cdn.media.ccc.de/.../wikidatacon2019-15-eng-Keynote_Questioning_Wikidata_hd.mp4

Hello. Just to check.

Can everyone hear me?

Grand.
I've never understood

why that's such a phenomenon
when people give talks

because if you can't,
what are you meant to say?

(laughter)

But yes, so as said, I'm Os.

I'm a PhD student
at the University of Washington,

where, according to the slide,

I study "Gender, Infrastructure
and (Counter)Power."

I'd ask you all to do me the indulgence
of pretending that

that's some very explicit, nuanced,
thoughtful, academic description

and not just what I write as a catch-all,

because I kind of study
a thousand different things

and fitting them all
into a few words is hard.

But most of the things I study

are around how systems
of knowledge enforce particular ideas

of how the world works,

and particular relationships of power

with a specific focus on gender.

I'm also an ex-Wikipedian.

I spent 15 years as an editor

which is maybe where my interest
in the nature of knowledge started,

and I really can't express
how happy I was to be invited

and how glad I am to be here
with all of you,

but particularly James Forrester

who is probably the only person qualified

to countersign
my passport renewal application,

cause it's running out soon
and I've been trying to work out...

(laughter)

You move to Seattle.
Everything is great.

Then you're like,
"Oh, the UK government

requires me to find an ex-priest,
civil servant, or member of parliament,

who's known me for at least 2 years

and who I can ship paperwork to."

That sounds plausible.

(laughter)

Anyway, but...

So I'm here as someone
who has spent a lot of time of...

a number of years--

which I don't like to think about
because it makes me feel incredibly old--

wrestling with the nature of knowledge

and the idea of knowledge--

to talk to you about

what Wikidata looks like

to someone from my background
and with my research interests.

And I'm not going to spend much time
on the story of Wikidata itself,

because if you're here,
having spent 24 hours

having it brain dumped into you,

you're familiar with it.

It's a big semantic data store

that aims to provide
machine-readable knowledge

in a centralized way.

And what this looks like
is a series of items

with associated properties or statements.

So the item for "apple"
has the property "fruit."

I mean, probably.

It's a Wiki so there's probably
a long-running edit war

of whether an apple is a fruit,

and there's 50 people
running 300 accounts between them,

and it's been going for years,

and at this point,
if you mention the word apple on Wikidata,

you're preemptively banned
as someone who, you know,

is secretly a sock puppet

and running an account on one
or another side of this.

So as a consequence,

it's also a classification system, right?

A way of sorting and organizing the world.

So, objects or people or concepts

are classified as worth
having a Wikidata entry or not.

A fruit or not.

And in each case

a series of criterion apply

to determine the properties
that an object should have,

and the values of these properties

and how the objects
all relate to each other.

So Wikidata is really an attempt to build

a universal classification system.

And classification systems
have been studied pretty extensively.

One prominent work
which I'd really recommend people read

if they're interested in this stuff
is Sorting Things Out,

which is book by Geoff Bowker
and Susan Leigh Star.

And they found that
in an ideal universe,

a classification system,

be it universal
or over a particular domain,

has three attributes.

The first is it operates on consistent
and unique principles.

So, there's a consistent pattern

of what should be in each category
and for what reasons.

The second is all the categories
are mutually exclusive.

And the third is
that the system is complete.

It contains total coverage of
what it describes.

And this doesn't mean
it has to have every single object

that fits into the system.

It just means that in the situation

where it lacks an object

and that object then shows up,

there should be
a consistent mechanism

to work out
whether it should be added or not,

and how it should be described

and so on, and so forth.

There is one small problem
with this which is that:

"No real-world
working classification system

that we have looked at
meets these simple requirements

and we doubt that any ever could."

Or to put it another way,

all classification systems fail.

All classification systems
have gaps and exceptions.

And obviously, the same is true
for all systems, full stop.

Anyone who has ever coded

or simply worked in an environment,

or studied in an environment,
or lived in the world

knows that we've yet
to design a single thing

that we've thought all the way through.

The problem is that when we take a system,

classification, or otherwise,

and put it out into the world

and give it power and authority,
and integrate it into other systems,

that already have power
and authority,

there are consequences
for what happens

when the system inevitably fails,

for how it reinforces or undermines
existing relationships of power,

for how it hurts people.

A universal classification system is,
in another words,

not merely doomed to failure,
it's also doomed to hurt people.

And the way that it is structured

is ultimately a series of ethical
and political choices as a result--

Who do you want to hurt?
How much?

What should be done
when people are injured?

And those choices have real consequences.

And so making these choices
often involves confronting the fact

that there's very rarely a single

simple machine-readable interpretation

of something that's true
for all people throughout all history.

