Skip to main content
Search form
  • 00:10

    STEVEN MCDERMOTT: My name is Steven McDermott.[Dr. Steven McDermott, Qualitative Analysisand Social Media Lead, HM Revenue and Customs]And for the last 18 years, I've been studying social media.I started off studying social mediaand studying the law in Singapore,looking at political discourse, looking

  • 00:30

    STEVEN MCDERMOTT [continued]: at Singaporean elections, but doing itin a way that was much more qualitativeand really had an ethnographic component to it.What I did was for the first I think it was five or six years,I didn't say that I wasn't Singaporean,I didn't say that I was.I just happened to be living in Singaporeand studying and teaching at the same time

  • 00:51

    STEVEN MCDERMOTT [continued]: and posting articles that I was interested.And suddenly, people started gathering aroundall of the jargon, if you like, in the blog.And that's why I really got started.It began with 12 of us during Singaporean elections basicallytalking about politics in the Singaporean context, mostly

  • 01:12

    STEVEN MCDERMOTT [continued]: anti-PAP, Peoples Action Party, the vast majority of it.And that's where my interest started a long time ago.Since then, I've worked as a university lecturer,a researcher.If you're academic, you've done that;you understand, one ternary contract after another.Very limited job security, if it even exists at all.

  • 01:36

    STEVEN MCDERMOTT [continued]: Yes?At one stage during my PhD, I was doing four jobs justto get by on a daily basis.After finishing my PhD in a panic,luckily I managed to get a job three yearsuniversity lecturer.Three years in, I realized, all I'm doing is teaching.

  • 01:57

    STEVEN MCDERMOTT [continued]: Teaching, teaching, teaching.What happened to the research?I was applying for grants, but not getting much.A little bit of money.After three years, I realized, if I carry on like this,I'm going to be de-skilled.I can't deliver to my students what I should be delivering- yes?Just focusing on teaching what I already know.So I decided to leave academia in order to do some research.

  • 02:22

    STEVEN MCDERMOTT [continued]: That's the state of affairs at the minute, I think.So I ended up going to, of all places, HMRC.Is that from the kettle to the fire, maybe not.And that's what I'm going to do today.I'm going to not give you a little peek behind the curtain,I'm just going to tear the entire curtain away and let

  • 02:45

    STEVEN MCDERMOTT [continued]: you see what we're doing.I'm going to also show you the reasoning behind why we'redoing what it is that we're doingand I'm going to show you the scale of what we're doing.I'm also going to introduce a little bit of how we'retrying to introduce things like machinelearning and artificial intelligence into it.I'm going to ask, at what level should westop the machine learning, at what levelshould we be stopping the automation?

  • 03:08

    STEVEN MCDERMOTT [continued]: We really need to be concerning ourselves with thingslike ethics and privacy.So a nice little overview.If you want to get in touch with me,the best way is through my Twitter handle @soci.I do respond.My official email address, it's OK.But if I'm traveling on the move,

  • 03:29

    STEVEN MCDERMOTT [continued]: it may be a little difficult to get signed into it.So what I'm looking at, the titleis "The Recognition of Citizens Voice with Social Media."A large part of the responses to contemporary problems,the first kind of mindset is, let's apply the tech to it.The tech will fix it.

  • 03:50

    STEVEN MCDERMOTT [continued]: It's a technological solution to somethingthat probably isn't a technological problem.That's how we'll fix it.OK?A lot of them have that similar mindset.What they're trying to do is address what theycall a crisis of listening.At the moment, civil servants within HMRC

  • 04:10

    STEVEN MCDERMOTT [continued]: are concerned about democracy.They're concerned that they are not in any wayrepresenting the citizenry of Britain, that somehow there'san amount of dislocation between the citizens and politicians,yes, MPs.What we're trying to do is somehow enable the departmentto be a listening organization that can be more inclusive.

  • 04:31

    STEVEN MCDERMOTT [continued]: My big concern with this idea of becoming a listeningorganization is, yes, collect all the information.But you're not really listening unless you're somehowincluding what citizens are saying and responding to it,whether that be affecting policy or even just affectingor fixing a service that the citizens are trying to us.So at the moment, they're very, very strong

  • 04:53

    STEVEN MCDERMOTT [continued]: in gathering the information.Yes?That's not a problem.The problem we're having is there seemsto be a gap or a disjunction between,OK, the citizens are saying things like,I'm seeing Error 404 on the websiteand I'm trying to submit my self-assessment.The problem then is we can pick up.How do we close the loop and go and fix it?

  • 05:13

    STEVEN MCDERMOTT [continued]: Or it could be people are arguing online,why aren't we getting enough tax from Google and Amazon, yes?Citizens are saying that.How do we move that, what the citizens are saying,and get it to the policy advisors--if you like, the executive committees--who then say to the politicians, this is what we're being told.So to me, it's almost-- the listening is fine.

  • 05:35

    STEVEN MCDERMOTT [continued]: We can gather the data, but we need to close the loop, closethe gap.Internally, they have acknowledgedthat maybe all of the solution isn't technological.They're also starting to argue maybe weshould start looking at the culture of HMRC more widely.As I said at the beginning, my background is largely academic.I'm not a born and bred civil servant.

