July 28, 2024 • 39 mins
The focus of computer technology historically has been on the manipulation and communication of data and information. Yes, there’s always been the monstrously obvious admonition of “garbage in, garbage out” when speaking of data. But as our dependence on data grows, the issues of data quality, of making data better, have grown in importance and complexity. Data, it turns out, is endlessly nuanced.
Government Data Generation and Usage
Government has an enormous interest in data. It is an issuer of data when it assigns account numbers, for example, to its citizens to ease service delivery. It is also a considerable consumer of data in order to establish policy, measure program efficiency, support planning, and, just as with any business or individual, for decision making of all kinds.
But this isn’t simple. The term “government” masks the fact that multiple agencies exist, each with its own goals, never mind data handling policies and procedures. Sharing data across industries is as nuanced as data sharing between enterprises or even more so.
Understanding how governments think about the data they consume and generate is key to long term data security and online identity.
Talking with Data Expert Ian Oppermann
In this fascinating and stimulating conversation, Steve and George discuss these topics with Ian Oppermann, the former data director for the state of New South Wales, a director for Standards Australia, and advisor to multiple startups.
Ian shares his insider’s knowledge of government agency priorities and the fact that sharing data across agencies is “extraordinarily hard.”
Just at the Beginning
Standards Really Really Matter
Ian’s participation in ISO standards development comes from his insight that data sharing requires very crisp definitions, detailed use cases, and specific guidance for each use case based on privacy and data custodianship requirements. And he points out that we are just at the beginning.
For example, the latest ISO standards tackle the basics of terminology definition and use cases, ISO 5207, and guidance of data usage, ISO 5212.
These standards do address the use case of AI but even at this stage the standards address the basics.
People Matter
As with many technology management concerns these days, the concerns are rarely about the tech itself. They’re about people, too. Here’s Ian:
“If you want to use [data] for important purposes, you actually need people who know and understand what data is, who know and understand what data governance is, and who know and understand how to actually use the data for appropriate purposes and then put guidance restrictions or prohibitions around the data products you create.”
Ian concludes with:
“But [for] the general use of data, we're only just beginning to understand the power, the complexity, the mercurial nature of data and starting to build frameworks around it.”
Take a listen if you care about data management and governance in large organizations. We are just at the beginning of getting this right.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:10):
Welcome to Making Data Better, a podcast about
data quality and the impact ithas on how we protect, manage
and use the digital datacritical to our lives.
I'm George Higley, partner atBlackstuff Consulting, and I'm
delighted to join my partner,steve Wilson.
Great to see you again.
Let's get right into it.
(00:30):
We've got a terrific guesttoday, ian Opperman.
Among his many accomplishmentsand we'll ask Ian to you have to
brag about them yourself he'son Wealth to Check a couple of
years he's advisor to a legaltech startup.
He was struck by your role as aprofessor at the University of
Technology in Sydney, mostrecently as chief data scientist
(00:52):
for the state of New SouthWales.
Nsw has been a leader in dataanalytics and open data.
So, ian, we're really delightedto have you here and look
forward to hearing your thinkingand a bit of the story of New
South Wales and the DigitalDriver's License and a lot of
other topics, all right.
So, Ian to you, love to hearabout what you're doing today.
Speaker 2 (01:11):
George, great to be here and great to be talking
about data quality.
As I mentioned during theintroduction, I was the Chief
Data Scientist for New SouthWales government for eight and a
half years, an experimentalrole which was really about
helping government think aboutdata differently and see data
differently.
Now left and I've joined astartup.
Of course, the thing you doafter being in government is
join a startup and that startupis all about helping governments
(01:33):
do government differently and,of course, using data, ai and
digital tools to do thingsdifferently.
Joined with a former minister,the former CEO of what's called
Service NSW and the formerminister's chief of staff, and
we're out there doing goodthings with data, digital and AI
to help other governmentsaround the world.
Speaker 3 (01:54):
Fantastic, and I myself had the privilege of
consulting to Service NSW on andoff, especially through COVID,
and seeing firsthand you knowthe agility of that organisation
.
It recently won an awardactually for service innovation,
which was just such a greatresult and a real testimony to
(02:15):
the quality of the people there.
The innovation included thingslike the Data Analytics Centre
which Vian I think you had atitle role of some sort there,
but maybe you could tell ourlisteners about the importance
of data and how New South Walesreally sort of walked the talk
in open data and open government.
Speaker 2 (02:33):
So I joined New South Wales government in 2015 as the
first ever chief data scientistand was given not only the gun
and the badge, but also a pieceof legislation that said you had
to share data with what'scalled the Data Analytics Centre
, and the centre was all abouttackling wicked policy
challenges or problems that arecomplex, subtle and ultimately
have people's behaviour at heart.
(02:53):
We adopted the philosophy thatif we're going to tackle a
wicked problem, we need todeliberately re-complicate that
problem and try to address asmuch of it as possible, and that
, I have to tell you, was reallyan unusual philosophy to take
in the government, whichtypically is about keeping it
(03:14):
simple.
That was helping governmentthink differently about how to
use data to understand problemsdifferently, use data sets a
myriad of different data sets tosee the problem differently.
Every data set is incomplete.
Every data set is less than100% quality.
Every data set doesn't give youfull coverage, but it's a point
of life, and the wholephilosophy was let's see how
many different ways we canunderstand that problem and then
(03:35):
use analytics and AI to reallymake sense of it and do things
differently.
The flow I was taking to ServiceNSW was now we better
understand the problem.
Let's deliver differently, andmy colleague, david Rees, was
really the champion of ServiceNSW.
He took the philosophy thatgovernment can, government can
do this, government can be agile, government can experiment,
(03:57):
government can try, test andlearn and build out services in
an agile way.
We had all been workingtogether for some years when
COVID hit.
And when COVID hit, we wereready to respond and that was
the really big differencebetween before and after that
beginning of thatexperimentation in 2015.
When COVID hit, we couldrespond with data-driven tools
(04:21):
and data-driven elements ofidentity and data-driven ways of
delivering services, becausewe'd spent years experimenting
up to that point.
Speaker 3 (04:30):
And of course there's a really good social contract,
I think in Australia at least,for the use of data, the
reasonable use of data.
You know our COVID response inNew South Wales is very strong
through New South Wales Health.
The public more or lessembraced the idea that contact
tracing was going to be a thingand you know we had such a rash
of private sector contacttracing cloud solutions.
(04:52):
It was a mess and I think itwas Victor Dominello himself who
said we should make this easyand uniform for citizens.
Speaker 1 (05:01):
But I also want to like tag.
Speaker 3 (05:02):
The point you just made about complicating things,
I think is genius.
I'm a mathematician andscientist by training and I
think about generalizing and inengineering we try to simplify,
look for simplifying assumptions.
In science, sometimes we'relooking for that grand general
theory.
Usually generalization makesthings more complicated, but in
(05:26):
our area of digital identity, wethink that generalising the
digital identity problem to adata problem surprisingly makes
things simpler.
That convergence I'm going tocome back to that later, the
convergence of governance ofdata and ID.
But you know, for now can we gobig and talk about data sharing
(05:46):
, because I think that that'sone of the defining themes of
the digital economy.
We're doing open banking inAustralia.
We've got open data.
How do you think about datasharing generally and what are
the societal parameters that youthink about?
Speaker 2 (06:03):
So, steve, you've addressed the entire Pandora's
box with your set of commentsthere.
Let me try and unpack thesub-elements of it.
The first was around sociallicence, and it's an interesting
position to be in governmentand be talking about social
licence.
Social licence is typicallyassociated with mining companies
and such things.
You don't get a choice aboutengaging with government.
(06:24):
When I started in 2015, therewas this real concern that if
government was doing dataanalytics, if government was
doing analysis of data at a sortof at scale, that there would
be this concern from the publicand the perception of public
perception really inhibited awhole lot of activities.
It wasn't the public saying no,we don't want you to do this,
(06:44):
it was inside governmentpeople's thoughts about what the
public might say.
That was really a limitingfactor.
So we went out there and testedit.
We got the data and the datawas, quite simply, we don't want
to answer the same questionbecause we want to deal with one
government, we don't want todeal with 15 different parts of
government.
So we actually spent some timeusing the data to understand how
people felt about governmentusing the data Absolute genius.
(07:08):
We also had the leadership youmentioned, victor W Lowe.
We had a minister who said thisis ridiculous.
Government cannot be thiscomplicated to work with.
It has to be simple.
It has to be like dealing withservice companies, because
government largely is a service.
So we had the most high-levelleadership constantly battling
the ridiculousness of engagingwith government, which gave us
(07:29):
all a license to do it betterand do it differently, to
reimagine how government canactually work.
When COVID hit, the ministerssaid we are going to do things
differently, we are going to dothings simpler, we are going to
make it easier to engage Into.