Anything in the universe

has multiple meanings,
and symbolisms, and nuances

to different people in different contexts
at different times.

But designing a classification system
and implementing it,

designing a system that can make a claim

to having consistent principles,

and covering everything it discusses,

inevitably involves
cutting down on this complexity

and making decisions about what
"the" meaning of a thing is going to be,

or what array of possible meaning

should be presented
and in what sequence.

And as a result,

it involves silencing voices
or rendering voices louder.

Again, this has consequences.

And to see what I mean
about this complexity

and context, and reduction,
and the consequences of it,

I'd like to set through some examples
from Wikidata itself.

The ones I've chosen
are all gender-related because again,

gender is both professionally
and personally sort of a key interest.

So, the first that I'll start with
is transexualism

which is described as a "condition

in which an individual
identifies with a gender

inconsistent or not culturally associated
with their biological sex."

Fairly unobjectionable and--

wait, no, it's classified as a disease,

and a psychiatric disease at that.

Now, I know what you're thinking,
which is this is appalling

but actually it's not as simple

as either of these statements
being true or false, right?

They're in a category of sort of,
"true, except."

So, take transsexualism
is an instance of disease, right?

Technically, this is true,

in so far as transsexualism

is the name of an entry

under the International Classification
of Diseases, version 10.

But we should add some complexity
and nuance to that.

So, the ICD
is a classification of literally

everything in the world
that you could have

that was in any way involved at all
in someone's injury or death.

It is in fact illegal to die of something
that is not listed in the ICD.

(laughter)

So it contains kind of a lot of things,

and transexualism is listed in it

so we classify it as a disease

because it's in a classification
of diseases.

So, here are some other things
that the ICD also lists as diseases

that it has specific entries for.

PA80: Shot by accident.

PA40.0: Fell off a boat, drowned.

(laughter)

PA41.1: Fell off a boat,
damaged the boat, and drowned.

(laughter)

PA40.1: Fell off the boat,

didn't damage the boat,

didn't drown,

still died of something.

(laughter)

And finally, QD50: Being poor.

(laughter)

So, if any of you
have ever fallen off a boat,

I'm very sorry but you have a disease

which you should really
talk to a doctor about.

What class of doctor,
I'm not sure.

It might be a psychiatrist.

Who knows?

So you know that's disease, right?

What about health specialty: psychiatry?

Well, that's also true, sort of.

So, psychiatrists are the people

who diagnose the presence
of gender dysphoria,

a disconnect between one's sense of gender

and one's sort of like,
embodied or perceived gender.

But again, context.

For example,
saying psychiatrists diagnose it

ignores the fact
that none of the treatments

are psychiatric.

You might as well list the specialties

as specialization in hormones

or plastic surgery,
or being a personal shopper.

All of these also have some role
in people's life trajectories.

They are not listed.

One other useful
potential factoid by the way,

is that the ICD 10 is actually

the old International Classification
of Diseases,

and the ICD 11 no longer lists
transsexualism at all,

much less as a disease.

But my point here is not that Wikidata
sometimes contains outdated information

or sometimes contains
false information,

it's that the statements
that are constructed from that information

as a consequence of what they leave out

and what the results are,

drop things and add risk.

So, one way of structuring

the information that
that entry contained is:

"transsexualism is a psychiatric disease."

And this leaves out
a lot of complexity,

some of which we've discussed.

But the greater issue is how it interlocks

and resonates with existing narratives,
and existing information.

For example, the idea
of transsexualism is a disease.

Does anyone know why
the ICD stops listing it as a disease?

Well, two reasons.

First is because calling
being trans a disease is not accurate.

It does not meet the definition
of being a disease.

In fact, the only reason
that anything to do with being trans

is still in the ICD is not
out of some objective

like, you know, examination
of biology or psychiatry

but instead purely pragmatism.

That if you stop listing it,

then insurance companies
in places like the U.S.

would stop covering medical care

that is associated with being trans.

And the second is that

the stigma associated
with having something classified

as a disease is substantive,

and when you list transsexualism
as a disease

and a psychiatric one at that,

you tap into really
long-standing assumptions

and false beliefs about trans people.

Assumptions and beliefs
that have a lot of power.

Like, if it's a disease
there must be something wrong

with trans people,
something that people should fix.

And if it's a psychiatric condition

then trans people should
be therapized out of being trans.

In other words, whatever the raw truth
or falseness of the statement,

stripping out its complexity
and contextuality,

lets people fit it into their own notions
of what it means.

And that doesn't end

in a neutral objective
classification system,

it ends in things like conversion therapy,

and it being legal
to beat people to death for being trans

when you find out that they're trans
after you slept with them,

because, you know,
something's wrong with them.

Like why would you
be considered reasonable

to have done this?

