Governing Big Data: Video Panel

hi I'm James McGuire and our topic today is governing big data making sense of all your data analytic procedures talk about that I'm joined by three thought leaders with us is Aaron Cal director of product adulation hi Aaron how you doing I'm doing great how are you James good good and why don't you give us like you know the 11 second elevator pitch as to what elation does I'm not sure everyone knows absolutely so elation is the data catalog so what we do is we crawl and index all the data sources inside your enterprise so the data consumers can find relevant data understand what it means and Trust it so they can produce excellent it seems like there's definitely a need for them also with us is sugu ins who is a data governance specialist with the info data group hello to you soon hi good evening well apart from men I'm also the president of de mer international so so that's our international data management organization membership from lots of membership we have the data management body of knowledge a certification in data management so my day job I do data governance my night job I do the safe methey excellent okay also with us is Mike Matchett analyst with the tunisia group hello to you Mike how you doing today Jim good and I know you do you cover your data analytics from many different you know stand points yeah we've been studying big day yeah in all things storage and data management from sort of enterprise infrastructure perspective historically for you know a couple decades now as today's group and I focus a lot on what happens when you bring that big data in and do analytics on it and you have to persist it and store it and replicate it and protect it and of course govern it so try to bring a little industry perspective to your today nice ok so what does it mean actually to govern Big Data and if you can talk about perhaps you know one one or more they really won one sort of best practice in terms of governing big daddy what's your companies be aware of and what's your take on that sure I might not be the right person to have go first since we see this a little bit differently inhalation may be a standard but I'm yep yeah so really you know but to some people governance is really about defining policies and then is process of stewardship of put a minute to practice ensuring that all data is carefully annotated with how it shouldn't shouldn't be used what for example personated avaiable information what's high-quality that's sort of then and and adulation we definitely believe in that more traditional approach but we also think a lot about what's going to have the impact on the consumer so the governor's metaphor we often use is think about how to get people to drive correctly right Olivia be driver's test and someone has to like you know wipe the handbook and write that test and it will take it but what's really going to change behavior or just-in-time interventions like stoplights and speed limits that are posted and it elation we're all about trying to make that putting those causes into action right okay so you're talking about actually putting policies in action but heck I'm this you know company a I just want to plug in a bunch of numbers and get insights to help me drive sales I mean I'm going to put policies in place oh my god okay that never occurred to me that's exactly right that did I think there can sometimes be a perceived tension between analysts and data scientists who want to move fast and get insights and drive revenue you can learn more traditional governance or IT role who are trying to do risk mitigation and what we've observed is the thing that they share is animals want to be fast the governance ninety folks want to be compliant but everybody wants accuracy right so we find if you really make that V it's a central point attic attic we make sure all this data we if we know what it is and where it comes from and what it was good and trustworthy that can be something that's central to both those constituencies so what's your take what does it mean to govern Big Data and what is it what is the best practice for for comfort for governing big data so I was a couple of things I agree with Erin about I and and like we do talk about the yo policies of data stewards and all of those good things and compliance but what we miss when we when we focus on that is we miss the opportunity that the huge amount of data can give us and and what I think in terms of best practice is that balance that balance of yes we need to comply yes we need to know what our bet is where it is what it's about you know what the context of it but once we know that we have a lot better idea of of making that day to live and work for us so Stata government for me is that balance it's not a we're only going to focus on the compliance with only going to focus on the on the on the punishment side almost we want to get to the stage where in a great example here in Europe we're sort of focusing on GDP our um and everybody's like looking at GDP on you you can see really a night when I'm you know - what would it would you be searched gee GDP our status for the general data protection regulation of so it's an it's the European regulation which says if you have the data that's personally identifiable you need to manage it according to this regulation if any of this data belongs to any citizen of the EU so pretty much that's it's enormous it's so broad and in the US if you kind of go well we're in the US we don't care we don't have personally identifiable data sorry you do number one and two you have it about Europeans therefore you might as well you know give up and actually manage your data appropriately but this is a really big thing and you're getting lots of people with that sort of like deer-in-the-headlights look about it right now and one of the things I'm trying to do is in the Weimar and all the rest of it is get people to understand this isn't something to be afraid of this is an enormous opportunity to look at the data we've got I understand what it is and really use it and you know in in terms of your data size and your analytics and stuff imagine if you actually knew what your data was and you didn't have to get you made all of these big reports and most