  • 05:57

    STEVEN MCDERMOTT [continued]: None of my family have ever been civil servants.I have no idea what it even means to be a civil servant.Yes, I've read all the statements I have to do,online learning.I ticked all the boxes and I read the core values.And I understand that.I understand my commitment to the citizensis to be honest and upstanding guy, honest and true. yes,I get all that.But on a day to day basis of the rules and regulations,

  • 06:19

    STEVEN MCDERMOTT [continued]: I don't have a clue, literally.And the team that I'm a member of, there's like 50 of usand we're all sort of ex-academic, largely.We don't know the rules about the hierarchy.They said to me, Steven, you're a grade 7.I'm like, I don't even know what that means.What does that mean?Well, it means you're lower than a grade 6

  • 06:39

    STEVEN MCDERMOTT [continued]: but you're higher than a-- what's that even mean?And then you have people who are SOs and HOs.This means nothing to me.All I need is people to work with me, yes?So mindset is very, very different.HMRC are recognized by bringing in academicsand reaching out to universities andother essential researchers.And so what they're saying here is, we need to help.

  • 06:60

    STEVEN MCDERMOTT [continued]: The current structure is too top-downto implement the cultural changes needed to overcomethis crisis of listening.So they brought in people like me, academics.I didn't understand the--what do you mean, I can't be doing X?I've just done it.

  • 07:20

    STEVEN MCDERMOTT [continued]: Is this a problem?Yes, Steven.This is a problem.We had not known this problem existed.So we were kind of brought in slightly as disruptorsa little bit.Yes, we're contained.I have one of those stride machineswhich are locked onto the network,and then they've given us these toys off stride machinesto go and play with.But the two must never connect.

  • 07:43

    STEVEN MCDERMOTT [continued]: So we can experiment on our off stride machines,off network machines, but do not mix the streams.Yes?So what they've implemented now and we're reallytrying to push through within HMRC to all of HMRCis this idea of some sort of real-time listening, which

  • 08:03

    STEVEN MCDERMOTT [continued]: is a monitoring response to social interactionsbetween HMRC representative, citizens, customers, and HMRCservices.And of course, I was going to showyou a real live gamble of it, but of course it doesn't work.Of course, tech never works.What we have is real-time updates of monitoring

  • 08:25

    STEVEN MCDERMOTT [continued]: and displaying visualizing customer interaction with HMRCadvisors, like HMRC customers, HMRC gov, HMRC digital.So that's in real time.We've also got slight, very small level of network analysisof that happening in real time.Any politicians or journalists who are tweeting about HMRC,

  • 08:47

    STEVEN MCDERMOTT [continued]: it appears on those screens.It appears in the press office for all to see.So it's real-time listening.But for me, that's fine.It's immediate.It allows me to respond quickly.It's almost preemptive at times.It also allows them to spot problemsthat people who are using the service

  • 09:08

    STEVEN MCDERMOTT [continued]: have spotted before the IT system has worked outwhat's going give you an idea, if anyone everfollowed HMRC, the issue of concentrics last year.The person whose job I took over came into work onemorning, switched on his computer,

  • 09:29

    STEVEN MCDERMOTT [continued]: and the system he's using to monitor this interactionhad crashed.And he went, that's not--what's going on?I can't even get it to start up.What had happened was the concentric issuehad hit the public first and was basically taking off allegedly.It exponentially just-- and the systemthat it was using to monitor couldn't handlethe level of interaction.

  • 09:49

    STEVEN MCDERMOTT [continued]: So he was able to give a few minute warning to the pressoffice, get ready, the proverbial has hit the fan.Yes?So this is sort of real-time listening and monitoring.So just to give you an idea of what my day is like,I spend all day every day reading Twitter, Facebook,

  • 10:11

    STEVEN MCDERMOTT [continued]: looking at Instagram, following whatpeople are saying about HMRC.I've taken what I did in the Singaporean contextof the ethnographer and I'm embedding myselfin HMRC's online social media presence.I don't interact with anyone.If you look at my bio, it will say HMRC, yes?But that's it.

  • 10:31

    STEVEN MCDERMOTT [continued]: That's all it is.No one comes to me with HMRC problems.I don't know the answers to them.I'm only there five months.But what I'm doing is I'm really tryingto listen to the customers and then feedthat back to everybody else as bestI can within a body of 70,000 staff members.

  • 10:52

    STEVEN MCDERMOTT [continued]: OK?So as well as doing that, what I'm also trying to dois I'm trying to keep up with trends that are happeningand how customers are moving from Facebook to Twitterand people are moving off Twitter to thingslike YouTube and Instagram.I'm trying to keep myself current and up to dateon what I'm doing.I've also got some sort of underlying strategy mentally

  • 11:14

    STEVEN MCDERMOTT [continued]: in my head of what I'm trying to achieve.I'm trying to remove the companies like social mediaplatforms like Brandwatch and Crimson and Pulsar.I'm trying to remove them, I'm trying to remove Twitter,and I'm trying to get directly at the customer.It's sort of a long-term strategy.