So a more accurate framing of this
might be this,

which is hard to fit into Wikidata.

And because we can't fit
that into Wikidata,

and we strip it down,

and we lose all that complexity,

we open up the possibility to, again,
reinforce these really dangerous notions.

So, let's look at another example,
also from gender,

and that is the entry for non-binary.

So, as Wikidata informs us,

non-binary is a range of genders

that are neither exclusively man
nor woman.

And there are some critiques
I have of the "also known as" section,

but that's not the biggest issue here.

No, the biggest issue here

is that at no point does this entire page
make any reference to trans people.

So, if you go to the entry
for transgender woman,

it says, "opposite to transgender man."

And if you go to the entry
for transgender man

it says, "opposite to transgender woman."

If you go to this entry,

it has absolutely no reference
to trans people whatsoever.

There is this complete disconnect
and distinction

between non-binary people
and trans people.

And this might be, seems to be,

a pedantic thing to be concerned about

but it's actually a really useful example
for a couple of reasons.

The first is that how non-binary people
relates to being trans

is really hotly debated.

Individual non-binary people
may or may not identify as trans.

As a consequence, it's really difficult

to make big categorical judgements
about a class of people.

Other people would say that non-binary
people aren't trans,

for whatever reason,
or that non-binary people are trans.

You know, you have to
make a decision at some point.

How are you going
to categorize this entry?

What attributes are you going
to associate it with?

But it's hard to do that in Wikidata

when by necessity
the structure of the platform

is so categorical and so fixed,

that you can't really say like,
for some people these things are related

and for others they aren't,
and it's actually very politically charged

but you should think about it.

There's no objective fact to fall back on.

It's very contextual and complex,
and disputed.

So, how do you fit this in?

Anyone?

But, this reductiveness
isn't just a question of,

"Oh well, we haven't fit all
the information in

so I guess it's not perfect."

Again, it fits into preexisting discourses
and the preexisting world,

and has the potential
to cause very real harms.

There's this very long history

of non-binary people
not being considered trans,

going back to, in fact, the foundational,

sort of medical and academic,
and authoritative works

on what being trans is
and how trans people should be treated.

And what this has resulted in

is non-binary people being cut
out of access to resources--

medical care, community membership,
any kind of support.

In fact until 2013,

being non-binary was not a thing
you could possibly be

while still getting access,
to transition-related medical treatment.

If you were, and you wanted access
you would have to go to your doctor

and consistently lie,
and hopefully get away with it.

So, if you want that diagnosis to happen

so that your health insurance
will cover things

or that your national health service
will cover things,

you could either be a man
or a woman,

and nothing else.

And right now there's a ton of backlash

to non-binary existences

from people who are thinking
that we are a threat,

or something new and novel

when we've been around for just
as long as any other kind of trans person

and just not discussed.

And again, the consequence of this

is that this silence is reinforcing
those preexisting ideas

of being non-binary has nothing to do
with being trans whatsoever,

and it creates and reinforces discourses
that cut people off from care,

and cut people off from community.

And finally, before I stop harping
on things about gender quite so much,

the hijra.

So, according to Wikidata

the hijra are the third gender
of South Asian cultures

and a sub class of non-binary.

Now, here's the thing.

Yes, hijra people fall
outside a simple man-woman binary,

but pretty much zero hijra people

would ever define themselves
as non-binary,

because it just doesn't make any sense.

In a western context,
non-binary people are, by definition,

not man or woman

but as a consequence
not trans man or trans woman.

Hijra includes trans women,

and also includes all intersex people,

all sterile people,
and a large number of gay people

while not including trans men

or people who are non-binary,
and were assigned female at birth.

All of this is really complex

and there are literally books written

on the framework of gender
and how that fits into it.

But the point is
there's not a simple mapping

of western gender notions

to gender notions
in the rest of the world.

Categorizing hijra people

as a subset of non-binary people

ignores the fact that most hijra people
do not see themselves that way,

would not see themselves that way,

and that the definitions of hijra
and non-binary

are completely incompatible.

But again this has the potential

to cause harm.

Because the fact of the matter

is that western notions of gender
are pretty regularly

and over a long period of time
exported to the rest of the world

often by violence.

We have these information systems.

We have classification systems.

We have standards.

We have, historically and currently, wars,

all of which are orientated
around this idea

of the western way of doing things
is the only good way

or is the best way
and the standard way,

and everyone should conform.

And so when we have these big projects
which are trying to fit the world

in to a very westernized idea
of knowledge, because they have to,

because that’s how classification systems
do universally work--

everything has to fit
into one consistent scheme.

It is perpetuating that kind of violence.

So, you could respond
to my concerns and examples,

and rambles with kind of a lot.

One line to take would be,
"Why does this matter?"

