History and other lies | Marleen De Kramer | TEDxUniversityofLuxembourg

"My topic is the importance of admitting that we don’t know everything, and not being too focused on telling the truth. It’s perfectly fine--in fact, it’s good scientific practice!--to say when we know something and when we’re only speculating. Like all sciences, the humanities need to move away from the “black box” model of knowledge creation."

Marleen has been bouncing around between European countries and different disciplines, completing a degree in Architecture at the Technical University of Wuppertal, Germany, a degree in Heritage Science at Queen’s University, Belfast, Northern Ireland and a Marie Sklodowska Curie Fellowship in Digital Cultural Heritage in Vienna, Austria. She’s now working toward a PhD in History at the University of Luxembourg, where she is doing virtual reconstructions of castles.

This talk was given at a TEDx event using the TED conference format but independently organized by a local community. Learn more at https://www.ted.com/tedx

I'm here to tell you

why I don't tell the truth about castles.

You might think it's my job.

After all, we expect professionals

to speak with authority
and give us clear-cut solutions,

and that makes us very, very nervous

because there's so much
we simply don't know about history.

And as a result,

a lot of things have become established
in our collective memory as the truth

simply because someone said it once,

it sounded convincing,

and nobody since has stood up to say,

"Well, we don't know exactly
what it was like,

but it wasn't like that."

Take Greek temples.

Everyone knows they're made
of beautiful shining white marble.

We've seen them that way for centuries,
from postcards to museums,

and that establishes certain
seeing habits in our heads,

where we've seen them this way so much
that anything different just looks wrong.

And yet today we know for a fact

that they were painted
in bright garish colors;

we're just a little unclear
on some of the details.

I've colored this one in myself
in about five minutes of research,

so it's likely to be wrong
in all the relevant places,

and it's still more correct
than the white one.

So why do we continue
to show them in white?

Well, there's two reasons for that:

One is that we as humans like certainty.

And so we would prefer
to be absolutely certain

even if it's the absolute certainty
that we are absolutely wrong

(Laughter)

than to say, "Well,

maybe it could have been approximately,

I think, something like …"

And then there's the fact

that when we're trying to establish
a new truth in people's heads,

we want it to be
the correct truth this time.

But even if we're not
entirely clear on all the details,

that doesn't mean
we can't make a statement.

If you ask me right now what time it is,

I can't tell you,

but I don't have to shrug my shoulders
and just say, "I have no idea."

I know this event is on
from 12:00 till 6:00,

so that eliminates
half the clock right there.

We've had our first coffee break,
we've not had the second,

so it's between 2:00 and 5:00.

I know there were people ahead of me,
and I'm not being told I'm out of time,

so it must be around 4:30.

Is that correct?

I don't know.

It might not be the truth,
but I don't have to tell you the truth.

I just have to know
how correct I'm likely to be

because how correct I am
can be very, very important.

Me telling you it's about 4:30
is pretty useless

if you want to know
whether you can still catch your bus;

and in that case,
we might have to ask more people,

we might have to fill it in
with more clues and so on;

and that's science.

We ask a question,

and then we fill in the unknown
and get more and more precise.

So the scientific method
is pretty well-established:

You ask a question
about the world around you,

you research what
you already know about it,

you design an experiment
to test what you don't know about it,

you gather the data, you analyze them,
and you reach a conclusion;

and that conclusion could be,
"I need more data."

Then you go back,
design another experiment,

run it again, gather more data,

and you get more data
and more data and more data,

and suddenly you're buried in data,

and you're dealing with big data,

where scientists now have this problem

that there's so many data

they can never read them all
in one lifetime.

They have to find new ways
to deal with that.

And then there's me.

This is me.

You can tell I'm not
the kind of scientist with a lab coat,

and my data problem is slightly different.

Basically I'm dealing
with one student's lab report

that they dropped on the floor,

lost half the pages
and then shuffled the rest,

and there's probably a coffee stain
on the relevant bit.

So what I've got is I've got
half a broken castle,

slightly burned,

I've got a legal contract from 1388
that was written by a guy

who managed to spell the name "Arnold"
four different ways in three pages,

I've got some rocks from the village

that may or may not have
belonged to this castle,

I have got a map
that was done by a guy

for whom this was a 10-minute squiggle
in an eighth-year campaign,

a painting that was drawn

about 300 years
after the castle burnt down,

and a book that was probably propaganda.

And of course I could go
look in the archives,

I can get another archaeological
excavation going, and so on,

but at some point, there's simply
no way to gather more data.

And then you expect me to take that

and mash it all up
into the truth about castles?

You want a reconstruction
that's so realistic

it feels like you're really there,

like every little pebble
in the courtyard is just right.

There's a reason that a lot
of sites and museums

don't use the word "reconstruction";

instead, you find a picture,

and next to it, it has the disclaimer
"Artist's impression."

And that doesn't mean
they didn't do any research;

it just means they didn't document
what they researched.