of them are wrong because you guessed what what you what you're doing so that for me is the best practice do to balance but sue I hate to break this to you I know you're living in Europe but we here in America don't believe in regulating anything she said if you can if you can get the data and make money off of it that's what matters the regulation stuff just slows us down so I know that's not a European outlook but it's what you know breaks that here oh well you know what you can regulate while you're making money I will give you update you don't need to yes yeah I mean if you know the regulation is about knowing where your data is knowing what your data is and actually looking after it appropriately if you want to make money the best way to make money is to do it right the first time how many times do you actually have to redo your analytics and your you know your calculations because you've got it wrong because you have the wrong data so you know what I'm sorry I put on your big-girl panties do the compliance and actually get value from your data I load your site I'll basic don't I because what Mike what's your sign say what what is it what does it mean to actually govern big data and and what might be a best practice for governing big data yeah I'm sure I'm sure I can distill this down to just one thing you know of court on that corner that's right right without being to flip but I think one of the things that people should automatically think of when they think of big data and they're thinking of volume and variety and and velocity they shall be thinking of veracity and we've looked at that for a long time and saying you just got to start thinking about data as something where you know what it is and you can show its provenance and you know whether it's true or not and in what it comes from I was a military intelligence officer a long time ago and in addition to classifying information and stamping a kind of security clearance on it one of the things you also should have been doing with that information is also stamping it with a rating of how believable it was or how much you could prove it to be true it is always an assessment and I think that's even truer Taino we put all this big data together is you have to have some understanding of the source of the data where it's from Hong it's updated all those things and people or data scientists running off with their machine learning and throwing things in Google Cloud get to these black box results and then come back later on and someone says what data did you use what did you put in the top of that black box and they go the black box does I certainly don't know where that data came from that I put into it you know and and if you don't know anything along those steps you're in big trouble so I think it's just a bet best practices don't let data scientists who are thinking and I think this broad before but running towards that you're getting that prediction line to 90% saying hey this is great because we can we can we found some optimization benefit by doing this run the show completely you have to be able to go back and not just for compliance reasons but you have to be able to go back you can talk about ethics privacy uh you know certainly meeting compliance law but even just getting it right and is gonna be a problem if you don't if you don't understand that data so that's what I would say and if I could have one more thing and you can cut this during yeah turn your editing sleeves okay yeah if you if you want but I think it's interesting as you bring up the talks about compliance and in this country we're having a big problem right now with sort of clash of of confusion ups to whether what we're seeking our freedoms to do what we want you know remove regulation and allowing people to capitalize on information and resources and do what they want versus and the freedom that you get from owning your own data and holding it private and secure and being able to have that freedom so you know we're we're we're we're the understanding what freedom means and I think tension there for sure yeah and so I think that has to be really hashed out and and and I you know super I have a lot more to add on it's just probably have a book to write on this on the balancing of opportunity cost with an opportunity you with individual freedoms and rights and and and so on but I think that is definitely a wide open concept to so enterprise is entering into this fray really are going to have to they really should be laying a groundwork to implement any kind of policy that emerges out of this right this is business we're plowing new ground and new rules and regulations are going to keep coming out and we don't know where it's going to end up five ten years from now but the more data you're accumulating the more it's going to pay to have metadata inabilities to implement policies and and have this provenance today so yeah you know I may relate I made up a question there tell me if this question they pretty sense but I want to ask it whether or not it makes sense like big data quality how do you manage data quality is that it first of all is that question make any sense and if it does please answer it totally okay what's my take on that so so I think there's a lot of ways to look at data quality I think the idea of you know but Rhapsody in trustworthiness like I was saying this makes total sense alright so consumers look at that so wanted to say let's look at the values right if we're saying you know date of birth from an enormous spike at age you know 47 that could be because the know values in some ETL transfer got defaulted to January 1st 1970 right not because it's actually what made people are it's just the way the data is skewed rather could be some place in the pipeline that breaks and things get stale or certain records could be duplicated multiple sources in MMA constitutions 100 reasons why they did it can be wrong that you might be able to tell from foe calling the values you might be able to tell from looking at the lineage and the provenance all of that is super helpful in the big part of revelation tries to surface from the big data layer but I think it's going to look at to though is people or a great source of information about about