Why does Wikidata participating
and validating

or invalidating particular discourses
have an impact on the world?

And the first answer is
it actually doesn't matter if it matters.

It matters that you acknowledge it,

So, right now the default framing
of Wikidata is

we're just collecting all of the knowledge
in a machine-readable form,

but you're not.

You're also making decisions

about what should be included
and what shouldn't,

and how knowledge should be represented.

What complexity is worth representing
and what isn't.

And those are ethical
and political choices,

and framing the project
as simply the result

of a million anonymous,
and interchangeable monkeys

with an equivalent number of typewriters

makes it impossible for us
to have conversations about it.

Wikidata's organizers and users
and funders must understand

that they're fundamentally
making charged decisions

that are not neutral
or objective at all,

and that is not bad but dangerous.

And so, okay, having accepted

that these are ethical
and political decisions,

you could say,
"Well, if people want their takes

on things included,
they should just contribute."

And marginalized communities
do contribute a lot, right?

There's a long history
of queer communities,

particularly, being
very early adopters of technology.

And so people could
just contribute to Wikidata.

Like Hijra people could create accounts
and start arguing

that actually the entry
shouldn't be a subset of non-binary

and so, and so forth.

The problem is that
this is unlikely to help

because they're the minority,

because many of the voices
and perspectives

that are currently silenced,

in the political and ethical decisions
being made,

are those of minorities.

So, I did some number crunching on this.

Wikidata has 20,000 active editors

from a human population
of seven billion give or take,

unless you believe that maths is a lie

and the world governments,
controlled by lizards under the Arctic,

is making everything up.

And there are approximately...
Um hmm?

(person 1) You mean they're not?

(laughter)

Look, I'll be honest.

If living in the U.S.
for the last five years

has taught me anything,

it's that any government assemblage
large enough to try and control

a big chunk of the human population

would in no way be consistently competent
enough to actually cover it up.

(laughter)

Like we would have found out
in three months--

and it wouldn't even have been

because of some
plucky investigative reporter--

it would have been
because one of the lizards

forgot to put on
their human suit one day

and accidentally went out
to the shops for a pint of milk

(laughter)

and got caught in a TikTok video.

(laughter)

So Wikidata has 20,000 active editors--

of whom we will assume none are lizards

in human suits or otherwise--

from a human population of seven billion,

and there are approximately
one million Hijra people in the world.

So if we assume a rate
of equal participation--

setting aside the extreme poverty
a lot of Hijra people live in

and the corresponding impact

on access to things
like reliable internet coverage--

then the combined efforts
of 20,000 Wikidata editors

would have to be overwhelmed
by 2.85 people.

That doesn't seem particularly plausible.

Okay, so then you might say,

"Well, what if we just have
other Wikibase instances

isn't that the whole thing
we're building towards?

You can set up your own Wikibase
with your own perspectives

and your own decisions
about how to classify things,

and what to prioritize,
and what not to.

Make your own site with your own standard
for what constitutes knowledge

and what information is important."

And people could do precisely that.

But the problem is
that Wikidata has a lot of heft behind it

which is why the decisions
that Wikidata makes have so much import.

There's the fact that it already exists.

It has a first movers advantage.

There's the Wikimedia brand.

There's the funding
from places like Google.

There's the relationships
with other institutions.

When the strategic plan for Wikidata

calls for engagement
and integration with museums,

that doesn't just result

in getting more data for Wikidata.

That also results in Wikidata

and the decisions its users make
permeating more of reality,

becoming more of a standard
of how data systems work,

and more of a place that is drawn from
to populate other spaces.

So I keep using this line,
"Not bad, but dangerous"

to describe classification systems
or to describe Wikidata,

and I want to reinforce

that I don't think that Wikidata
is inherently bad.

But I do think that its dangers are vast

and are not being properly attended to.

Just by looking at gender,

we saw three examples,
which I pulled very, very quickly,

of situations where even setting aside

the sort of objective "accuracy"

of the information that
a Wikidata entry might contain,

the information it chooses to contain
and chooses to prioritize perpetuates

or silences particular discourses,
and particular ideas

that have weight in the rest of the world,
that do harm in the rest of the world.

And I picked those examples

not because they're surprising
in any way,

or not because they're unique,

but simply to point out that
if I could find that many problems

with resonances in wider violent systems

in such a tiny sliver of content,

imagine how many others
are lurking out there.

And the goal of Wikidata,

the goal of universal classification

if these dangers are not attended to

could ultimately result,
or will ultimately result,

not in simple like neutral classification,

but imposition.

In saying this is the way
the world works

and if you don't like it

then congrats, you should try
and fit into it.

And I really wish that I had
a sort of simple answer for this.

I don't.

It's one of the advantages

of switching to academia

instead of working
in an engineering department.