We don't know who they talked to,
which books they read,

which conclusions they drew,

and which other theories they discarded.

Now, imagine for a moment

that we would treat a text the same way.

You go into the museum.

There's a plaque, and it says,
"Author's impression."

The author thinks there might
have been a castle here.

You wouldn't take that very seriously.

So why do we treat text
so differently from models?

It's because we've come to a consensus
on what makes a scientific text,

and it's quite simply this.

When you're writing
a scientific document,

you put in footnotes,

you cite works by previous scholars,

you show your argumentation -

you simply give your document provenance -

because showing you a picture of the truth

isn't going to help you
without me explaining why it's true.

The truth is, all of these
are correct at the same time.

That's the truth,
but it's not a very helpful truth,

because without context,
data are not information.

So I'll give you a little context.

So for a little context,

this first clock shows
the time in Luxembourg,

and the second one
has the time in Tokyo,

the third one is one
of those annoying clocks

everyone had in their kitchens
about 10 years ago

that actually run counterclockwise,

and the fourth one is not a clock,
it's a barometer -

you just wouldn't know that
by looking at it.

So in historic research,

when we deal with images,
we know what to do:

We give those provenance
through metadata and paradata.

Metadata you've probably heard.

Metadata are data about the data.

You can see those when
you're browsing your computer,

and you can see who made a file,
when it was made,

when it was last opened, and so on.

Paradata are slightly more complex.

Paradata are data
that give context for the data,

so like how they were gathered,
how they were processed,

which decisions were made
about them, and so on.

The metadata for this image

would be that it was taken by me
on the first of June, 2017

on a Sony compact camera.

The paradata are that it was
picture 111 in a series of 128

and I took it on my first
research trip to this castle.

And I love to show this picture

because this picture has everything in it
that is wrong with models.

You walk up the stairs in this castle,
you come to the attic,

and there's a big glass box
with this model sitting in it.

And what I love about it

is that there are no data
attached to it whatsoever.

You don't have a scale bar.

You don't have a date
it was made or who made it.

You don't have a date
it's supposed to represent.

There's nothing even to say

that it's supposed to be this castle
that you're standing in.

And if you're talking
about decision-making processes

in the reconstruction,

if you take a closer look
at that center tower there,

it becomes very, very obvious

the size of that tower was not based
on an archeological excavation

or because there was
a foundation there or something.

No, that's the size
of the toilet paper roll they had.

(Laughter)

And so this model makes me happy

because it's everything
I'm trying to avoid.

And I'm not the only person
trying to avoid this kind of thing.

A lot of intelligent people
are working and avoiding this.

There are some hugely
complex systems these days

that go into great detail on data,

metadata, paradata,
how they all relate, and so forth;

and my favorite one
takes about six months to learn.

Now that's bad enough for me
as a researcher,

but imagine that you,
as a museum visitor,

have to go on a six-month
training course

to understand what you're seeing.

So, instead, I have a system
that's just good enough for me.

I simply take my model,

and I tell you which parts are true
and which ones are not.

So probably true is the easiest.

That's the category of things
that I think are true

because they're still there,

so that could be things
like the castle ruins.

Next, pretty close to true,

we have a lot of evidence for those.

So for example,

I was saying foundations,
towers on foundations -

we fill in the gaps
what we have good evidence.

Third stage, extrapolation,
could be true - maybe not.

That's where I'm working
on secondary and tertiary data,

like the maps and images.

And then there's my favorite category -

the stuff that's not really true.

Now, these things I need
to put in my model

because the model would be
missing something without it.

If I didn't put these in,
I would be telling you a lie,

but I have no idea what to really put in.

It's an interesting problem.

So that's things like I know
the great hall had paintings on the walls,

I will never know what exactly
was painted on them,

so I have to make something up,

but if I left them as a blank stone
the way they are now,

that would be making a statement.

And then, of course, I need to attach
my metadata and my paradata,

and tell you why it's in that category.

And finally, I need
to make very, very sure

that you don't only know
why it's in that category

but which part exactly
I'm talking about.

If you remember that clock from earlier,

well, I can tell you for a fact
that it's Friday afternoon.

I can also tell you
with absolute certainty

that sometime in the last two millennia,
we had a castle on this hill.

What I cannot tell you

is whether in that window, in 1548,

we had an archway
and that archway had a stone

and that stone was exactly
312 millimeters wide.

It could have been 317,

but my drawing is going to say
one way or the other.

And that is the really, really
interesting point for future researchers

because if I've told you
I have no idea what was here,

they can use that point to research,
and then they can say,

"Look, we found more data,
and actually you're completely wrong.

It was 483 millimeters."

And I can say, "Hooray!"

because that advances
our state of collective knowledge.

So if I'm doing science properly,

I want people to be able
to prove me wrong.

So that's why I'm not going to tell you
the truth about castles,

and why I make it
very, very clear to you

when I'm just making it up.

(Laughter)

Thank you.

(Applause)