  • 11:38

    STEVEN MCDERMOTT [continued]: I personally have a sort of goals-oriented strategy.For me, I'm trying to work out whatis the future of social media analytics,where are we heading?Because of my own background, my kindof ethnographic understanding, for me context is king.I'm going to show you a lot of computer-generated machine

  • 12:01

    STEVEN MCDERMOTT [continued]: learning images and visualizations,quantitative aspects of social media analytics.And to me, they're fine up to a point.They do the job of something, yes?To me, they're descriptive, they can be done in real me, they're not even 50% of the picture.To me, we need everything else in the qualitative perspective.

  • 12:22

    STEVEN MCDERMOTT [continued]: Part of that is the machines-- thisis a very simple way of summarizing-- the machinesdon't understand.Simple.The machines are very good at syntactic logic and numbercrunching, yes?They understand logic and reason.But what they don't get is semantic meaning.They don't understand what a word means.

  • 12:43

    STEVEN MCDERMOTT [continued]: They can never understand.They don't get ethics.So the quant is fine.They don't understand context.I think-- and I can go in length of these topics, but I won't--we're entering a second big data revolution. for me,the first was the invention of the medical profession and allthe assemblages and buildings and professions that came

  • 13:05

    STEVEN MCDERMOTT [continued]: around as a result of that.I think people are waking up to surveillanceand the end of privacy and the use of smartphones and smartdevices.So people are really concerned about their privacy.They've lost it.Privacy is dead.We're trying to reclaim it and claw it back.We're in the midst now of what theycall fake news, misinformation.

  • 13:27

    STEVEN MCDERMOTT [continued]: For me, I'm kind of a more sort of George Orwell kind of guy.What is it, news is something that someone somewheredoesn't want printed?Everything else is PR.I've lived in that world for, what, 40-odd years,this idea of news being fake.When did we all wake up to that?It's nothing new.

  • 13:50

    STEVEN MCDERMOTT [continued]: I'll come back to this later whenwe're looking at how we're tryingto capture bots and deal with bots in social media.I also like to look at suppose that--I was going to say good.It can't be good--successful strategies and influencing social media

  • 14:10

    STEVEN MCDERMOTT [continued]: are not good.OK?And let me also, and I think I agree with this,there are calls for some sort of Hippocratic oath for datascientists.Within the profession of HMRC and the civil service,I don't at the minute-- because I don't have the levelof clearance--I work with colleagues who have accessto everybody's information--

  • 14:33

    STEVEN MCDERMOTT [continued]: anybody who has a National Insurance number in the UK.And that's a massive level of responsibility.Yes, we have data guardians in place and ethicsand there are rules and regulations and systemsin place to check, yes?But to me, the rules and regulationsare fine in trying to quantify and systematize

  • 14:55

    STEVEN MCDERMOTT [continued]: what the response should be or what the infringement isgoing to be.I mean, you need an oath, sort of a yes,I adhere to certain values and shall do no harm.And finally, as I like to call it,we have the GDPR, or the monster coming over the hill, in May.

  • 15:17

    STEVEN MCDERMOTT [continued]: Now, I've met with the lawyers externally,I've discussed it with academics and technicians internally,I've discussed it with privacy activists in the UK.And in May, there was a deluge of demands for data subjects,as they referred to, to have their data removed,

  • 15:39

    STEVEN MCDERMOTT [continued]: to have none of their data involvedin automated process in making decisions about them stoppedon a mass scale.A lot of organizations seemed to be working on the premisethat it's OK, the British public won't hear about GDPR.We can keep it quiet.

  • 15:60

    STEVEN MCDERMOTT [continued]: And then, we get a little stamp on our website,it says, GDPR-ready.Good.They're in for a monumental shockto the core of everything that's meant for the internet upto this stage.It's going to be a fundamental shift.To me, it's the viper in the sense

  • 16:21

    STEVEN MCDERMOTT [continued]: it's gonna be the end of that not just from governmentorganizations, but also corporationslike Google, Facebook.Yes?A fight back, if you like.And it's only the beginning.To me, that's the context.Where I think we're moving is we seemto be entering a world now where data is becoming currency

  • 16:42

    STEVEN MCDERMOTT [continued]: or data is [INAUDIBLE].Data is the new oil.No, it's not.We seem to be moving into a worldwhere our data is being sought, gathered, exchangeby data brokerage companies.I have a very subtle definition for Twitterand Facebook and Instagram.You list them all, yes?They're data brokerage firms, each and every one of them.

  • 17:05

    STEVEN MCDERMOTT [continued]: And what they do is they gather our data,they gather everything about us-- about you,about our relationships and interactions-- they tell us,don't worry about it, it ain't worth anything.Yeah, right.And then they repackage it, sell it to marketing and advertisingcompanies through middle people like Axion and other companies.

  • 17:26

    STEVEN MCDERMOTT [continued]: And then that's how advertising is targeted towards us.So I think we're entering into a new world that'ssome sort of surveillance capitalism.But I think GDPR is a pushback against that,or the beginning of a pushback against it.