accuracy and quality and the analogy I would rise to Google so before Google the way search engines worked if I want to know if something is relevant to a keyword that I type is they see how often that keyword occurs in the content we all said to be other words right and the fact of the matter is that's looking at the data itself not the metadata on the usage right so - you know that's with you know Alta Vista or something like that you would see a lot of results that weren't very good in the same way that a restaurant might say you know best tasting food that might not be true what I want to see is the big line at the door are people in there enjoying it are people on Yelp saying good things and that Google people are developing right exactly exactly right oh we're going to look at the behavior what do people know instead of interviewing hundreds of people though about which website is the best upset about dogs or about medicine they instead looked at the links they were already making roaming behavior already and dogs of the mine has been havior exactly right so so in the is the way we would look at the world's because okay what are the things people are doing with data the activities in the day-to-day life that show their preferences and their sense of quality so if the best analysts are all creating their dashboards using this particular data set and the other data said that no one's using only some summer intern is using that can be a good indication as to what's high-quality so I think quality you can look at both the content both the content the provenance and the usage downstream as input into that process the idea of data curation is being the task of taking all those inputs and kind of making that final amalgamation a human judgment for what's good much not so do you believe there is such a thing as managing big data quality I think about you know systems to suck up enormous fields of data structured unstructured data and it's like the more the merrier is you know managing big data quality this makes sense what what is it yeah very good you go soon as I would you know see what is your take on um so absolutely you know I keep getting asked oh I keep getting told by all the big data groups who say yeah we've got so much data we actually don't need to do quality on it right now we've got so much that the sheer volume will will will even out the data quality curve and I'm like well okay if you've got three billion records of data of those three billion 2.99 billion actually have great email addresses as a good as an example then you're right anything that's wrong will will actually will actually you know foot fall out however when you have bad data quality already you know then you are going to have a problem you're you're absolutely going to have an issue and that you have to look at your quality you've got to decide you I think you have to actively go yes I realize I've got big datasets but I've done a little bit of Internet analysis I've tried to pull some stuff out and my results are not what I'm expecting and trust me a good data person knows the results they're expecting it is something it's almost a gut feel that we have that we can look at and we can say no that's not right and when you're when you're making when you're looking at that and you're looking at your data and you're going that isn't right it's time to look at the quality absolutely something if it's a quality and I don't care how big your data is you know we've had we've had big data for for four years so did anybody con you into this is a new thing we've had enormous tons of data lying all over the place for a long time so in my mind well so he was wood how else of those large database companies been a business away if we haven't had massive stories of data for you know four years if not decades but let me use the e word a second screen Mike ethics and Big Data it's almost like it's an amazingly hot potato issue I mean what role does up explain but I guess the real question is are our companies really thinking in terms of ethics because so big big yeah yeah so of course that's a leading question because there that's well know yeah obviously nowhere near enough companies are thinking about ethics the way they should be and big data does compound the problem and you know that just to give a few kind and a broad brush examples without without going over things that are in the popular press you know everyone today is concerned with the discussion between artificial intelligence and sentience maybe taking over the world and is that a good or bad thing and people are arguing back and forth I'm going you're missing that you're missing the point there we don't have our official assistance today it's you know it's it's probably not going to happen that way what you should be more concerned with his machine learning and how that's being used because what people are doing is saying let me take this big massive data draw a pattern out of it using some mathematical algorithm and then apply that pattern as if it was the rule of law and there's no recourse or or insight or judgment or even defense of what that that algorithm is coming up with it's a it's a it's not only the algorithm is a black box to most people although you can study those things in terms of probabilities now but if the application of that is is beyond the realm of most business folks when they get that they pretty much the application of law of a machine learning recommendation for example I mean you if you're not the data scientists that built it and understand the variations in the in the probabilities and things and that goes back to the data if you don't know what data went into it you can't draw a good conclusion but what we're finding today is people are putting these things into practice selling them as services to others who just put blind faith into it so people are losing their jobs or not getting a job they on the zip code they're in because of some big data pattern the people there have been teachers who've been fired because the wrong data went into the systems and there's a third party and there was no recourse legal or otherwise to even go and argue what data did you use how do I come up with this pattern it was just a blind recommendation