The line here, of course, is wewant people to be able to live
their lives within theconstraints of COVID and what
(07:50):
people are likely to do sosimple things like QR code
check-ins and QR code check-outs, simple things like delivery of
vouchers for people to try andstimulate the economy was all
done because we all felt that wewere enabled and allowed to do
things differently.
And, of course, service NewSouth Wales was the vehicle that
allowed that delivery.
(08:11):
Data Analytics Center was theplace that allowed us to
understand where best to placethose resources.
But all of that requires datasharing.
So, steve, I'm catching up withyour series of questions.
All of that relies on datasharing and data sharing,
especially within Google, isextraordinarily hard.
There are three broad sets ofreasons why people aren't
(08:31):
sharing data.
They're unwilling, they haveconcerns about all sorts of
things, they're unable, theydon't have the frameworks or
they believe they're not allowedor actually not allowed to do
data sharing.
And two of those changed reallydramatically during COVID.
People were willing to sharedata.
They really understood by thatstage the importance of doing it
.
They also had legislation thatsaid explicitly you will and you
(08:53):
must share data in order tomake things better.
And so, if we got down to theframeworks of the unable, how do
you share data with appropriategovernance, especially data
about people?
So privacy, especially datawhich is sensitive, so sensitive
subjects and particular healthis a sensitive subject.
Building out those data sharingframeworks that we had been
(09:13):
experimenting with for yearssuddenly became important to put
into action.
From an advancement of thematurity and data sharing and
use within government, covid wasamazing.
It was, of course, the turn ofthe maturity and voucher use
within government, covid wasamazing, but it was, of course,
a terrible pandemic.
But from the mechanism to moveus forward, it was amazing from
the perspective of everyone'smaturity and focus improved for
(09:37):
the volume of voucher.
Speaker 3 (09:40):
Yeah, absolutely Necessity, classically being the
mother of invention in thatcase.
Speaker 2 (09:46):
So we had been inventing for years, but it was
all kind of interestingexperimentation.
The difference was we had todeploy it.
We had to actually stopexperimenting and actually do so
.
In this case, necessity was themother of implementation.
Speaker 3 (10:00):
Nice one.
I like that twist.
Look, there's other thingsgoing on, a lot going on.
We've got a Data Availabilityand Transparency Act at the
federal level in Australia.
We've got national dataarchives.
We've got things like the fearprinciples that are supposed to
(10:21):
provide a governance for datasharing that are supposed to
provide a governance for datasharing.
We had that sort of metaphorabout trying to describe data as
an asset or as a resource.
But how do you on the economicimportance and maximize the
economic outcome of data as aresource?
Speaker 2 (10:42):
You see you're touching on some very, very big
issues.
Let me try and approach thisfrom a couple of different
perspectives.
You talked about the DataEvaluating Transparency Act.
Legislation describes things atthe principle level and that's
great.
That says this is how we'regoing to do this, but it's a bit
like describing the principlesof changing a tire or a bit like
(11:03):
the principles of how you'regoing to wire up your house.
It's useful in the sense ofsaying this is what we should be
doing, but it's not practicalin the sense that you consider a
group of people and considerthem talking about changing a
tyre on their car at aprinciples level.
But someone needs to pick upthe tools.
Someone actually needs to getin there and do the doing, and a
(11:24):
lot of the work so far has alldone.
Whether it's theIntergovernmental Agreement on
Banacharing versus theBanacharing-Rose Act doesn't
help unless someone is willingto pick up the tools and a lot
of the good intentions.
During COVID we had theIntergovernmental Agreement on
Banacharing that says you areallowed, we are willing and
(11:45):
you're allowed.
So out of the unwilling, unable, allowed, it spoke to the two
ends of it.
It didn't help with theframeworks and starting to
consider data as actuallysomething which is not just an
Excel spreadsheet, not justsomething that you do a little
bit of analysis on If you wantto use it for important purposes
.
You actually need people whoknow and understand what data is
(12:06):
, who know and understand whatdata governance is, and who know
and understand how to actuallyuse the data for appropriate
purposes and then put guidance,restrictions or prohibitions
around the data products youcreate.
That thinking still doesn'tlargely exist in the world, so I
don't agree.
There are purposes.
People use data for a very,very narrow, specific purpose or
(12:29):
they use data in a very narrow,specific context.
But the general use of data.
We're only just beginning tounderstand the power, the
complexity, the material natureof data and starting to build
frameworks around it.
Most people think of data as anExcel spreadsheet.
That's the mental model mostpeople have.
The practical reality is ifthat data is important or to be
(12:53):
used for an important purpose.
It's really important tounderstand where the data came
from, how it got to, what thechain of custody was, what the
authorizing frameworks are, theelements of data quality, and
then assess whether that data isfit for the purpose you want to
use it for.
And then when you createsomething from that data whether
it's a summary or a subset, oran alert, an alarm, a decision
(13:17):
or an action or even just aninsight.
It's really important tounderstand how to contextualize
appropriate use of that dataproduct and put guidance,
restrictions or prohibitions onhow you use that insight so that
it can appropriately be appliedfor important purposes.
If you're using data for fun,it doesn't matter.
If you're doing correlationsbetween by consumption and
(13:38):
political outcomes, it doesn'tmatter because that's a fun
pilot trick.
It matters when it matters.
And if it really matters, it'simportant to understand the data
before you use it, as you useit and after you create those
data products.
That understanding is openingjust a door to a lot of people,
I mean for these analogies ofthe past.
You know diaries of the newworld and that's such a terrible
(14:00):
, terrible analogy.
It doesn't fit in so manydifferent ways.
I guess the assumption is thatit's plentiful and you can
create value from it.
But that's completely wrong.
Data is much more like the newelectricity.
You have to really evokeyourself if you apply data
inappropriately.
And we all, of course, knowthat.
You know we all have that basicunderstanding of electricity
(14:22):
from a very early age.
We know not to stick a forkinto a toaster.
We know not to stick our tongueinto a light socket.
But each and every one of uscan make a toaster.
Each and every one of us canchange a light bulb, but if we
go to rewire a house, you knowyou need to get an expert in
electrician.
If you're going to build apower station, you know you need
a group of really experts to dothat.
(14:44):
But with data, we all think thatwe're the home handyman.
We just pick up the power toolsand drill holes in walls and do
silly things with data.
We need to get to the pointwhere we understand that data is
extraordinarily powerful or canbe extraordinarily powerful,
and when it's extraordinarilypowerful, we need to build the
right frameworks around it toensure that we do it in a way
which is appropriate and we putrestrictions, guidance around
(15:08):
the data products we create.
The alternative is what happensnow, at least within government
People won't share data becausethey're concerned about what
you're going to do with it andnot only you, but the next
person and the next, next personwill do with the data or the
data products or they'reconcerned that someone can stand
on one leg and squeak andre-identify an individual.
(15:28):
So we either don't share or weboil the goodness out of it and
release open data, which is notthat useful and really really
open data sets are just not thatuseful.
You asked a very big question.
You made a very big statement.
Data sharing can be done insystematic ways those people
sitting around looking at thetire and using principles to try
and get that tire changed.
(15:49):
Someone needs to pick up theframework, someone needs to pick
up the tools and do the doingaround data sharing frameworks
that actually make sense.
But consider, before you usethe data, the fitness for the
purpose as you use the data, andwhat to do with those data
products once you create them.
Speaker 3 (16:07):
So you said a number of things that are just
fascinating.
You're touching on research,and I think it's a word that is
underestimated.
I think that the word has beenbastardized by the slogan do
your own research, as if theperson in the street can analyze
data and understand it andreach robust decisions.
(16:28):
I think that we've cheapenedthe respects institutionally for
professional research, butthere's a fork in the road here
and we could go both ways, andyou've been involved with both
of these.
One of them is the economics ofdata, or the beautiful book
Infonomics that you pointed metowards, ian, a couple of years
ago, so we'll put this in theshow notes.
(16:48):
But Infonomics, I think, was aat work about 2018 and I think
it was focused on almost like ataboo topic, which is how do you
, how do you realise value, howdo you monetise data, now that
there's a lot of darksurveillance, capitalism and a
lot of dark agendas behind thatidea.
But I think that Laney wastrying to make that idea maybe,
(17:11):
but I think that Laney wastrying to make that idea maybe
respectable or governable.
So that's one thing, and wecould talk about the economic
outcomes whether or not usingeconomics to get social outcomes
is a good idea, and the otherone is data quality.
You drew attention recently tothe new ISO 8000 international
standard on data quality, and Iguess that's going from
(17:35):
principles to tools, isn't it?
So can you walk us throughwhether or not data quality as a
standard is something that'sgoing to lead to some good
outcomes as well?
