this provides you with a lot more rich data about how you know the drilling down into who a Fraser McGruer actually is and who he is more like is he more like a uh a white middle-aged person or is he more like uh other people who have different characteristics other than the age and his his racial background

Speaker A: 00:20:21

sure nice and anything else you'd like to say Peter um yeah so I think another thing that these

Speaker C: 00:20:26

new uh providing platforms let you do and Amazon definitely do this Netflix do this

Speaker A: 00:20:33

by the way you say Amazon in a very American way but go on sorry Amazon yeah yeah like that um

Speaker C: 00:20:39

so they they what they let you do is um produce a pilot program and then uh and then gather data on the success of that pilot which gives you a great deal of information about whether or not it's worth producing series one two and three um so what this means is it allows you to conduct a greater level of experimentation than was ever possible before pilot programs on terrestrial tv etc would only ever really give you a loose figure for viewing but would not tell you anything about who viewed it and whether they liked it so it wouldn't tell you anything about how you then market that program and what the target market is likely to be so um yeah I mean Amazon spits out new pilot programs sort of on a weekly basis um and uh some of them are excellent some of them are pretty atrocious but again it depends on your particular taste okay let's wrap up just by

Speaker A: 00:21:30

quickly asking either on on whatever platform whether it's Netflix or Amazon Instant or whatever

Speaker C: 00:21:35

ite I think it started around: 2009

Speaker B: 00:22:13

probably also on it's probably on lots of different okay different and for yourself Chris uh well obviously Stranger Things was a big was a big hit with me uh but uh beyond that um I'm uh not not shamefully watching but um uh watching uh Pole Dark at the moment and thoroughly thoroughly

Speaker A: 00:22:31

enjoying it oh that is good yeah interesting interesting and there's also it's nice there's a sort of terrestrial program there um yeah for me I think the most recent thing was Narcos um which I don't know I I do enjoy it but I don't think it's that good really but I think it's a little bit cliched but I'm just into your drugs aren't you yeah that's you know yeah well assassinating stuff in Latin America that's me you know um and then the other thing as you know I was quite ill over the summer and I spent about two months in bed and and so I ended up watching Vikings which is a series on on Amazon and I think I watched something like 36 36 episodes in quite quick succession and I've really a bit like with the Olympics I've got a really bad association strong association with feeling terribly sick and there might be a reason why you're sick if you watch 36 hours of television on the trot well I think it's called appropriately called binge watching because I did feel like vomiting at the end of it anyway um okay well look guys thank you very much um as always you've been listening to the Cognitive Engineering Podcast with Aleph Insights with Chris Wragg and with Peter Coghill um thank you for listening and until next time goodbye