we know that the way a lot of it's going it's going to be machine learning driven and if the machine learning algorithms only 90 percent or so on that may be better than the human could do in real time but that leaves that 10 percent hanging out there and and I'm not talking just about being some optimal but I'm talking about actually implementing things of they're prejudicial things that are right that are actually against our civil liberties against they're against our laws in this country as well as possibly impinging on human rights worldwide when you do this so I think there's great ethical considerations and I wouldn't just say itself but I would say what big is getting use for which is in creating these machine learning and optimization things we gotta be really careful with that part going forward right I mean there's a really obviously it's not a stretch to say that you know big data inch of an orwellian stay are pretty close to each other or could be pretty close to each other aryn GGC what do you see in the area of ethics and big data our company's aware of it are they thinking about right there they are they worried about it or they mainly running right over it yeah you know I think obviously like individuals companies follow on a broad spectrum as how much they're concerned with ethics versus you know more of a get-rich-quick mentality and like you said you know these cultural norms might differ in Europe versus the US and how we think about freedom to which is freedom from and that sort of thing we're just people Mike was saying I think that there's a risk but also an opportunity so obviously the risk comes from an application where decisions being made entirely by the computer in a way that could be either either mistaken or there could be reinforcing an existing bias right so just to spell out I think Michael would really Bella but a particular example let's say that I look at the hiring decisions the human hiring managers make based on interviews and I try to have that Fawcett's be automated by looking at a bunch of different features and variables from a survey or a LinkedIn profile or whatever else well if there's a bias based on gender or race something like that even if it's conscious bias in the humans that bias is going to be encoded in the training data and therefore that bias is going to be put into that algorithm but the opportunity is to say you know big good it could be a tool to correct that instead of to reinforce them for example if there's less bias in performance evaluation how you're actually doing them a job then there is when you're looking at somebody and saying oh they don't look like the kind of person to do this job I'm going to judge them if you trained on that you could actually take cumin to our bias and instead of replace them with a computer that has the same biases and is biased faster because the computer it could be something that isn't biased because it's actually able to look at what the real evidence shows instead of just what the assumptions are so I think I do a mixture of companies who are thinking about how to have a big data and machine learning be a tool for good folks who aren't thinking about it that much at all and I hope we'll see a trend toward the former are you saying that a say for example a program that might be searching for applicants that human bias could be built into the actual searching mechanism so that they understand age okay yeah so basically if you machine learning model is you you take you take a bunch of data in the world right you say okay given these inputs it would you know how do we wait them to predict the output and if that output is being done in a subjective way that has biases in it the computer is going to naturally replicate that rather than overt comment right if you have somebody doing a pre big data model the more kind of a rule based algorithms if that person in charge of the algorithm knows the law and knows cognitive biases and everything else they could make sure that they never take into account race or gender things that are correlated with that in their algorithm very conscientiously if you just say we're going to do unsupervised machine learning and the bulk of the big data will solve the problem it won't solve that problem right yeah so your your our token European and so we have to ask you about ethics too and we know that you may have a different view on us but what did you take on a big day and ethics our company is really there with it or is it mainly being ignored in the Western Province I think here in Europe we're probably a lot more conscious of the effects I think generally as as a culture we're sort of more formal we're a little bit more reserved we're more I think open to considering that ethics is something that we sort of take we've going to technologists off and of course I mean part of that is whether we like it or not we are being driven by the data protection acts that are around and the new DPR that tells you you've got and on your data ethically we've got to handle our dates immorally so whether whether we want to or not there's an enormous compliance drive towards it and organisations who who have taken that card and are really doing it are the organizations where where ethics is already there and they're ready looking at it and going alright so but what they use big data how can I use it how can I get value out of it but how can I not tread on anybody else's so so it is there and yes absolutely I was Erin and and my there are organizations who are not doing it and her doing their damnedest to to avoid anything that could be sort of smack of any kind of ethics but in general I think there's a big drive and again from my perspective of Zima International our new our new body of knowledge came up at the end of June and we have a chapter on ethics and ethics in that book and it's probably you know it's kind of quite rare at the stage to have it written down about it but we decided that this was a topic that was really close to our heart and that we needed to bring to the fore it doesn't matter what you do as a data management person or an organization that handles data you have a duty and a responsibility to use that data appropriately and to consider what the