  • 17:48

    STEVEN MCDERMOTT [continued]: Now I'm going to move towards the methodsif you like that setup, what I thinkis the problem of all these social media platformsand social media analytics more generally.My worry is that because it's a very algorithm,automated, quantitative centric approach to social media

  • 18:11

    STEVEN MCDERMOTT [continued]: is what we are generating are calculatedpeople or calculated data people and calculated publics.My worry is that not just media analysts,but policymakers and decision-makerswill start and take these dash boardsas some sort of truth or fact.You can apply the same critique to quantitative methods

  • 18:35

    STEVEN MCDERMOTT [continued]: in general, yes?For me these data points, these truths, these facts,these visualizations are constructions,social constructions.So what I'm trying to head towards is,yes, I'll use these platforms, but I do soin a very critical way.I just don't trust what they're telling me.

  • 18:58

    STEVEN MCDERMOTT [continued]: It gives me a nice descriptive pattern.I mean, it's the beginning of the analysis.It's not the conclusion.It's not here we go, paper published.Well done.It's the quant, the visualizationI'm going to show you of where I start my analysis.All they do is it's almost like a sampling.

  • 19:18

    STEVEN MCDERMOTT [continued]: I think of like a sampling technique.Where does it think I should start, why does itthink I should start here?The worry is that our social mediais in various ways becoming how we are defined, managed,and governed.

  • 19:39

    STEVEN MCDERMOTT [continued]: For me, I find it denaturalizing and unfamiliar,but not objective, unbiased, or mutual.So here's a sample of analytic dashboards.And there are myriad companies set up in Londonand all over the UK who will quite happily generatethese dashboards for you on a monthly basisfor about 5,000 pounds a pop.

  • 20:02

    STEVEN MCDERMOTT [continued]: Every month.And depending on how many Twitter users or Facebookmentions you get, then they increase the price.And this goes on and on.I can set these dashboards up using freely available softwarein 10 minutes.For me, a lot of these are a fiction, they're a construct.

  • 20:28

    STEVEN MCDERMOTT [continued]: So now what I'm going to take you throughis HMRC's dashboards.I said I'm not going to peek behind the curtain,I'm going to just tear the curtain off and show you.What we use is a company called Brandwatch, which is probablythe leading company in terms of social media and analytics.

  • 20:48

    STEVEN MCDERMOTT [continued]: There are very many out there and many just as goodor better than them, OK, offering a myriadof different visualizations.What I'm trying to get at is while we fixate ourselveson this drive for real-time analytics,and yes they're moving forward, is this idea

  • 21:09

    STEVEN MCDERMOTT [continued]: that surely we need to also maintainat least a historical idea of how our brand image, howour interactions with our citizens, with our customersis changing on a day-to-day basis.A lot of these visualizations give you whatis known as a static picture.And you think, oh, our interaction with the publicis this static and fixed thing.

  • 21:29

    STEVEN MCDERMOTT [continued]: For me, it's moving and evolving.So it's like you've got a living organism.It's like a micro ecology, if you like,and it evolves on a minute-by-minute basis.The worry is with the quantitative aspect of it,results in various fixed, static visualizations.And then the policymaker or interpreter thinks, a-ha,

  • 21:51

    STEVEN MCDERMOTT [continued]: we know the public now.Let's do A. And before they've got information B,the public has changed, the citizenry have changed.So if you like, this is the interactionsHMRC has had for last year, from the 1st of January, 2017,to December 31st.Total mentions, 3 million.

  • 22:13

    STEVEN MCDERMOTT [continued]: Unique authors, 438,000.And surprise, surprise, the dominant theme is--who would have guessed it--Brexit.Followed by Theresa, Labour, Britain, Tory,SMAC (Single Market and Custom).It has dominated my life for the last six months.And by my life, I mean every waking moment of my life

  • 22:33

    STEVEN MCDERMOTT [continued]: is Brexit.I have tried using filters on Brandwatch to get it off.I've gone in and set categories and rules in Brandwatch to say,just don't show it.I know it's there, but don't show me it.And I can't get the thing off.So I thought, OK.Leave it.Leave it.Just deal with it.

  • 22:53

    STEVEN MCDERMOTT [continued]: Look at it.So what do we do?What I may end up doing is I'm looking up, in this one,over that full year again, yes, what were the peak topics,what really excited the British public in relation to HMRC?When I started this, I thought, the first one'sgot to be football.Has everyone heard about football and Rangers

  • 23:14

    STEVEN MCDERMOTT [continued]: and HMRC and the tax?I thought, that's got to be because that's so, they are just favorite football teams.People who were there if you liketweeting not at HMRC Customer Services, but just with #HMRCare any other derivation of it.Quite small.while British public, the Customs Unit,quite small as well.

  • 23:35

    STEVEN MCDERMOTT [continued]: The prime minister's speech in Florence,which I think I had 45 minutes to get readyfor; this, the Paradise papers; the autumn budgets;and since then we're back at Customs Union with Brexitand have been ever since.