Speaker 2 (17:47):
Once again, you've struggled me until I had to do a
box.
Let me see if I can stepthrough some of those, but that
lady's got infinite knowledge.
Speaker 3 (17:53):
It's a treasure chest , isn't it?
It's not all problems andpoisonous snakes.
Speaker 2 (17:59):
It's a chest of some sort with some stuff inside it,
and what you make of it iswhether it's treasure or
Andorra's box.
So that lady tried to build aframework that said data is
valuable.
We all think about that analogydata is the new oil.
As much as I hate it, theintention there is to try and
say it's valuable, but no one'sreally sure how it's valuable.
So what we've done tried to dowas put a framework in place
(18:21):
that said you can use data foroperational purposes.
You can use data for strategicpurposes.
You can use data to createinsights and all of that's sort
of the internal uses of data.
And there are more than simpleoperational purposes.
You could and should be usingyour data for strategic purposes
.
Most organizations do not.
Most organizations don'trealize the value of their own
data.
They also center around someexternal value aspects to data.
(18:45):
What would someone claim for it?
Or what markets would youcreate?
Or what new value would youcreate if you mixed your data
with someone else's data?
And then, of course, there areelements of putting up potential
loss as well.
What would it cost to reproducethe data?
What would be the impact of younot having exclusive access to
that data and I thought that wasreally, really good.
And then what I'd been tryingto do at the time was work out a
(19:09):
government version of that.
So he talked verybusiness-focused.
I thought what would governmentdo with data?
And I'd also been trying towork out how we can build
frameworks, literally anaccounting framework for data.
So it was a very nice meetingof minds around that idea.
We are, to this day, stillwithout an accounting standard
which values data and tell.
(19:30):
Facebook that their data is notvaluable.
It is the majority of the valueof the company.
Tell LinkedIn their data is notvaluable and it's the majority
of the value of the company, butwe don't have a way of
measuring it from a financeperspective.
So that problem is still beingtackled.
But I think we're gettingbetter at saying it's in here
somewhere.
The problem's in here somewhere.
We don't quite know what it is,but it's in here somewhere.
(19:52):
The problem's in here somewhere.
We don't quite know what it is,but it's in here somewhere.
Data quality is a reallyinteresting element and as much
as I hate the oil analogy fordata, I often try to help people
conceptualize by saying thatdata comes in different
qualities and you can still usepoor quality data, but what you
can do in terms of reliance onthe outcome is quite different.
So it's like putting inAustralia we have 91 grade, 95
(20:13):
grade, 98 grade petrol and wehave diesel and you need to know
whether your car needs to haveor can run on 91, 95, 98, or
whether it runs on diesel.
You need to make sure youabsolutely don't mix them up,
but you can still run your car,your high-performance car.
It can still run on 91.
It just doesn't work so well.
Or you can run your clapped-outcar on 98.
(20:35):
It might be good things for theengine.
The point is you need tounderstand if it's fit for the
purpose you want to use it for.
We trust when we go to a petrolstation, we trust that all the
work around, where it came from,how it got there, that chain of
custody, that chain of problemswe trust all that away because
we trust the brand of petrolstation.
But we still have to make theassessment for ourselves.
(20:57):
Is it fit for the purpose Iwant to use it for?
Would I inadvertently putdiesel into my petrol engine?
Would I inadvertently put 91into my high-performance engine,
which requires 98?
So we still have to make thatdecision.
Data policy is, if you thinkabout all the effort, all the
engineering, all of the process,control, all the governance
(21:20):
that goes into getting thatpetrol into the bowser with data
, we just say I've got an Excelspreadsheet, let's just pop it
in and see whether we get someinsights out.
We need to put the same effortinto the data in order to get
the same reliance on the dataproducts that we create.
And again, those data productscan be really, really broad.
(21:41):
It could be a plot, it could bean insight, it could be an
alert, an alarm, a decision foraction.
It could be any one of thosethings, but in order to rely on
that, it's really important thatwe've done that work around the
gardens.
There are several things that Ilooked at in the past as chief
data scientist, where I ranwhat's called the AI review
committee, and AI is a use ofdata.
(22:03):
We looked at really importantuses of data like sepsis
prediction, emergency departmentpatient emergency department is
about to go through a sepsisepisode.
An algorithm says alert andsomeone rushes over and treats
the patient.
That is literally a life anddeath use of data.
The alert is not a decision.
That arc of alert alertdecision action.
(22:26):
It's not doing something, butit raises the attention of a
human being who responds in aheightened state of awareness to
say this patient is about to gothrough a sepsis episode.
Getting that wrong either falsepositive or false negative can
be really visibly impactful tothat person.
It could be literally life ordeath.
(22:46):
In order to trust the data, inorder to trust the data product,
we really need trust in thatgovernance process that got us
to that point, because ifsomeone's life depends on that
alert.
Even though it's a human alert,that alert leads to a
heightened state of awarenessaround responding to a
life-critical condition.
(23:07):
We really need all thatgovernance to be in place.
So data quality is reallyimportant and the standard you
enter the 8,000-series standardscomes in many, many parts.
It actually says to us thatdata quality is not the issue of
whether you've got day-to-day,month-to-month year-to-year or
month-to-month day-to-dayyear-to-year.
It's not about simple thingslike that simple format of
(23:30):
things.
It's about the entiregovernance process of data.
How does data flow into yourorganization?
What are the controls and thechain of custody, the chain of
authorizing frameworks?
How did the data get collected?
It speaks to if I hesitate tosay the life cycle of data,
because life cycles in data areeither trivial or infinite.
(23:50):
So it tries to speak to lifecycle of data, because life
cycles in data are eithertrivial or infinite.
So it tries to speak to theelement of life cycle and it
says these are all the thingsyou need to consider.
Let me revisit that analogy ofthe petrol station.
It allows us to re-imagine thatfrom a perspective of data and
say there's a lot more to thisthan just bank format.
(24:12):
There's a lot more to it thanwhether you've got the year
right to one decimal point orthe number right to one decimal
point.
There's a lot more to it andit's really important to know
and understand that.
If your use of data isimportant, that you really have
to at least be aware of thesestandards preferably have
someone you've organised and youunderstand these standards or
(24:34):
know where to go to get thatextra help in order to enact
these standards, to haveappropriate confidence and
therefore trust in the databefore you use it, the data as
you use it and the data afteryou, the data products you
create, what you carry aroundthose and the data.
Speaker 1 (24:54):
After the data products, you create what you
can around those.
Well, ian, steve and I, as wetalk about data sharing, we're
often using the term data supplychain, because you've got
issuers of data and they oftenget the data to the party that
needs it to make a decision,which should be multiple
entities.
In between touching that data,you talk about data provenance
(25:20):
and recording the evidence ofthat supply chain journey.
One of the major producers ofcritical data is government
itself, creating Medicarenumbers, driver's license
numbers and many of those mybirth records and the like.
Many of those numbers are usedvery commonly in commercial
(25:43):
contexts, whether that was mydesign or not, but they still
are very important for use foronboarding purposes, for example
, creating an accountrelationship.
So, stephen, I've been thinkingabout government, as, of course,
you're issuers of these numbers.
Thus far, they've really beenissued either on paper or in
(26:05):
plain text, and plain text, ofcourse, is the target of every
breach and the whole planet isaffected.
The number of producers or it'sall led to go any closer to
(26:32):
identifying me as a safe driveror as a driver?
Or why is this driving me toSouth Wales?
Why don't we turn that into averifiable credential that has
security.
It's presented out of a wallet,out of a security device, so
that hardware and software willbe together.
That's the way we think itwould be very helpful for
(26:56):
government to expand its remit,if you will, in terms of that
data issuer Really providing amore useful to society version
of the data that it's alreadyproducing.
How hard is that?
Speaker 3 (27:12):
How hard is it to get .
Speaker 1 (27:14):
Again.
You're laughing.
That tells me something already.
How hard is it to getgovernment to expand its
thinking about?
If we are issuers of data?
How can we get that intosociety in a more useful way?
Speaker 2 (27:35):
So if I were still with government, you would just
ask the career-limiting question.
So thank you for asking that.
Speaker 1 (27:41):
Well, that's also delighted.
You aren't in government anylonger.
Speaker 2 (27:44):
Yeah, it's not that long ago, so I'm still a little
shy with questions like that.
It's really hard for a coupleof reasons.
Let me wind back.
Let me try, and perhaps thatquestion must have a stake.
I'll be going, so I'm still.
I'm sure I've written questionslike that.
It's really hard for a coupleof reasons.
Let me wind back.
Let me try and approach thatquestion much like I have with
Steve.
You've asked a really importantand big question.
The first part of that is, ofcourse, that's what we need to
do.
We are stuck in the 20thcentury thinking about creating
(28:08):
honeypot kind of sense, where webring all the data together and
make ourselves a target forcyber attack.