outcome is if you do not what one of my clients that I'm working with has the most insanely sensitive data about people and if the states are got out it could effectively destroy knives and I mean it would individually destroy life and which in a way is harder than the data that just literally is let loose in hundreds of millions you know you know you know a lot of the breath data breaches are millions of Records these could actively destroy individual flies so we have to think about it we have to be realistic and as data management people data management professionals we have a duty to make sure that if there is something being done unethically we actually call it out we step up and we say guys there is a problem you cannot continue to do this and we have to be the ethical evangelist and whether we like that or not just to be the right the right person to have a profession that we're proud of we need to do that good point yeah Mike would you take look looking to the future say we're you know we're having this conversation in the year 2022 we're just not that far from now and we're talking about you know big data governance you know what will be all settled by then what will still be an issue then whether its ethics whether it's procedure whether it's it's you know logistics what what is the future a big data governance you know in several years from now so I'm just gonna I'm gonna talk about two things that may just go around the edges of that and that's to talk about we the core the core that is more their bailiwick but one thing one thing I've been looking at is you know with big data you can mask out PII fields you know personally identifying information out of one data set and out of another and another but the trick with big data is you can put all that data together and reverse-engineer back personally identifying information if you got big enough data sets okay and and I don't know how to engineer our way out of that because you know there's it's a very difficult thing to say I'm going to get the benefits of big data by pulling all those data together and if I mask all the individual and identifying information upstream I should be okay but then to realize that in the final product when I put it together I basically can reassemble information and I guess it's on ethics and hits on data quality in sort of things so that I'm not sure how we engineer our way out of by 2020 because I think our power of analytics only goes up by then and our power to draw conclusions and I think the Box only gets smarter in that sense of being able to create a so that's one concern I have is I think that just gets a gets to be a bit more difficult problem as we go along the second one I think people are going to face that they're not necessarily aware of is what I've been calling a data paucity or poverty of data and that is you might be a company or startup coming along now and saying I got this great idea and I can do this and this and change the world and improve everyone's but if I need access to that map data and supply chain I don't get it I can't have it data and and we haven't quite seen that yet but I think that's coming for some folks who say like I've got I've got a way to really improve that and I don't have the rights to get into that I don't have the money to buy into that supply chain of data from from folks I know data itself is becoming a currency if you will and I think a lot of folks who are sort of saying like no the big data makes a little better we're going to find out that once we become more of an information market you have to buy and sell and and get access that data that that becomes its own challenge and even specific if I make a widget and I even put making an IOT widget for example so Internet of Things think so generates its own little data but I sell that downstream who - people were embedded in larger and larger objects who has the rights to the data coming off that widget who's going to get access to the data they're gonna get all the context needed to understand that and that just because a huge I think a bunch of bigger problems we get to 20 billion twenty twenty two hundred billion twenty trillion Internet of Things objects generating data into by 2020 right so this is an is not like today problem necessarily but by 2020 I think the data flows and the this kind of bigger and bigger issue and your take on the future of data governance or perhaps big data itself and with the world of ideas yeah where's it touch well I think it was really cool as be on his panel a couple years ago we really felt like we said oh data governance is about quality it's about empowering analytics instead of just preventing misuse we sort of felt like there was there wasn't anybody with a lot of people and now I'm hearing sue and Mike talk instead of saying the exact same things all about finding understanding an accuracy so I think already we're trending toward new 2020 we're just thinking about it more holistically and we're collaboratively is it an adversarial relationship between good sort of data using the data controllers I think it will become more of a curation and consumption market which is a lot of a mutual benefit and and I think there's a few types we're like data really cannot be misused you know in a Stu mentioned a certain use cases where the data falling into the wrong hands could be truly devastating but mostly attentive czar really aligned right getting access to data that is accurate is trustworthy being able to find that most people also want to use it correctly so you can think about it almost like going to the supermarket and being told you know this contains not eat it you know only if you're not allergic but we're not going to like you know forbid it right Alice where it's informed consent and an alignment sentence and I think that's goes off so it's going to happen and obviously I have a bias but I think tools like data catalogues become the marketplace for going between these two roles people who know about data people who are going to use the data so that you can both avoid that downside risk of a lawsuit or something even worse while also getting that benefit of leveraging that data to make better decisions better product to make people's lives better so