  • 23:56

    STEVEN MCDERMOTT [continued]: So this is kind of telling us the volumes.In terms of the HMRC services--you can't see it, it's nice and small--tax avoidance, self-assessment, tax credits,corporate tax evasion.OK?Those are the terms.What we're able to do is track over time.

  • 24:16

    STEVEN MCDERMOTT [continued]: The top on here is weekly, the bottom one is daily.So you see little peaks.And what happens usually is we'reable to track what our customers are becoming interestedin as the deadlines for submissions approach,obviously.Yes?And the biggest one every year isself-assessment at the end of January.

  • 24:36

    STEVEN MCDERMOTT [continued]: And it goes through the roof.This is from January of 2017.Yes?So it's what the customers were getting in touch with usand discussing.We can also drill right down into all the mentions,if anyone mentions HMRC, and we get a lot of these.Most of them are not very flattering of HMRC,as you would expect.

  • 24:58

    STEVEN MCDERMOTT [continued]: You can't really look at social media analytics and be--if you don't like people swearing at youon a regular basis, keep away from social media analytics,especially if you're working for HMRC.You get a lot of it, quite a lot.So a lot of times, I have this problem.Someone will say to me, Steven, can you send through for X Com

  • 25:19

    STEVEN MCDERMOTT [continued]: the tweets and mentions that have come through todayabout a particular topic?I spend quite a lot of time going,should I delete that one?Should I-- I reckon I should delete that one.Do we want the minister or the chancellor to see that?No, we don't want it.Right.Then it's not so much censorship, it's

  • 25:40

    STEVEN MCDERMOTT [continued]: that I worry about what people are saying.So we can go from that sort of macro levelright down into what people are actually saying.This isn't just Twitter; this is everything.This is Twitter, Facebook, Reddit,all local and national newspapers, radio broadcasts,BBC.If it's published in the UK and it's about HMRC,

  • 26:00

    STEVEN MCDERMOTT [continued]: it will appear in this feed.In terms of who for the last year discussedand where they're geographically located, it was [INAUDIBLE]..You've got a scale here.The darker blue is between 18,000 and 2 millionhigher end.So people who discuss HMRC are not simply located in the UK,

  • 26:23

    STEVEN MCDERMOTT [continued]: although we tend to focus our own analysisand real-time analytics on just UK-based accounts.But it is a global brand.A list of all the different companies that we're following.Twitter; forums; Reddit; TripAdvisor; Instagram;and one of my favorites, the Daily Mail;

  • 26:44

    STEVEN MCDERMOTT [continued]: CareerBoard; Express; Reed; and YouTube.We can always count on the Daily Mail to be negative.There's a story come out of the Daily Mail?OK.All right.This will be negative as usual.We're now moving into the realm of people posting pictures.

  • 27:05

    STEVEN MCDERMOTT [continued]: So this is also within Brandwatch,but they're slowly rolling this out.And what it is, is people posting pictures on Twitter,this one anyway is Twitter-focused,which contains the logo of HMRC--the logo.And what we're looking for is people using itfor phishing scams or people pretendingto be a representative from HMRC and trying to scam people out

  • 27:27

    STEVEN MCDERMOTT [continued]: of their money.So we're also able to forward this onto cybersecurity teams.We look at Facebook.What I've noticed with Facebook engagement,it is dipping unless we pay for it.If we pay for it, it takes off.But once we withdraw the money, it dies back down again.

  • 27:52

    STEVEN MCDERMOTT [continued]: That's the Facebook engagement.We're also looking at LinkedIn.We've got quite a global audiencefollowing HMRC on LinkedIn.Most of it tends to be around our recruitment issues,job offers.This is the one I really like within Brandwatch,it will slide off a bit.

  • 28:12

    STEVEN MCDERMOTT [continued]: This is the one I spend all day looking at.This is where I start.This one's quite good.This one is fading topics.Trending topics, topics that are starting to take off.This was taken a few weeks ago.Of course, it's [INAUDIBLE].

  • 28:33

    STEVEN MCDERMOTT [continued]: So I spend all day refreshing this every every 15 minutes,just watching to see what's coming up.So we've got leaving the Customs Union, Trans-PacificPartnership, Good Friday Agreement, foreign minister.The other day I had had one up here.It appeared here and I thought--if I say the name, you're going to know who I'm talking about.Martin Lewis.

  • 28:54

    STEVEN MCDERMOTT [continued]: Heard of Martin Lewis?Martin Lewis appeared here and I went, clicked on it.He tweeted, I'm appearing on This Morningto discuss tax-free childcare.I went, alarm, alarm.On to the Press Office.Did you know?It just came back saying, thanks.

  • 29:14

    STEVEN MCDERMOTT [continued]: You did know as if they did.So this is like a real-time what are the main trending topics.But I want to show you at this point is a lot of thisis still like a black box.Yes?I will see the background working as a whole.I once met Brandwatch technicians here in London,and I asked one of their display guys, a technician, I said,

  • 29:38

    STEVEN MCDERMOTT [continued]: what algorithm are you using to cluster these?And he went, sorry, what?What algorithm are you using to cluster these topics?And when we're doing text mining, whichone are you using it?MCMC.That must be bankrolled.