Once upon a time you had to beinteresting for someone to put
the effort into trying to attackyou from a cyber perspective.
Nowadays we're all justcollateral damage.
You don't have to be at allspecial.
Every data set is going to beattacked at some point.
It's just the nature of theworld we live in.
(28:28):
So let's not create those datasets in the first place.
The problem is most people ingovernment can't see a different
way forward.
Most people wouldn't think thatthat's just the way you do
stuff.
You just build data setstogether.
So there is an importantmindset change that must happen.
That isn't yet happening.
(28:49):
So that's the real problem.
Virtualization of data is a wayof doing this.
Being able to give people youwere hinting at it, let me make
it explicit giving people theirown explicit control of their
most important data assets, suchas their identity, is really an
(29:09):
important way of thinking abouthow we should go forward.
Most of the time, I don't needto ask you the question what's
your date of birth?
I'm asking you the question infront of government are you old
enough to enter this licensepremises?
Is A greater than B?
And if the answer is yes, thenoff you go.
I don't need to know your dateof birth.
I don't need to know where youwere born.
I don't need to know all thesort of stuff that people
(29:30):
typically will take from you.
If you try to enter a licensefor purposes, I don't need to
know whether you have aresponsible.
I don't need to know your dateof birth, whether you've got a
responsible service of alcohollicense.
All I need to know is is itvalid?
Do you have it?
Is it valid?
Is it within date?
And the answer to that questionthen allows that next step to
happen.
That thinking is a step beyondwhere most people are, because
(29:52):
we're so used to having physicalidentity that we hand over,
which has lots more informationthat is actually required to
answer the question are you oldenough to go on this license
premises?
But that's where the work needsto go.
Sorry, george.
Speaker 1 (30:05):
So I totally get your answer.
What is it that people like usneed to do to change that
mindset?
I mean, honestly, that's kindof the mission of why we might
even be doing this podcast.
We want this discussion to beout there.
Speaker 2 (30:24):
Okay, many, many times I've sat around a table
with government colleagues andlistened to their plans for how
they're going to do things infuture and thought to myself
you're kidding, you'reabsolutely kidding.
Really, this is what we plan todo and then try to help people
(30:44):
understand there is a way ofdoing things differently.
The hardest thing in the worldto do is change someone's mind
If they're set on a path ofaction they've got their
planning, they've got theirprocesses, they've got their
resources, they've got theirbudget and you try to say to
someone don't do it that way, doit differently.
It's really hard to do.
You're an annoying barking dogon the side of the road where
(31:06):
they're on a mission to dosomething.
It genuinely takes a crisis.
It genuinely takes a crisis tohelp people think that they can
do things differently.
Covid was an amazing agent forchange for the way we think
about data.
Data breaches, unfortunately,are an amazing agent for change.
When it impacts a group, theysuddenly listen, they're
(31:26):
suddenly ready to do thingsdifferently.
But watching everybody else dothings or getting good advice,
that's.
That's interesting, george, butI'm busy, busy doing grown-up
stuff over here.
I don't really need your advice.
So, unfortunately, in somecases, you have to live through
a couple of crises,unfortunately.
You have to help people whenwhen they fall in the hole.
You have to help people whenthey fall in the hole.
(31:47):
You have to help people out.
Let's say this is allowed doingthings differently and the
really, at least in Australia,there's been a tendency for
governments to de-skill over thelast decades.
So we have a lot of generalistsin government who are not
technical, who don't get thenext level of detail and who
just think well, you know, thisis what we do, this is how we do
(32:09):
it.
So there isn't a simple answer,but being there to help this
thing, these are elements of howwe get going.
So I hope this podcast, I hopesomeone listens to this podcast
and says, oh, maybe I can dothings differently, maybe I can
look at this differently, butit's very hard to change the
mind of someone.
It's really hard to change themind of a public servant.
(32:30):
It's really, really hard tochange the mind of a public
servant who's got all theirpieces, their program logic laid
out and their budget and theirresources are ready to go.
It's really tough to do.
Speaker 3 (32:40):
We've got other catalysts that I guess are
programs that governments haveextensively committed to, like
Digital ID now and open banking,do you find?
Well, it's a leading question,ian, but we find patterns
through these programs that arecommon and they suggest to us
(33:02):
maybe there's an emergingsophistication about how to
govern data, what the parametersare that are governable, what
can be standardized, what shouldbe decentralized and left to be
determined on a local sort ofbasis.
One of the anxieties I haveabout ISO 8000 is that we had to
(33:23):
see this process or thismovement become too robotic in
the way that ISO 9000 did.
It poisoned the well by makingquality management like an
algorithm or a spreadsheet andit took a lot of the judgment
and intuition and humanity outof quality management.
(33:44):
And you know 30 years and Ithink ISO 8000 is going to do a
better job.
But are you getting to thepoint where you can imagine how
this weird, intangible thing isgoing to be governable in a
general way?
Speaker 2 (34:01):
Thank you, steve.
Once again you've answered theright question.
On the record, I love standards.
I'm a huge fan of standards.
Without standards, we areelectrocuting ourselves, we are
sticking our tongue in the light, we're doing all that dumb
stuff.
There is a tendency forstandards to get a little
brilliantly focused, a littleheavy, and you touched on an
(34:23):
example right there.
I think there are otherstandards that get so cumbersome
that they become less usefulthan they could be, but here's
one I prepared earlier.
We've just finished, aftermultiple years of work at ISO
IEC, so at JTC1 standard on datausage 2.5207, terminology and
(34:45):
use cases.
Believe it or not, defining theterminology, getting the
definitions right, is somethingthat largely didn't exist or
happened in a multitude ofdifferent environments, and we
fought like cats in a bag aboutgetting those definitions.
Because the cloud people thinkthis, the document exchange
people think this, the AI peoplethink this, the smart cities
people think this.
We said let's get that down.
(35:07):
And then guidance for usage,and one of those usages, of
course, is AI.
So we connected to the world ofthe AI folks, we connected to
TAO, we connected to all thisstuff and we said you know, it
needs to be simpler.
It also needs to acknowledgethat there is uniqueness in
every circumstance.
There is some of that judgmentthat's required, but the formula
(35:28):
for these standards is, quitesimply what do we need to know
about the data before we use it?
What do we need to understandabout the data as we assess it
for fitness purpose?
What are the guidance,restrictions or prohibitions we
need to put around the dataproduct?
We've read the thing that I'vebeen sort of slipping into the
conversation to know is exactlywhat's in that standard, and I
think it balances the slippinginto the conversation to know is
exactly what's in that standard, and it acknowledges the fact
(35:50):
that context matters, yourenvironment matters, your
sensitivities matter, yourparticular circumstances about
your data matter, but most of it80% of it, by the way actually
is a repeatable process.
But the 10% is your judgmentand your context and your
environment and you need to dosome level setting about layers
of control very high control,high control, moderate control,
(36:12):
low control, no control.
You get to decide, but this ishow you decide.
This is a recipe for you todecide.
You still make that dish, youstill bake that cake.
It's your cake, but this is howto turn the oven on.
This is the temperature to turnthe oven at.
This is how long to think aboutit should go on the oven for.
But you're still doing thatsubjective confinement to not
(36:34):
make it robotic, because data isreally complicated.
It's so complicated that infact, there are seven degrees of
infinite complexity in datasharing and use.
Seven degrees of infinitecomplexity, data sharing and use
, which we're going into a lotof trouble.
But that basically comes downto every data set can be used
for an infinite number ofpurposes.
They might not be fit forpurpose, but you can use it for
(36:55):
an infinite number of purposes.
Context changes the sensitivityof data, so you bring that data
set close to another data setor you know something about
something in the data.
The context of use changes it.
All those ambiguities, all thosepossibilities, try to create
the things you create.
Whether it's a salary data setor a self-set or an insight,
(37:17):
that same issue exists, thatcontext matters, and you can use
that insight for a range ofdifferent purposes and there's a
couple more steps in there, butit means that it's so mercurial
, it's so difficult, so slippery, it's like smoke, it's like
trying to eat your arms aroundsomething which is really
dangerous.
So if you can flatten all thoseambiguities or uncertainties
(37:40):
into a couple of different types, layers of control, then you
can start to do something usefulwith it.
That's what these new standardstalk about.
It says this is what we mean,these are the definitions and we
mean the same thingconsistently and these are the
ways of thinking about how youuse data.
And therefore it's 5.7,.
(38:02):
Terminal Newscast is 5.212, aguidance for usage, and one of
those usages just happens to beAI.
Speaker 3 (38:08):
Brilliant.
Let's get those details fromyou and put them in the show
notes later on and see how muchof that material is accessible
to our audience.
It's fantastic guidance.
Thanks, Ian.