You can just show up places

and go, "Everything
is really complicated."

Someone should do something about that.

Could I have a grant please?

(laughter)

But all I can really do

is point you back to
Bowker and Star's conclusion,

which is that this isn't ultimately
about Wikidata,

this isn't a problem with Wikidata

this is that the class of systems

that Wikidata is a part of
has never been done safely

and there is no reason
to think it could be.

And so my call is ultimately

not for a particular change,

or for all of you
to just go home and give up.

It's for the project collectively

and for you all individually

to determine how comfortable you are

with participating and building a system

that makes a claim to universalism,

that makes a claim to neutrality
and truth in data,

when we know that that's neither possible

nor harmless when it fails.

and if you are not comfortable
with that, working to articulate

what other ways of doing this
there might be.

And these could look like, for example,

giving primacy
to those local Wikibase installs.

Saying that ultimately

we need to give individual communities

and individual contexts
and spaces primacy

in defining what matters to them,

and how they wish to be defined.

And the conversation about
which perspective should be included

in some central repository should wait

until we have
the full range of perspectives.

So, that's everything from me.

Thank you, everyone,
for sitting through this.

I think we have about 20 to 25 minutes--

(moderator) 25 minutes for questions,
so, please, plentiful.

Thank you very much.

(applause)

(person 2) Thank you so much
for this wonderful presentation

about the problems inherent
in classification systems.

One of the examples you had
is really cool

from a mathematical point of view,

when you were showing
that transgender male

is the opposite of transgender female--

or transgender female
is the opposite of transgender male

and the opposite of cisgendered female.

That makes cisgendered female
be the same as transgender male,

because opposite of is the same--

if A is opposite of B
and C is the opposite of B,

A and C are the same.

So actually that's a place
where it should be different from

and not opposite of,

and that involves
a lot of mathematical issues

when we go to actually ask queries
of the database,

so it's really important
that you've pointed out things like that.

Yeah, another example
of that which I thought was fun

was transsexualism was defined
in part further down--

which I wanted to include,

but couldn't find a way
of fitting it into the flow--

as the same as sex-reassignment surgery.

Which is unintentionally hilarious

because a diagnosis of transsexualism

was historically a prerequisite
for sex-reassignment surgery.

So it's not so much a chicken
and an egg problem

as the chicken is carrying the egg.

(laughter)

Yeah. So yeah, these--

When we look at Wikidata
and how much it uses mathematical,

or pseudo-mathematical language of, like,

opposite of, distinct from,
in the set of...

Yeah, reality is more complex

than the mathematics
we have to represent it.

I don't have a smart answer there
except to say

that I used to be
a quantitative researcher

and I left,
and there is a reason for this.

(moderator) Next question.

Who raised hands?

I see a hand over there?

(person 3) Hello.

First of all.
Thank you for this presentation.

It was very eye-opening.

I want to tell you,
but first of all--

there's a Wikimedia--
I don't know if you know

about the community LGBT+ user group.

So it's a user group,

and they have this mailing list,

and they discussing actually
the issue of sex and gender in Wikidata,

and there is some proposals made

by LGBT+ people to improve it.

So, but it's not fully done yet.

So, there are some plans,
people working on it.

It would be great
if you want to chime in there

and give your opinion

because I'm pretty sure
you're more expert than most of us.

But I want to give a critique
of this thing that you said

about hijra people that said

out of 20,000 editors of Wikidata,

assuming 2.8 of them will be hijra

and they need to overcome
all of these 20,000 people

but this is not true.

Lots of people, I say assume 20,000 people

are just unaware of an issue.

They are not bigots

or they are not going to actively

not let people do this.

And lots of them would help
if you tell them.

Like, as you [inaudible]
that edits Wikidata,

I have no idea about this issue

and if I knew it
I would have fixed it.

So, yeah.

Yeah. I totally get what you mean.

And I want to be clear that I'm not saying

there are 20,000 people,

many of whom are in this room,

although only a tiny percentage

who are vehement bigots
and cultural imperialists.

Instead what I'm getting at
is the fact

that the consensus model,
and discussion-based model

that the WikiProjects are based on

has a couple of flaws,

and one of the big flaws

is that it assumes that all of the voices
worth representing are there

and are represented
somewhat proportionately.

Consensus started off
as a model in Quaker communities

where literally everyone impacted
by a decision was in the room,

because everyone impacted
by a decision could fit in the room.

And so my point
with this 2.85 number is not to say

you have to argue
with the entire population of Wikidata

every time you want to make any decision,

but instead to say
that the consensus model

and the majoritarian model
of what knowledge should be represented

runs fundamentally into a problem

when the people
who are being underrepresented

are underrepresented.

For another example, and a real one,

Myanmar as a country.