  • 30:00

    STEVEN MCDERMOTT [continued]: It's a black box, yes?And that to me is a concern.For me, I treat them from critical perspectiveas these are fictions, these are social constructs.The modeling algorithms knows nothingabout letters, words, sentences; nothing about narrative form;nothing about meaning; nothing about politics.

  • 30:22

    STEVEN MCDERMOTT [continued]: All it sees are the letters.And that's where the human being comes in.I think that's where the human being is nevergoing to be replaced.Yes, there are a lot of promises about AIand artificial intelligence and machine learning.But I'm not overly worried about itat the minute because of that gap

  • 30:42

    STEVEN MCDERMOTT [continued]: between syntactic and semantic meaning.As I said, they lack understanding.What I'm concerned with on a more political levelis what's happening is the datafication of people.Once we start treating these opinions as if they'resort of disengaged from the speaker,

  • 31:04

    STEVEN MCDERMOTT [continued]: are we really listening or are wejust datafying it, churning it into numbersso we don't have to really comprehend and listen?I mean, I've been struggling with this problem for a while.How do you prove that you're listening?I get a lot of blank stares, so nobody's listening.Well, how do you prove as an organization

  • 31:25

    STEVEN MCDERMOTT [continued]: that you're listening?Prove, show evidence that you're listening.That's a concern when I think of if we'removing towards the datafication, the quantified,I think maybe we're not really listening.

  • 31:48

    STEVEN MCDERMOTT [continued]: So a lot of decisions are being built on these datafied people,these datafied social interactionsand social relationships.For me, they do not accurately or adequatelyrepresent how people engage with and experience the social mediaor our services.My worry is that planners, strategists, designers,

  • 32:08

    STEVEN MCDERMOTT [continued]: developers, and policymakers take social media dataas an object or representation of a truth thatcan be extracted from and reflects their audienceor customers.But they are not mutual and they are not objective.A lot of what I'm seeing in big organizations--and not just where I work now, but all the time-- they'refixated on customer satisfaction scores, on the number.

  • 32:30

    STEVEN MCDERMOTT [continued]: Are we close to our target or not?Oh, no.It's 78.9 And this goes on every day in my life.What do you mean, it's 78.9?Have you cross-checked this?Steven, why did it go from 79 to 78.9?I say, well, the customers say they're nothappy with the sign-in function.

  • 32:51

    STEVEN MCDERMOTT [continued]: Yeah, but why aren't they?Why, Steven?Because they're not happy with the sign-in feature,we need to change it.Yes, I hear you.But why?Oh.Oh.That's a fixation with numbers, yes?I'm not anti-numbers, I'm just saying they tell us maybe

  • 33:12

    STEVEN MCDERMOTT [continued]: less than 50% of the story.They're not neutral.They have their place.For me, we need to be doing something different.For me, it's about being criticalof social media research platforms and dashboards.Certain demographics are not on Twitter,they're not on Facebook, they're not on Instagram.

  • 33:33

    STEVEN MCDERMOTT [continued]: They're not representative.That's fine.There's a place, social, ethical,and political concerns at the coreof what we're trying to do.For me, for a while I've argued for ethics as a method,constantly trying to get my colleagues-- though it's notworking at all-- to stop talking about dataand instead refer to people or, heaven forbid, citizens.

  • 33:56

    STEVEN MCDERMOTT [continued]: Just change the language.You want to change culture, you changethe language people are using.So when people say, data, I go, people.Sometimes I say, citizenry.And they look at me.but it's about looking and being concernedwith the language we're using whenwe're referring to what people are doing on social media.If we keep talking about it in terms of data,

  • 34:16

    STEVEN MCDERMOTT [continued]: we're going on with datafication.If we keep referring to them as people, citizens, opinions,understandings, the why, we open it up.We make it truly listening being able to respond.For me, I try to do something different.

  • 34:37

    STEVEN MCDERMOTT [continued]: Instead of relying solely on Brandwatch,which I do on a day-to-day basis,because of the nature of my role,I tend do use something called research.And what I do is I have my own servers running collectingdata, data that I can unpack, that I can open up and start

  • 34:59

    STEVEN MCDERMOTT [continued]: reading and checking, what am I sending to Facebook's API?What am I really requesting?Dos Facebook offer something elsethat I can also get back and add that to my list on my API?For me, it's knowing them quantitatively andqualitatively, knowing them as a mass, as a group,and as individuals.

  • 35:19

    STEVEN MCDERMOTT [continued]: We see them as groups and therefore subject to groupdynamics at times.Just for those-- most people are probablyaware of this paintball accident that is this.Social network rough.So what I'm doing here is people who arereferring to HMRC customers.Each little tiny, tiny dot is a Twitter account

  • 35:40

    STEVEN MCDERMOTT [continued]: in this instance.And then they're grouped accordingto a particular algorithm called modularity class,which I can check against MCMC.Then I monitor and watch it over time as it evolves.