Speaker 2 (38:21):
It was a lander of love.
A lot of landers and a lot ofissues.
People who try to get theirarms around it start to talk
about things like what's thedifference between data and
information.
That one's easier to dial out,but then information and
knowledge and wisdom.
It was three in the morning.
We were really close to the endof the standard.
(38:42):
Someone said we haven't talkedabout wisdom.
I said no, the ISO definitionof wisdom.
We'll probably understand it.
If not, we're ending thisconversation.
Wait, I love it.
Speaker 1 (38:56):
Well, I think it behooves us to exercise some
wisdom at this point and tobring this to a conclusion.
Ian, thank you very much.
I'm really grateful for yourinsights and your thoughts.
I'm going to be thinking abouthow we shift government minds
(39:16):
when they already have all themomentum in the world for a
particular vision and we'reusing them over slightly.
Speaker 3 (39:23):
Thank you very much.
Thanks so much, ian.
Here's to data and informationand, if we can get, to wisdom.
Thanks for helping us along theway Later.
Thank you.
Welcome to Making Data Better, a podcast about
data quality and the impact ithas on how we protect, manage
and use the digital datacritical to our lives.
I'm George Higley, partner atBlackstuff Consulting, and I'm
delighted to join my partner,steve Wilson.
Great to see you again.
Let's get right into it.
(00:30):
We've got a terrific guesttoday, ian Opperman.
Among his many accomplishmentsand we'll ask Ian to you have to
brag about them yourself he'son Wealth to Check a couple of
years he's advisor to a legaltech startup.
He was struck by your role as aprofessor at the University of
Technology in Sydney, mostrecently as chief data scientist
(00:52):
for the state of New SouthWales.
Nsw has been a leader in dataanalytics and open data.
So, ian, we're really delightedto have you here and look
forward to hearing your thinkingand a bit of the story of New
South Wales and the DigitalDriver's License and a lot of
other topics, all right.
So, Ian to you, love to hearabout what you're doing today.
Speaker 2 (01:11):
George, great to be here and great to be talking
about data quality.
As I mentioned during theintroduction, I was the Chief
Data Scientist for New SouthWales government for eight and a
half years, an experimentalrole which was really about
helping government think aboutdata differently and see data
differently.
Now left and I've joined astartup.
Of course, the thing you doafter being in government is
join a startup and that startupis all about helping governments
(01:33):
do government differently and,of course, using data, ai and
digital tools to do thingsdifferently.
Joined with a former minister,the former CEO of what's called
Service NSW and the formerminister's chief of staff, and
we're out there doing goodthings with data, digital and AI
to help other governmentsaround the world.
Speaker 3 (01:54):
Fantastic, and I myself had the privilege of
consulting to Service NSW on andoff, especially through COVID,
and seeing firsthand you knowthe agility of that organisation
.
It recently won an awardactually for service innovation,
which was just such a greatresult and a real testimony to
(02:15):
the quality of the people there.
The innovation included thingslike the Data Analytics Centre
which Vian I think you had atitle role of some sort there,
but maybe you could tell ourlisteners about the importance
of data and how New South Walesreally sort of walked the talk
in open data and open government.
Speaker 2 (02:33):
So I joined New South Wales government in 2015 as the
first ever chief data scientistand was given not only the gun
and the badge, but also a pieceof legislation that said you had
to share data with what'scalled the Data Analytics Centre
, and the centre was all abouttackling wicked policy
challenges or problems that arecomplex, subtle and ultimately
have people's behaviour at heart.
(02:53):
We adopted the philosophy thatif we're going to tackle a
wicked problem, we need todeliberately re-complicate that
problem and try to address asmuch of it as possible, and that
, I have to tell you, was reallyan unusual philosophy to take
in the government, whichtypically is about keeping it
(03:14):
simple.
That was helping governmentthink differently about how to
use data to understand problemsdifferently, use data sets a
myriad of different data sets tosee the problem differently.
Every data set is incomplete.
Every data set is less than100% quality.
Every data set doesn't give youfull coverage, but it's a point
of life, and the wholephilosophy was let's see how
many different ways we canunderstand that problem and then
(03:35):
use analytics and AI to reallymake sense of it and do things
differently.
The flow I was taking to ServiceNSW was now we better
understand the problem.
Let's deliver differently, andmy colleague, david Rees, was
really the champion of ServiceNSW.
He took the philosophy thatgovernment can, government can
do this, government can be agile, government can experiment,
(03:57):
government can try, test andlearn and build out services in
an agile way.
We had all been workingtogether for some years when
COVID hit.
And when COVID hit, we wereready to respond and that was
the really big differencebetween before and after that
beginning of thatexperimentation in 2015.
When COVID hit, we couldrespond with data-driven tools
(04:21):
and data-driven elements ofidentity and data-driven ways of
delivering services, becausewe'd spent years experimenting
up to that point.
Speaker 3 (04:30):
And of course there's a really good social contract,
I think in Australia at least,for the use of data, the
reasonable use of data.
You know our COVID response inNew South Wales is very strong
through New South Wales Health.
The public more or lessembraced the idea that contact
tracing was going to be a thingand you know we had such a rash
of private sector contacttracing cloud solutions.
(04:52):
It was a mess and I think itwas Victor Dominello himself who
said we should make this easyand uniform for citizens.
Speaker 1 (05:01):
But I also want to like tag.
Speaker 3 (05:02):
The point you just made about complicating things,
I think is genius.
I'm a mathematician andscientist by training and I
think about generalizing and inengineering we try to simplify,
look for simplifying assumptions.
In science, sometimes we'relooking for that grand general
theory.
Usually generalization makesthings more complicated, but in
(05:26):
our area of digital identity, wethink that generalising the
digital identity problem to adata problem surprisingly makes
things simpler.
That convergence I'm going tocome back to that later, the
convergence of governance ofdata and ID.
But you know, for now can we gobig and talk about data sharing
(05:46):
, because I think that that'sone of the defining themes of
the digital economy.
We're doing open banking inAustralia.
We've got open data.
How do you think about datasharing generally and what are
the societal parameters that youthink about?
Speaker 2 (06:03):
So, steve, you've addressed the entire Pandora's
box with your set of commentsthere.
Let me try and unpack thesub-elements of it.
The first was around sociallicence, and it's an interesting
position to be in governmentand be talking about social
licence.
Social licence is typicallyassociated with mining companies
and such things.
You don't get a choice aboutengaging with government.
(06:24):
When I started in 2015, therewas this real concern that if
government was doing dataanalytics, if government was
doing analysis of data at a sortof at scale, that there would
be this concern from the publicand the perception of public
perception really inhibited awhole lot of activities.
It wasn't the public saying no,we don't want you to do this,
(06:44):
it was inside governmentpeople's thoughts about what the
public might say.
That was really a limitingfactor.
So we went out there and testedit.
We got the data and the datawas, quite simply, we don't want
to answer the same questionbecause we want to deal with one
government, we don't want todeal with 15 different parts of
government.
So we actually spent some timeusing the data to understand how
people felt about governmentusing the data Absolute genius.
(07:08):
We also had the leadership youmentioned, victor W Lowe.
We had a minister who said thisis ridiculous.
Government cannot be thiscomplicated to work with.
It has to be simple.
It has to be like dealing withservice companies, because
government largely is a service.
So we had the most high-levelleadership constantly battling
the ridiculousness of engagingwith government, which gave us
(07:29):
all a license to do it betterand do it differently, to
reimagine how government canactually work.
When COVID hit, the ministerssaid we are going to do things
differently, we are going to dothings simpler, we are going to
make it easier to engage Into.
The line here, of course, is wewant people to be able to live
their lives within theconstraints of COVID and what
(07:50):
people are likely to do sosimple things like QR code
check-ins and QR code check-outs, simple things like delivery of
vouchers for people to try andstimulate the economy was all
done because we all felt that wewere enabled and allowed to do
things differently.
And, of course, service NewSouth Wales was the vehicle that
allowed that delivery.
(08:11):
Data Analytics Center was theplace that allowed us to
understand where best to placethose resources.
But all of that requires datasharing.
So, steve, I'm catching up withyour series of questions.
All of that relies on datasharing and data sharing,
especially within Google, isextraordinarily hard.
There are three broad sets ofreasons why people aren't
(08:31):
sharing data.
They're unwilling, they haveconcerns about all sorts of
things, they're unable, theydon't have the frameworks or
they believe they're not allowedor actually not allowed to do
data sharing.
And two of those changed reallydramatically during COVID.
People were willing to sharedata.
They really understood by thatstage the importance of doing it
.
They also had legislation thatsaid explicitly you will and you
(08:53):
must share data in order tomake things better.
And so, if we got down to theframeworks of the unable, how do
you share data with appropriategovernance, especially data
about people?