The English Wikipedia claims
that it was called Burma

until a couple of years ago.

And the reasoning for this
was very simple.

The BBC didn't like calling it Myanmar

and a load of editors--

(person 4) [inaudible] completely wrong.

Sorry.

(laughter)

You run into this issue of like...

I know it's not the precise thing,
but it's just...

- (person 4) : [inaudible] it's actually--
- (moderator) I give you the mic, sir.

- Yes?
- (person 4) I'm sorry,

that's just incredibly playing
being ignorant and that...

- Okay. Go for it.
- (person 4) That's an absolute terrible,

terrible mischaracterization
of the political situation in Myanmar.

Okay. Go for it.

(person 4) Anyways, so basically
what it is is that the country--

in the Burmese language

the country can be referred to as
Myanma or Bama.

Yep.

Myanma tends to be a more
formal register

and Bama tends to be
a little bit more informal register

but both are acceptable terms
for the country.

The term Burma came obviously
from the term Bama,

but what happened was

there is no official...

The country was officially referred
to, in English, as Burma

up until 1988-- 1989, excuse me,

when the military government
of the country

basically decided,
the military junta of the country decided

that the country should be
referred to as Myanma.

Ostensibly, this was as an attempt
to make the country name

more acceptable to minorities
within the country.

However, this is a bit
of historical revisionism

because Myanma and Bama
specifically refer

to the majority ethnicity in the country.

So, it was basically the government
of Burma at the time--

trying to make the people
equivalent to the country,

therefore implicitly saying--

(person 4) Almost the opposite,

but in a really weird way.

They basically declared that
Bama was in reference to the ethnicity

and Myanma was in reference
to the country,

when historically they both
represent ethnicity

and the country.

That makes sense.

(person 4) But what happen was
because Democrat advocates

within the country
believed that the military junta

did not have the power

to be able to change
the name of the country

in any language,

because they were not
empowered by the people of the country.

and were explicitly
a military junta that they...

therefore the country should continue

to be referred to Burma in English.

Because of the fact that essentially
to call it Myanmar is essentially to say

the government of Burma and Myanmar
at the time was legitimate.

After the fall of the-- well not fall,

but after like the semi return
of civilian government in 2014,

this question came up,

"Okay, should we call this country
Burma or Myanmar in English?"

and essentially,
the facto leader of the country,

Aung San Suu Kyi,

said that there's nothing
in the Burmese constitution

that says you know,
what you should call it in English

so call it whatever you want.

I mean the name of the country

is officially the Union of Myanma
in Burmese,

but as far as in English
you can call it whatever you want.

But generally before the return
of the civilian government in Burma,

to refer to it is as Myanmar
was essentially

to legitimize the military government.

And so therefore,

to call it Burma was generally considered
to be a specific political act

to not give that government legitimacy.

Yeah. So, I'm not saying that
that isn't a rationale for it.

I'm saying that
on the English Wikipedia specifically,

the page went through seven requested
move discussions

over four years
and a mediation cabal decision,

and an attempted structured mediation,

and a review of one the closures
of the move discussion,

and that when you look
at the discussions,

most of the sort of argument
back and forth

is not about
the nuanced political situation

of the country

but it's instead about
what is the common name in media sources

and what do
different institutions call it.

And that when you look at the discussion,

you can see a clear point
where pretty much every news organization

that isn't the BBC
in the English Language,

that's considered like a major
western news source

has switched their language sources,

and the debate
essentially becomes a debate

of whether we should listen
to the Wall Street Journal

or the BBC.

So the point I'm making
is not about the specific politics

of the situation, but instead the fact

that it's really easy for those decisions

to actually become almost a proxy dispute
of how much do we love the BBC,

and that when you look at the discussions

you see this really nice case study

in the issues of having
those conversations

and having those nuanced,
and often insider perspectives

when most of the discussions
are centered around

how much we love the BBC

and are coming from people
who are outside the context.

So, it's not--

My point in all of this is basically

that even if you're not fighting
20,000 people,

even if you're only arguing
with 20 people,

probabilistically, 19 of them

are going to be people
who have very strong opinions,

who don't necessarily bear
any negative consequences

of whichever change happens,

but have a particular world view
and have decided to stick in it,

and so the proposals
by the LGBTQ+ group

to change the Wikidata criteria

might be amazing, I might love them,
I might not love them,

I haven't read them.

But the base premise of this is...

We got the people who show up
on Wikidata right now,

and those are the representatives
of all queer people

and this is the universal rule
of what should be done

with the content of all queer people

is almost a microcosm
of the same problem.

- (moderator) We have another question.
- Yep.

(person 5) Hi.