  • 36:01

    STEVEN MCDERMOTT [continued]: For me, what I'm trying to really applyis a qualitative and ethnographic approachto social media analytics.So at the moment, I'm not engagingwith people who are talking about HMRC because I thinkthat might freak them out.So a lot of people, if they were to refer to HMRC CustomerServices, then we respond, yes?

  • 36:23

    STEVEN MCDERMOTT [continued]: But if they just said, HMRC, I don't like you very much,I don't respond.Because I'm worried I might freak them out.So if someone says, I don't like you very much, HMRC,you took all my money and the road outsidedoesn't work, I don't jump in and go, hi.Where do you live?We'll come fix the road.

  • 36:44

    STEVEN MCDERMOTT [continued]: Because they're going to go, surveillance,you're killing my privacy.Yes?How dare you listen to my conversations.Well, how can we be a listening organizationif we don't listen?So we're trying to work out, when should I engage?So overall I'm a ethnographer, but very muchobserver position.And what we're trying to in turn work on is, just how far can I

  • 37:08

    STEVEN MCDERMOTT [continued]: go and say, actually, we're trying to deal with your issueso others don't have the same problem.For me, in terms of the strategy, whatI'm trying to overcome is with this move to surveillancecapitalism, if you like, is we have our customer;

  • 37:29

    STEVEN MCDERMOTT [continued]: we have us listening to them; and we have Brandwatch;and then we have Twitter, Facebook, and all the myriadof different organizations trying to make money outof datafying our customers.And what I'm trying to do is remove Brandwatch,remove Twitter, remove Instagram,and go directly to it.At the moment, HMRC has an app.

  • 37:51

    STEVEN MCDERMOTT [continued]: We do.It's a little bit interactive.You can do certain things on.But then that's used to keep rolling that out and improvingon it.So I'm coming close to the end.So my approach is--this is my title, if you like, this is my goal--

  • 38:14

    STEVEN MCDERMOTT [continued]: qualitative analysis and social media is my job title.I leave in a world of quantitative peoplewho love numbers, or statisticians.And I had the good fortune of meetinganother group of social researchers just recently.It was like seeing a lifeboat appearing on the horizon.

  • 38:36

    STEVEN MCDERMOTT [continued]: What I'm trying to do, what we're trying to do,I and my team, is listening to what people say to usand what they say about us.We're trying to present qualitative analysisto drive data-rich decision-making andevidence-based strategy to improve services.So I've been socially intelligent,listen to what they say, but also

  • 38:57

    STEVEN MCDERMOTT [continued]: using computer-aided qualitative dataanalysis software, which I'm going to show you what I use.Focused on understanding insights, meanings, and valuesand really trying to get the why of human behavior.So I spend a lot of time reading certain people on Twitterwho have real solid grievances about HMRC or about government

  • 39:21

    STEVEN MCDERMOTT [continued]: policies.And I'm reading them and listening to them.And I'm trying to work at the moment--I'll show you something in minute--I'm trying to work out, why is this individual tweetingor posting on Facebook 150 times a day,from 7:30 AM in the morning till 8:00 PM that night every day?

  • 39:44

    STEVEN MCDERMOTT [continued]: Every day?When I first came across them, I thought, that's a bot.It looks like a bit, it acts like a bot.I go and look.No, that's a human being.Every day.I'm thinking, so why are they doing that?What's their motivation behind it?The tools that I use are Brandwatch and Vizia.

  • 40:09

    STEVEN MCDERMOTT [continued]: I do a little bit with Survey Monkey, not so much.Natural language processing with Tropes--T-R-O-P-E-S. And the networks, the graphs I create are fromGephi.To get them up to scale bigger, up into the millions, UCINEtand Pajek.A lot of these are not free, not free, not free,

  • 40:30

    STEVEN MCDERMOTT [continued]: but give you free access for a bit.Free, free, free, free, free, free for a bit, free, free.So I use a lot of freely available open source softwareto do it.What I'm trying to do is find small patterns in big data.

  • 40:56

    STEVEN MCDERMOTT [continued]: This is using Gephi.This is an analysis of civil servants in the UKwho have Twitter handles, looking at people within HMRCare in blue, gdstream, UK civil servantsin more general, number 10 Downing Street.To me, this is looking at sort of backstream of communicationbetween civil servants, an informal way or pathway

  • 41:19

    STEVEN MCDERMOTT [continued]: of communication that's going on.I also analyzed unstructured customer servicesfeedback using Tropes.Not to conclude at this, this is getting of my analysis.They say it's easy?My first few weeks, they gave me customer feedback unstructured.I analyzed it.

  • 41:40

    STEVEN MCDERMOTT [continued]: Apparently, the identification and authorization processis easy.Now, I just did it last week and it wasthe biggest pain in my butt.I was like, aargh.Before you begin, get these five things.One hour later, I've got four.So I thought, how can they be saying it's easy?So before quantifying data sets higher-ups, it's easy now.

  • 42:07

    STEVEN MCDERMOTT [continued]: You can relax.I thought, no, no.Let's dig deeper.So wherever they say it's easy, I'm going right into the dataand just simply reading down through.What they were saying was, it's easyonce you get past that silly identification authenticationprocess.So what have they done at HMRC?They're now fixing or trying to fix the identification process.