So privacy, especially datawhich is sensitive, so sensitive
subjects and particular healthis a sensitive subject.
Building out those data sharingframeworks that we had been
(09:13):
experimenting with for yearssuddenly became important to put
into action.
From an advancement of thematurity and data sharing and
use within government, covid wasamazing.
It was, of course, the turn ofthe maturity and voucher use
within government, covid wasamazing, but it was, of course,
a terrible pandemic.
But from the mechanism to moveus forward, it was amazing from
the perspective of everyone'smaturity and focus improved for
(09:37):
the volume of voucher.
Speaker 3 (09:40):
Yeah, absolutely Necessity, classically being the
mother of invention in thatcase.
Speaker 2 (09:46):
So we had been inventing for years, but it was
all kind of interestingexperimentation.
The difference was we had todeploy it.
We had to actually stopexperimenting and actually do so
.
In this case, necessity was themother of implementation.
Speaker 3 (10:00):
Nice one.
I like that twist.
Look, there's other thingsgoing on, a lot going on.
We've got a Data Availabilityand Transparency Act at the
federal level in Australia.
We've got national dataarchives.
We've got things like the fearprinciples that are supposed to
(10:21):
provide a governance for datasharing that are supposed to
provide a governance for datasharing.
We had that sort of metaphorabout trying to describe data as
an asset or as a resource.
But how do you on the economicimportance and maximize the
economic outcome of data as aresource?
Speaker 2 (10:42):
You see you're touching on some very, very big
issues.
Let me try and approach thisfrom a couple of different
perspectives.
You talked about the DataEvaluating Transparency Act.
Legislation describes things atthe principle level and that's
great.
That says this is how we'regoing to do this, but it's a bit
like describing the principlesof changing a tire or a bit like
(11:03):
the principles of how you'regoing to wire up your house.
It's useful in the sense ofsaying this is what we should be
doing, but it's not practicalin the sense that you consider a
group of people and considerthem talking about changing a
tyre on their car at aprinciples level.
But someone needs to pick upthe tools.
Someone actually needs to getin there and do the doing, and a
(11:24):
lot of the work so far has alldone.
Whether it's theIntergovernmental Agreement on
Banacharing versus theBanacharing-Rose Act doesn't
help unless someone is willingto pick up the tools and a lot
of the good intentions.
During COVID we had theIntergovernmental Agreement on
Banacharing that says you areallowed, we are willing and
(11:45):
you're allowed.
So out of the unwilling, unable, allowed, it spoke to the two
ends of it.
It didn't help with theframeworks and starting to
consider data as actuallysomething which is not just an
Excel spreadsheet, not justsomething that you do a little
bit of analysis on If you wantto use it for important purposes
.
You actually need people whoknow and understand what data is
(12:06):
, who know and understand whatdata governance is, and who know
and understand how to actuallyuse the data for appropriate
purposes and then put guidance,restrictions or prohibitions
around the data products youcreate.
That thinking still doesn'tlargely exist in the world, so I
don't agree.
There are purposes.
People use data for a very,very narrow, specific purpose or
(12:29):
they use data in a very narrow,specific context.
But the general use of data.
We're only just beginning tounderstand the power, the
complexity, the material natureof data and starting to build
frameworks around it.
Most people think of data as anExcel spreadsheet.
That's the mental model mostpeople have.
The practical reality is ifthat data is important or to be
(12:53):
used for an important purpose.
It's really important tounderstand where the data came
from, how it got to, what thechain of custody was, what the
authorizing frameworks are, theelements of data quality, and
then assess whether that data isfit for the purpose you want to
use it for.
And then when you createsomething from that data whether
it's a summary or a subset, oran alert, an alarm, a decision
(13:17):
or an action or even just aninsight.
It's really important tounderstand how to contextualize
appropriate use of that dataproduct and put guidance,
restrictions or prohibitions onhow you use that insight so that
it can appropriately be appliedfor important purposes.
If you're using data for fun,it doesn't matter.
If you're doing correlationsbetween by consumption and
(13:38):
political outcomes, it doesn'tmatter because that's a fun
pilot trick.
It matters when it matters.
And if it really matters, it'simportant to understand the data
before you use it, as you useit and after you create those
data products.
That understanding is openingjust a door to a lot of people,
I mean for these analogies ofthe past.
You know diaries of the newworld and that's such a terrible
(14:00):
, terrible analogy.
It doesn't fit in so manydifferent ways.
I guess the assumption is thatit's plentiful and you can
create value from it.
But that's completely wrong.
Data is much more like the newelectricity.
You have to really evokeyourself if you apply data
inappropriately.
And we all, of course, knowthat.
You know we all have that basicunderstanding of electricity
(14:22):
from a very early age.
We know not to stick a forkinto a toaster.
We know not to stick our tongueinto a light socket.
But each and every one of uscan make a toaster.
Each and every one of us canchange a light bulb, but if we
go to rewire a house, you knowyou need to get an expert in
electrician.
If you're going to build apower station, you know you need
a group of really experts to dothat.
(14:44):
But with data, we all think thatwe're the home handyman.
We just pick up the power toolsand drill holes in walls and do
silly things with data.
We need to get to the pointwhere we understand that data is
extraordinarily powerful or canbe extraordinarily powerful,
and when it's extraordinarilypowerful, we need to build the
right frameworks around it toensure that we do it in a way
which is appropriate and we putrestrictions, guidance around
(15:08):
the data products we create.
The alternative is what happensnow, at least within government
People won't share data becausethey're concerned about what
you're going to do with it andnot only you, but the next
person and the next, next personwill do with the data or the
data products or they'reconcerned that someone can stand
on one leg and squeak andre-identify an individual.
(15:28):
So we either don't share or weboil the goodness out of it and
release open data, which is notthat useful and really really
open data sets are just not thatuseful.
You asked a very big question.
You made a very big statement.
Data sharing can be done insystematic ways those people
sitting around looking at thetire and using principles to try
and get that tire changed.
(15:49):
Someone needs to pick up theframework, someone needs to pick
up the tools and do the doingaround data sharing frameworks
that actually make sense.
But consider, before you usethe data, the fitness for the
purpose as you use the data, andwhat to do with those data
products once you create them.
Speaker 3 (16:07):
So you said a number of things that are just
fascinating.
You're touching on research,and I think it's a word that is
underestimated.
I think that the word has beenbastardized by the slogan do
your own research, as if theperson in the street can analyze
data and understand it andreach robust decisions.
(16:28):
I think that we've cheapenedthe respects institutionally for
professional research, butthere's a fork in the road here
and we could go both ways, andyou've been involved with both
of these.
One of them is the economics ofdata, or the beautiful book
Infonomics that you pointed metowards, ian, a couple of years
ago, so we'll put this in theshow notes.
(16:48):
But Infonomics, I think, was aat work about 2018 and I think
it was focused on almost like ataboo topic, which is how do you
, how do you realise value, howdo you monetise data, now that
there's a lot of darksurveillance, capitalism and a
lot of dark agendas behind thatidea.
But I think that Laney wastrying to make that idea maybe,
(17:11):
but I think that Laney wastrying to make that idea maybe
respectable or governable.
So that's one thing, and wecould talk about the economic
outcomes whether or not usingeconomics to get social outcomes
is a good idea, and the otherone is data quality.
You drew attention recently tothe new ISO 8000 international
standard on data quality, and Iguess that's going from
(17:35):
principles to tools, isn't it?
So can you walk us throughwhether or not data quality as a
standard is something that'sgoing to lead to some good
outcomes as well?
Speaker 2 (17:47):
Once again, you've struggled me until I had to do a
box.
Let me see if I can stepthrough some of those, but that
lady's got infinite knowledge.
Speaker 3 (17:53):
It's a treasure chest , isn't it?
It's not all problems andpoisonous snakes.
Speaker 2 (17:59):
It's a chest of some sort with some stuff inside it,
and what you make of it iswhether it's treasure or
Andorra's box.
So that lady tried to build aframework that said data is
valuable.
We all think about that analogydata is the new oil.
As much as I hate it, theintention there is to try and
say it's valuable, but no one'sreally sure how it's valuable.
So what we've done tried to dowas put a framework in place
(18:21):
that said you can use data foroperational purposes.
You can use data for strategicpurposes.
You can use data to createinsights and all of that's sort
of the internal uses of data.
And there are more than simpleoperational purposes.
You could and should be usingyour data for strategic purposes
.
Most organizations do not.
Most organizations don'trealize the value of their own
data.
They also center around someexternal value aspects to data.
(18:45):
What would someone claim for it?
Or what markets would youcreate?
Or what new value would youcreate if you mixed your data
with someone else's data?
And then, of course, there areelements of putting up potential
loss as well.
What would it cost to reproducethe data?