I think there's another problem

with the consensus-based approach we have,

is that sometimes we have consensus

on really difficult issues
on how to deal with that

and [inaudible] that on Wikidata,
and nobody is reading the discussion.

Typically, the project Names,

which is a really, really old
WikiProject on Wikidata--

and names are a really,
really complicated issue in the world.

Not every people of the world
have a given name,

not every people have a family name,
not, well, you have an idea.

And there are so many
writing systems out there,

and we have, actually, a system

which was working
for many cases in the world

on how to use properties,

what items should look like,

how to link these together
and everything--

We have eight pages--

nobody is reading that,
and someone just added

Latin script family names
to a Chinese researcher.

So, we don't have the names
of these researchers

but we know for sure
that the value added was wrong.

I don't have the correct value,

but I know this one
is not the correct value.

And it's not just discussing the issue

because we have big discussions

and we have actually modeling

which is mostly working on
and even qualifier on things to deal

with more complicated cases

but people are just,
"Oh, given names suggest a property,

I will just add that."

- No.
- Yeah.

I think it's not just
how to model thing,

it's really how to explain
to people the model,

and that's a technical part--
we could have tools with suggestions

and I think the constraint thing
which went live last year

is a great thing for that.

But even when we know to model thing,

it's how to make
this model known to people.

That's a bit technical issue
on how to do that better.

(moderator) So, there was just remark.

There's no real question for you?

Or that's a question to you?

- How to do that.
- (person 5) Yeah, it's a question.

(person 5): Sorry,
even if we have the discussion,

(moderator) Yeah, sure.

(person 5) My question,
if I was not clear, is that

even when everyone is in agreement

on how to model complicated cases,

how do we make technically
the model known for project

with the scope of Wikidata,

so people are not adding
the wrong value in good faith?

Because our problem is both.

We have trouble
modeling complicated realities,

and we have trouble explaining
to users, how to follow the model

we actually have.

Yep.

I will say that
if I could solve that problem

which is to reframe it,

how to reliably and consistently
enculture new users

into having the same view
and understanding

of the project space,

then they would let me graduate

and also give me a job.

It's the second oldest problem
in internet spaces is how to do that.

The oldest problem is writing a system

that will automatically detect insults.

I will say that...

You can look back at Wikipedia,

or before that,
there was the phenomenon

of eternal September on Usenet

which was, "Oh these people keep--
AOL disks have gone everywhere

and now there's newcomers

all the time who don't know
how things work around here,

and everything is drowning
in people hitting "Reply All."

Generally speaking,
the place that I would look for that

is there is a discipline called,
"Computer-supported collaborative work,"

and one of their big questions

is this question of onboarding,
and of like...

making the culture known to people.

But it may not be something
that is directly solvable,

or that we want to directly solve, right?

So, Susan Leigh Star
who wrote Sorting Things Out,

one of her other contributions

was generally the study of infrastructures

of which I would argue Wikidata
is definitely one,

and of the things that she argued

was that infrastructures
make themselves known

through using them.

So like, basically the only way
to work out how a system works

is to engage with it,
and trip over, and fall flat on your face,

and learn not to fall over that way again.

And I think everyone everywhere,
including new users,

including people
coming from other projects,

wants a way of approaching this
where they don't have to fall over.

But I'm not sure if that exists,

and I think that a better place
we might look is maybe to ask

what are the consequences
of people screwing up

and how do we make screwing up
an understandable

and a more expected component
of the user experience.

(moderator) Okay thanks.
Next question.

(person 6) Thank you.

So, first, thank you very much
for your presentation to us.

Again, someone said, eye-opening.

I was looking at the specific item
on transsexualism,

and it's actually even more interesting

because I was looking
at different Wikipedias,

how they dealt with the issue.

And I just look at three.

So, apparently, what
we are seeing on Wikidata

actually reflects pretty much
what happened to some extent

at some level on English Wikipedia,

whereas if you look
at Portuguese Wikipedia,

the actual item connects to transgender,

and on French Wikipedia
it connects to trans identity

whereas transsexualism is a redirect
in both Portuguese and French.

And I was looking at the history
of editing on the Wikidata item,

and if you look at--
there were several sort of wars

but the discussion page
is actually only one line,

but there were several conflicts
between editors,

particularly with the French

that were opposing
the use of transsexualism.

If you look at the names of the items
on each language,

the only one on which
you don't have transsexualism

is French for trans identity,

and then someone came,
and did what you said about

it's the opposite [inaudible],
trans identity,

and then there is a different item that--

Oh yeah.

(person 6) So, it's a complete
global fight over...

basically it's reverberating conflicts

that are apparently also

the manifestations of conflicts
that happen on each Wikipedia.

Yes, that also reflect conflicts
in local cultures,

and in different parts
of the world, yeah...