  • 42:31

    STEVEN MCDERMOTT [continued]: The scale of what I'm dealing with, thisis for June to November, 2017.I'm trying to deal with 1,500,000 comments coming in.Previously, some little teams of themstill try and manually tag and read the comments.The group I now work with, the Digital Data Academy,

  • 42:53

    STEVEN MCDERMOTT [continued]: we have a team of at least six peopledoing this on a daily basis.You're thinking, that's torture.Yes, every day.And manually tagging all these comments as they're coming in.I said, stop.This is not fair.This is inhumane.stop doing it.

  • 43:15

    STEVEN MCDERMOTT [continued]: So what we're trying to deal with is,how can we analyze 1 million plus comments,70 million words, and come up with something other than easy?So what I'm working with at the minute in HMRC DDA,I'm working with the University of Bradford,the University of Manchester and their text mining teams.We're trying to work on a better machine learning, HMRC-centric.

  • 43:38

    STEVEN MCDERMOTT [continued]: Let's get past sentiment analysis,let's get past whether HMRC customers are happy or sadbecause most of them are not very happy with usbecause they gave us money, and let's get out somethingcalled levels of certainty and uncertainty.What does our customers know, what do they not know,what are they trying to know?

  • 44:01

    STEVEN MCDERMOTT [continued]: And finally, we're also doing network analysis.This is part of the real-time dissemination of social mediaanalysis.This is part of the Vizia display, also part of Vizia,a way of categorizing the comments that are coming in.My current move is what we're finding

  • 44:21

    STEVEN MCDERMOTT [continued]: is Twitter and Facebook find for certain type of demographicsand others are moving on to things like Instagram, Reddit,YouTube is very popular in terms of the videosand the number of views the get.But the comment section is switched off,which I want to be switched on.The idea is to go to the customer on the platform

  • 44:42

    STEVEN MCDERMOTT [continued]: of their choice, not go to YouTube and or Twitter and say,I hear you're having a problem.Would you like to come over here to the phone,but going to the platform of choice the customer has.Just quickly, I'll finish on this.What we're starting to do with the University of Bradfordand a small team there is we're trying to take peopleposting images tagged with or located with or near the HMRC

  • 45:08

    STEVEN MCDERMOTT [continued]: logo, looking for phishing scamming sites.So as things like SA or self-assessment approaches,scamming and phishing attempts increase as well,and people are vulnerable to panic.Customers will post them on Instagram saying,here's a picture of that I got claimingto be from HMRC or on an email.

  • 45:29

    STEVEN MCDERMOTT [continued]: Is it?And they're responding to it and following it up.Or people claiming to be authorized HMRC representativeswhen they're not.I'm going to finish.Let's finished.A big move, as well, we're looking for internally

  • 45:49

    STEVEN MCDERMOTT [continued]: on our team is looking for bot detectionand trying to deal with misinformation and fake news.So far what I've been doing, or did for a few months, isI've been manually going through the lists of Twitter accountswhich are regularly every little dot is a post about HMRC.

  • 46:10

    STEVEN MCDERMOTT [continued]: The more frequent, the more systematicyou see, it's an indication that it might be a machine.So in that system, I'm likely going downthrough each and every one of them,checking is it a bot or not?Then if I think it is, if I think it's a bot,if I think it's a human being, I walk away and leave it.And finally, what we're trying to achieve

  • 46:36

    STEVEN MCDERMOTT [continued]: is some way of becoming a listening organization.But I think what we really need to includeis becoming socially intelligent but looking for understanding,insights, meanings, and values and trying to getat the why of human behavior.If you need to contact me, the best wayis either @soci or my gmail account.

  • 46:57

    STEVEN MCDERMOTT [continued]: And I'm more than happy to help wherever I can.Thank you very much for listening.[APPLAUSE]


Dr. Steven McDermott, PhD, discusses his social media research at Her Majesty's Revenue and Customs (HMRC) on the use of social media, machine learning, and considerations of ethics and privacy.

Video Info

Publication Info

SAGE Publications Ltd
Publication Year:
SAGE Research Methods Video: Data Science, Big Data Analytics, and Digital Methods
Publication Place:
London, United Kingdom
SAGE Original Production Type:
SAGE Lectures
Copyright Statement:
(c) SAGE Publications Ltd., 2019


Steven McDermott

Segment Info


Segment Num: 1


Segment Start Time:

Segment End Time:


Things Discussed

Organizations Discussed:

Events Discussed:

Places Discussed:

Persons Discussed:

Methods Map

Sentiment analysis

Sentiment analysis is an application of natural language processing that focuses on automatically identifying emotions or sentiments expressed in text data.
Sentiment analysis
Recognition of Citizens' Voice with Social Media

Dr. Steven McDermott, PhD, discusses his social media research at Her Majesty's Revenue and Customs (HMRC) on the use of social media, machine learning, and considerations of ethics and privacy.

Copy and paste the following HTML into your website