What would be the impact of younot having exclusive access to
that data and I thought that wasreally, really good.
And then what I'd been tryingto do at the time was work out a
(19:09):
government version of that.
So he talked verybusiness-focused.
I thought what would governmentdo with data?
And I'd also been trying towork out how we can build
frameworks, literally anaccounting framework for data.
So it was a very nice meetingof minds around that idea.
We are, to this day, stillwithout an accounting standard
which values data and tell.
(19:30):
Facebook that their data is notvaluable.
It is the majority of the valueof the company.
Tell LinkedIn their data is notvaluable and it's the majority
of the value of the company, butwe don't have a way of
measuring it from a financeperspective.
So that problem is still beingtackled.
But I think we're gettingbetter at saying it's in here
somewhere.
The problem's in here somewhere.
We don't quite know what it is,but it's in here somewhere.
(19:52):
The problem's in here somewhere.
We don't quite know what it is,but it's in here somewhere.
Data quality is a reallyinteresting element and as much
as I hate the oil analogy fordata, I often try to help people
conceptualize by saying thatdata comes in different
qualities and you can still usepoor quality data, but what you
can do in terms of reliance onthe outcome is quite different.
So it's like putting inAustralia we have 91 grade, 95
(20:13):
grade, 98 grade petrol and wehave diesel and you need to know
whether your car needs to haveor can run on 91, 95, 98, or
whether it runs on diesel.
You need to make sure youabsolutely don't mix them up,
but you can still run your car,your high-performance car.
It can still run on 91.
It just doesn't work so well.
Or you can run your clapped-outcar on 98.
(20:35):
It might be good things for theengine.
The point is you need tounderstand if it's fit for the
purpose you want to use it for.
We trust when we go to a petrolstation, we trust that all the
work around, where it came from,how it got there, that chain of
custody, that chain of problemswe trust all that away because
we trust the brand of petrolstation.
But we still have to make theassessment for ourselves.
(20:57):
Is it fit for the purpose Iwant to use it for?
Would I inadvertently putdiesel into my petrol engine?
Would I inadvertently put 91into my high-performance engine,
which requires 98?
So we still have to make thatdecision.
Data policy is, if you thinkabout all the effort, all the
engineering, all of the process,control, all the governance
(21:20):
that goes into getting thatpetrol into the bowser with data
, we just say I've got an Excelspreadsheet, let's just pop it
in and see whether we get someinsights out.
We need to put the same effortinto the data in order to get
the same reliance on the dataproducts that we create.
And again, those data productscan be really, really broad.
(21:41):
It could be a plot, it could bean insight, it could be an
alert, an alarm, a decision foraction.
It could be any one of thosethings, but in order to rely on
that, it's really important thatwe've done that work around the
gardens.
There are several things that Ilooked at in the past as chief
data scientist, where I ranwhat's called the AI review
committee, and AI is a use ofdata.
(22:03):
We looked at really importantuses of data like sepsis
prediction, emergency departmentpatient emergency department is
about to go through a sepsisepisode.
An algorithm says alert andsomeone rushes over and treats
the patient.
That is literally a life anddeath use of data.
The alert is not a decision.
That arc of alert alertdecision action.
(22:26):
It's not doing something, butit raises the attention of a
human being who responds in aheightened state of awareness to
say this patient is about to gothrough a sepsis episode.
Getting that wrong either falsepositive or false negative can
be really visibly impactful tothat person.
It could be literally life ordeath.
(22:46):
In order to trust the data, inorder to trust the data product,
we really need trust in thatgovernance process that got us
to that point, because ifsomeone's life depends on that
alert.
Even though it's a human alert,that alert leads to a
heightened state of awarenessaround responding to a
life-critical condition.
(23:07):
We really need all thatgovernance to be in place.
So data quality is reallyimportant and the standard you
enter the 8,000-series standardscomes in many, many parts.
It actually says to us thatdata quality is not the issue of
whether you've got day-to-day,month-to-month year-to-year or
month-to-month day-to-dayyear-to-year.
It's not about simple thingslike that simple format of
(23:30):
things.
It's about the entiregovernance process of data.
How does data flow into yourorganization?
What are the controls and thechain of custody, the chain of
authorizing frameworks?
How did the data get collected?
It speaks to if I hesitate tosay the life cycle of data,
because life cycles in data areeither trivial or infinite.
(23:50):
So it tries to speak to lifecycle of data, because life
cycles in data are eithertrivial or infinite.
So it tries to speak to theelement of life cycle and it
says these are all the thingsyou need to consider.
Let me revisit that analogy ofthe petrol station.
It allows us to re-imagine thatfrom a perspective of data and
say there's a lot more to thisthan just bank format.
(24:12):
There's a lot more to it thanwhether you've got the year
right to one decimal point orthe number right to one decimal
point.
There's a lot more to it andit's really important to know
and understand that.
If your use of data isimportant, that you really have
to at least be aware of thesestandards preferably have
someone you've organised and youunderstand these standards or
(24:34):
know where to go to get thatextra help in order to enact
these standards, to haveappropriate confidence and
therefore trust in the databefore you use it, the data as
you use it and the data afteryou, the data products you
create, what you carry aroundthose and the data.
Speaker 1 (24:54):
After the data products, you create what you
can around those.
Well, ian, steve and I, as wetalk about data sharing, we're
often using the term data supplychain, because you've got
issuers of data and they oftenget the data to the party that
needs it to make a decision,which should be multiple
entities.
In between touching that data,you talk about data provenance
(25:20):
and recording the evidence ofthat supply chain journey.
One of the major producers ofcritical data is government
itself, creating Medicarenumbers, driver's license
numbers and many of those mybirth records and the like.
Many of those numbers are usedvery commonly in commercial
(25:43):
contexts, whether that was mydesign or not, but they still
are very important for use foronboarding purposes, for example
, creating an accountrelationship.
So, stephen, I've been thinkingabout government, as, of course,
you're issuers of these numbers.
Thus far, they've really beenissued either on paper or in
(26:05):
plain text, and plain text, ofcourse, is the target of every
breach and the whole planet isaffected.
The number of producers or it'sall led to go any closer to
(26:32):
identifying me as a safe driveror as a driver?
Or why is this driving me toSouth Wales?
Why don't we turn that into averifiable credential that has
security.
It's presented out of a wallet,out of a security device, so
that hardware and software willbe together.
That's the way we think itwould be very helpful for
(26:56):
government to expand its remit,if you will, in terms of that
data issuer Really providing amore useful to society version
of the data that it's alreadyproducing.
How hard is that?
Speaker 3 (27:12):
How hard is it to get .
Speaker 1 (27:14):
Again.
You're laughing.
That tells me something already.
How hard is it to getgovernment to expand its
thinking about?
If we are issuers of data?
How can we get that intosociety in a more useful way?
Speaker 2 (27:35):
So if I were still with government, you would just
ask the career-limiting question.
So thank you for asking that.
Speaker 1 (27:41):
Well, that's also delighted.
You aren't in government anylonger.
Speaker 2 (27:44):
Yeah, it's not that long ago, so I'm still a little
shy with questions like that.
It's really hard for a coupleof reasons.
Let me wind back.
Let me try, and perhaps thatquestion must have a stake.
I'll be going, so I'm still.
I'm sure I've written questionslike that.
It's really hard for a coupleof reasons.
Let me wind back.
Let me try and approach thatquestion much like I have with
Steve.
You've asked a really importantand big question.
The first part of that is, ofcourse, that's what we need to
do.
We are stuck in the 20thcentury thinking about creating
(28:08):
honeypot kind of sense, where webring all the data together and
make ourselves a target forcyber attack.
Once upon a time you had to beinteresting for someone to put
the effort into trying to attackyou from a cyber perspective.
Nowadays we're all justcollateral damage.
You don't have to be at allspecial.
Every data set is going to beattacked at some point.
It's just the nature of theworld we live in.
(28:28):
So let's not create those datasets in the first place.
The problem is most people ingovernment can't see a different
way forward.
Most people wouldn't think thatthat's just the way you do
stuff.
You just build data setstogether.
So there is an importantmindset change that must happen.
That isn't yet happening.
(28:49):
So that's the real problem.
Virtualization of data is a wayof doing this.
Being able to give people youwere hinting at it, let me make
it explicit giving people theirown explicit control of their
most important data assets, suchas their identity, is really an
(29:09):
important way of thinking abouthow we should go forward.
Most of the time, I don't needto ask you the question what's
your date of birth?
I'm asking you the question infront of government are you old
enough to enter this licensepremises?
Is A greater than B?
And if the answer is yes, thenoff you go.
I don't need to know your dateof birth.
I don't need to know where youwere born.
I don't need to know all thesort of stuff that people
(29:30):
typically will take from you.
If you try to enter a licensefor purposes, I don't need to
know whether you have aresponsible.