And I'd argue that, I mean,

I'm British so I have a tendency
to say, "Wait, fighting with the French?"

"Yes, Please!"

(laughter)

But I'd say there's almost something
more fundamental than that,

and you can make an argument
in the other direction.

I can, as a trans person, make an argument
in the other direction and say,

"Actually, it's the French
and Portuguese who have it wrong."

Because the actual question is

is the entry transsexualism about

the medical classification,
or the state of being,

or the historic medical classification,

or the historic term
for the state of being,

or are these different entries,
or the same entries?

When are things distinct enough
to be different objects,

and how do we negotiate that fight

between people who think
that the medical status

and the identity are the same thing,
or different things.

But yeah, there is no easy answer

but yeah, I suspect
if you look at a lot of these examples,

and if you look
at a lot of controversies,

generally on Wikidata

what you're going to see is
these fights over...

These almost negotiations

are the local community norms,

and beyond that are the cultural norms.

Which is a problem because again,

when we're talking about marginalized
or minority groups,

we would expect them to also
be marginalized within Wiki communities,

and also within Wikidata,

and so Wikidata is sort of...

building on these
preexisting prioritizations

of whose knowledge matters,
and under what circumstances

and in what form.

(person 7): I wanted to touch
on something you mentioned.

Everything is complex
and I think modeling it right,

getting it right on Wikidata

is not the sum of the issue.

As you said, Wikidata is infrastructure,

and as [Hermione] said,

we have gotten it right perhaps
in some things, in some other topics,

and still can't
actually practice it right.

Yep.

(person 7): So I want to suggest that

this is a prevalent condition
of the human race.

And however well we model something,
even if we model gender

ten times more complexly
than we do today,

most SPARQL queries involving gender
would not bother

- with the qualifiers right?
- Yeah.

And would still generate very,
very flattened, very simplified results.

Google's use of our data
in the infamous Google infoboxes

will also flatten the data
and ignore qualifiers.

That is not going to change.

Wikidata will continue to be used
in simplistic ways.

Indeed, the majority of use,

probably, will be that simplistic thing.

My point is, it's probably not fixable

and we shouldn't stop trying.

I mean we should try to get it right

and understand that a lot of the use is,
despite our best efforts,

going to be simplistic and wrong.

Yep. I would agree with that.

I guess I would say that

you know,
it's not about like,

my issue here is not about
it being you know,

there is one true
incredibly complex answer.

At some point I just gave up

even in my thesis which is about
transness and technology

of defining transness.

I just gave up.

And I instead took what is referred to
as a pragmatist view,

which is basically that
it is whatever the people

in the situation that you're studying
believe it to be,

and however they construct
the world as if it were,

and what I'm getting at this

is not that there is
some universal definition

of anything which,
if sufficiently complicated,

would be enough,

but instead that I think
that the scale is the problem,

and the universalism is the problem.

Maybe we should keep trying,

or maybe we should stop.

Maybe we should instead say
that, again,

there should be a Wikibase install
in every self-defined community

that wants it and they can define things,
and articulate things

to their own satisfaction.

But then we end up in more political

and fraught debates of a reformist
versus radical actions,

and how you open a box
with a crowbar that's already inside it,

and I end up quoting Foucault for an hour,

and everyone gets sad.

Including me because I hate Foucault.

So this might be a discussion
for elsewhere.

But generally agreed, I just--

I would raise questions about

whether we should keep trying

for a better form of universalism,

or whether the problem
is that universalism.

I'm guessing we have
a time for one more? Yeah.

(person 8): This is a short question,
possibly complex answer.

One of the most popular

and used properties is sex
or gender on Wikidata.

Could you speak to whether you find

that merging useful,
productive, problematic?

Sure, I mean I think it's always

going to be reductive
cause it's a merging.

But I also think
that it is deeply tiresome

in a way that's kind of interesting

insofar as it reveals
the limitations of Wikidata,

though Wikidata claims to be building

towards this like big objective
set of knowledge,

but ultimately kind of
smushed these things together

because I mean they haven't asked

most people who have entries
what their gender is,

and/or what their sex is,

and so they just merge them

so that inference is easier.

But generally speaking, yeah,
I say that the merging

of the two together
is reductive and dangerous

but...

Again it's not...

There is no good way of doing it.

I think this is a particularly bad way

of treating them
as interchangeable things,

and treating them
as forever-linked things,

but I can't suggest a better way
that remains--

that continues to have Wikidata

even tracking this information
or the information contained

in that at all.

(moderator): Okay.
I think we have to conclude here.

I still saw some raised hands

so hopefully you'll be around.

Yeah. I am a grad student.

I have functionally no life, so...

(laughter)

(moderator): Perfect. Okay.
So please come and talk.

Thank you very much.

(applause)