I hope that clarifies a bit it makes sense so you're taking the future and there's something that I think maybe Mike and perhaps even error instead that I created image of my mind of like the future of big data may be affected by the hazards have NASA other words getting access to the data is going to take more and more money and so the competitive edge at big data which will be ever more important will be available into those with large amounts of money is I hope that's not the case yeah if I think actually what we'll see is just as things like you know Google and you know the first web browsers like Netscape made the cost both in money and technical know-how of getting access to information get lower I think will actually be a data democratization in the coming years Susan for data catalogues it will help people in organizations and across get access to the data that to happen don't need to be assurance to both GDP are and other lenses that nothing get leaked but I think you know if you're in a safe environment where you can sort of self-serve but I do that was the Amazon Marketplace right I can do it good search and see many products and many vendors but I get those star ratings and Amazon makes sure that we're not selling poison to kids or anything so it cut out a nice environment where it's safe to self-serve so you take on the future of data Government's governance what are we going to be talking about civil years now on this topic so I think that the one thing that we as data management people kind of forget is where we're we're already you know we're in we're in data management we're doing data governance we're doing Big Data we're doing one of those things there is a generation of youngsters people coming up behind us with incredibly different attitudes towards data now if you have teenagers or you have access to a bunch of teenagers I challenge you to go and spend a couple of hours watching them and watching how they work with their stuff how they work with their technology how they work with the data that they actually are having but they they multitask and they share and they do also think but they do it in a very targeted fashion and I have a suspicion that 5-10 years time we'll be sitting on our rocking chairs looking at this thing going while we never saw that coming I think there's going to be an explosion of a lot of self-service and a lot of different ways of looking at data I think that people are going to get more intelligence about their own data you know typically you go into an organization Mesa and it's either well that the CEO owns it or the organization owns it all or I'm a department in charge of marketing I own it I actually know who owns the data the person whose data it is I own my data you own your data and the generation coming up behind us know that they know that and they're not going to let organizations take charge of their data so I think that yes governance is going to happen but I think it's going to be a different kind of governance I think that the right of the generation of people who already know and are intelligent enough to know how important their data is in the world they're going to manage things very differently I'm actually quite interested to be able to sit back in ten years time and go okay now kiss I want to see what you make of this you know things like GDP are doom is a part of me that's going I wonder if it's going to exist in ten years time I absolutely wonder you never want to need it well while I all know that we'll need it well it's not a highway up to because those teenagers yeah yeah those teenagers are a hell of a lot brighter than we give them credit for and they think differently to us we are still of the oh my god let's look after it and protect it whereas they're almost of the mindset of well it's mine I'll protect it the way I want and I may choose to share it with you but when I do watch out because if you mess it up here I come and they like that you know so though I have an unusual view I do get that right I want to believe you're right about that and certainly you know younger people today are more tech-savvy than maybe we would have been pressed when we were like that age on the other hand they're more immersed than than we ever were at that age and I'm you think about you know everything just gets posted Instagram are though is the every sixteen-year-old making such a careful decision about what he/she posted to Instagram so that you know certainly the data is not as freely available out there so they are making that decision because if you challenge this and you turn around and say to them hey teenager a General Peter whatever you do realize that this is not now this is not private and everybody who's got access to you can see it and they can pass it on and there's an attitude of yes I know I chose to do it you know we are oh my god forget that picture go out and another challenge I see and I hate to bring it up but Facebook you know I keep telling people if you want to post something on Facebook please remember that you can't get rid of it I don't care what Facebook says they don't delete data therefore if you post a picture of yourself drunk-and-disorderly on a Saturday night it's out there forever okay and it me I'm kind of going I don't want to do that because what if my possible new job the boss there goes that happens to go on Facebook and have a look and see me drunk and disorderly on effect they're not in my own time what if you won't hire me okay whereas a teenager is going if that dude comes along and actually looks at me on a Saturday night that's his problem because it's might come off as long as I get to work on time I do my job properly who cares what I do on a Saturday night a big change of mindset and this is going to change how base is used in the future all right well yeah sorry I really no no that's what you know I get your point and I'm I see the clog the big clock is telling us we are more or less at a time I feel like I've learned a lot I thank the three of you for your expertise I'll send you the link when world you know when it's all done we get all tweet about it and I thank you very much

This transcript was generated automatically from the video's captions and may contain errors.