I don't need to know your dateof birth, whether you've got a
responsible service of alcohollicense.
All I need to know is is itvalid?
Do you have it?
Is it valid?
Is it within date?
And the answer to that questionthen allows that next step to
happen.
That thinking is a step beyondwhere most people are, because
(29:52):
we're so used to having physicalidentity that we hand over,
which has lots more informationthat is actually required to
answer the question are you oldenough to go on this license
premises?
But that's where the work needsto go.
Sorry, george.
Speaker 1 (30:05):
So I totally get your answer.
What is it that people like usneed to do to change that
mindset?
I mean, honestly, that's kindof the mission of why we might
even be doing this podcast.
We want this discussion to beout there.
Speaker 2 (30:24):
Okay, many, many times I've sat around a table
with government colleagues andlistened to their plans for how
they're going to do things infuture and thought to myself
you're kidding, you'reabsolutely kidding.
Really, this is what we plan todo and then try to help people
(30:44):
understand there is a way ofdoing things differently.
The hardest thing in the worldto do is change someone's mind
If they're set on a path ofaction they've got their
planning, they've got theirprocesses, they've got their
resources, they've got theirbudget and you try to say to
someone don't do it that way, doit differently.
It's really hard to do.
You're an annoying barking dogon the side of the road where
(31:06):
they're on a mission to dosomething.
It genuinely takes a crisis.
It genuinely takes a crisis tohelp people think that they can
do things differently.
Covid was an amazing agent forchange for the way we think
about data.
Data breaches, unfortunately,are an amazing agent for change.
When it impacts a group, theysuddenly listen, they're
(31:26):
suddenly ready to do thingsdifferently.
But watching everybody else dothings or getting good advice,
that's.
That's interesting, george, butI'm busy, busy doing grown-up
stuff over here.
I don't really need your advice.
So, unfortunately, in somecases, you have to live through
a couple of crises,unfortunately.
You have to help people whenwhen they fall in the hole.
You have to help people whenthey fall in the hole.
(31:47):
You have to help people out.
Let's say this is allowed doingthings differently and the
really, at least in Australia,there's been a tendency for
governments to de-skill over thelast decades.
So we have a lot of generalistsin government who are not
technical, who don't get thenext level of detail and who
just think well, you know, thisis what we do, this is how we do
(32:09):
it.
So there isn't a simple answer,but being there to help this
thing, these are elements of howwe get going.
So I hope this podcast, I hopesomeone listens to this podcast
and says, oh, maybe I can dothings differently, maybe I can
look at this differently, butit's very hard to change the
mind of someone.
It's really hard to change themind of a public servant.
(32:30):
It's really, really hard tochange the mind of a public
servant who's got all theirpieces, their program logic laid
out and their budget and theirresources are ready to go.
It's really tough to do.
Speaker 3 (32:40):
We've got other catalysts that I guess are
programs that governments haveextensively committed to, like
Digital ID now and open banking,do you find?
Well, it's a leading question,ian, but we find patterns
through these programs that arecommon and they suggest to us
(33:02):
maybe there's an emergingsophistication about how to
govern data, what the parametersare that are governable, what
can be standardized, what shouldbe decentralized and left to be
determined on a local sort ofbasis.
One of the anxieties I haveabout ISO 8000 is that we had to
(33:23):
see this process or thismovement become too robotic in
the way that ISO 9000 did.
It poisoned the well by makingquality management like an
algorithm or a spreadsheet andit took a lot of the judgment
and intuition and humanity outof quality management.
(33:44):
And you know 30 years and Ithink ISO 8000 is going to do a
better job.
But are you getting to thepoint where you can imagine how
this weird, intangible thing isgoing to be governable in a
general way?
Speaker 2 (34:01):
Thank you, steve.
Once again you've answered theright question.
On the record, I love standards.
I'm a huge fan of standards.
Without standards, we areelectrocuting ourselves, we are
sticking our tongue in the light, we're doing all that dumb
stuff.
There is a tendency forstandards to get a little
brilliantly focused, a littleheavy, and you touched on an
(34:23):
example right there.
I think there are otherstandards that get so cumbersome
that they become less usefulthan they could be, but here's
one I prepared earlier.
We've just finished, aftermultiple years of work at ISO
IEC, so at JTC1 standard on datausage 2.5207, terminology and
(34:45):
use cases.
Believe it or not, defining theterminology, getting the
definitions right, is somethingthat largely didn't exist or
happened in a multitude ofdifferent environments, and we
fought like cats in a bag aboutgetting those definitions.
Because the cloud people thinkthis, the document exchange
people think this, the AI peoplethink this, the smart cities
people think this.
We said let's get that down.
(35:07):
And then guidance for usage,and one of those usages, of
course, is AI.
So we connected to the world ofthe AI folks, we connected to
TAO, we connected to all thisstuff and we said you know, it
needs to be simpler.
It also needs to acknowledgethat there is uniqueness in
every circumstance.
There is some of that judgmentthat's required, but the formula
(35:28):
for these standards is, quitesimply what do we need to know
about the data before we use it?
What do we need to understandabout the data as we assess it
for fitness purpose?
What are the guidance,restrictions or prohibitions we
need to put around the dataproduct?
We've read the thing that I'vebeen sort of slipping into the
conversation to know is exactlywhat's in that standard, and I
think it balances the slippinginto the conversation to know is
exactly what's in that standard, and it acknowledges the fact
(35:50):
that context matters, yourenvironment matters, your
sensitivities matter, yourparticular circumstances about
your data matter, but most of it80% of it, by the way actually
is a repeatable process.
But the 10% is your judgmentand your context and your
environment and you need to dosome level setting about layers
of control very high control,high control, moderate control,
(36:12):
low control, no control.
You get to decide, but this ishow you decide.
This is a recipe for you todecide.
You still make that dish, youstill bake that cake.
It's your cake, but this is howto turn the oven on.
This is the temperature to turnthe oven at.
This is how long to think aboutit should go on the oven for.
But you're still doing thatsubjective confinement to not
(36:34):
make it robotic, because data isreally complicated.
It's so complicated that infact, there are seven degrees of
infinite complexity in datasharing and use.
Seven degrees of infinitecomplexity, data sharing and use
, which we're going into a lotof trouble.
But that basically comes downto every data set can be used
for an infinite number ofpurposes.
They might not be fit forpurpose, but you can use it for
(36:55):
an infinite number of purposes.
Context changes the sensitivityof data, so you bring that data
set close to another data setor you know something about
something in the data.
The context of use changes it.
All those ambiguities, all thosepossibilities, try to create
the things you create.
Whether it's a salary data setor a self-set or an insight,
(37:17):
that same issue exists, thatcontext matters, and you can use
that insight for a range ofdifferent purposes and there's a
couple more steps in there, butit means that it's so mercurial
, it's so difficult, so slippery, it's like smoke, it's like
trying to eat your arms aroundsomething which is really
dangerous.
So if you can flatten all thoseambiguities or uncertainties
(37:40):
into a couple of different types, layers of control, then you
can start to do something usefulwith it.
That's what these new standardstalk about.
It says this is what we mean,these are the definitions and we
mean the same thingconsistently and these are the
ways of thinking about how youuse data.
And therefore it's 5.7,.
(38:02):
Terminal Newscast is 5.212, aguidance for usage, and one of
those usages just happens to beAI.
Speaker 3 (38:08):
Brilliant.
Let's get those details fromyou and put them in the show
notes later on and see how muchof that material is accessible
to our audience.
It's fantastic guidance.
Thanks, Ian.
Speaker 2 (38:21):
It was a lander of love.
A lot of landers and a lot ofissues.
People who try to get theirarms around it start to talk
about things like what's thedifference between data and
information.
That one's easier to dial out,but then information and
knowledge and wisdom.
It was three in the morning.
We were really close to the endof the standard.
(38:42):
Someone said we haven't talkedabout wisdom.
I said no, the ISO definitionof wisdom.
We'll probably understand it.
If not, we're ending thisconversation.
Wait, I love it.
Speaker 1 (38:56):
Well, I think it behooves us to exercise some
wisdom at this point and tobring this to a conclusion.
Ian, thank you very much.
I'm really grateful for yourinsights and your thoughts.
I'm going to be thinking abouthow we shift government minds
(39:16):
when they already have all themomentum in the world for a
particular vision and we'reusing them over slightly.
Speaker 3 (39:23):
Thank you very much.
Thanks so much, ian.
Here's to data and informationand, if we can get, to wisdom.
Thanks for helping us along theway Later.
Thank you.
Popular Podcasts
United States of Kennedy
United States of Kennedy is a podcast about our cultural fascination with the Kennedy dynasty. Every week, hosts Lyra Smith and George Civeris go into one aspect of the Kennedy story.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com