Once Upon a Time

The Behaviour Space of Stories

Shawn Graham

Carleton University

shawn.graham@carleton.ca

https://slides.com/shawngraham/dhbenelux scholar.social/@electricarchaeo

Act 1: Story shapes are real

Act 2: ...and we are in them...

Act 3: ...and we can only escape through the generosity of failure

...a story in three acts.

“People think that stories are shaped by people. In fact, it’s the other way around…

Stories are a parasitical life form, warping lives in the service only of the story itself.”

— Terry Pratchett, Witches Abroad

behold! parasites!

Two People

“Tell me,” said Magrat, “you said your mummy knows about the big bad wolf in the woods, didn’t you?”

“That’s right.”

“But nevertheless she sent you out by yourself to take those goodies to your granny?”

“That’s right. Why?"

— Terry Pratchet, Witches Abroad


triples = [
    ['Red Riding Hood', 'lives_in', 'Village'],
    ['Red Riding Hood', 'is_child_of', 'Mother'],
    ['Mother', 'gives_basket_to', 'Red Riding Hood'],
    ['Red Riding Hood', 'visits', 'Grandma'],
    ['Grandma', 'lives_in', 'Forest House'],
    ['Wolf', 'lives_in', 'Forest'],
    ['Wolf', 'meets', 'Red Riding Hood'],
    ['Wolf', 'eats', 'Grandma'],
    ['Wolf', 'disguises_as', 'Grandma'],
    ['Wolf', 'waits_in', 'Forest House'],
    ['Woodcutter', 'hunts', 'Wolf'],
    ['Woodcutter', 'saves', 'Red Riding Hood'],
    ['Grandma', 'is_at', 'Forest House'],
    ['Wolf', 'is_at', 'Forest House']
]

good ol' wikipedia, eh? https://en.wikipedia.org/wiki/Knowledge_graph_embedding

Models I used to create the embeddings:

DistMult - uses diagonals; can't represent asymetric relations well

ComplEx - like DistMult but with complex numbers, can handle asymetric relations

RotatE - models relationships as rotations in complex space. Can handle inversion/composition tool.

<- across 3 dimensions, contrasting between actors and places

<- across 3 dimensions, organizing across domestic/wild divide, and who acts upon whom

<- across 3 dimensions, notice the close overlap of wolf/grandma in dimension 0 and dimension 2

Raw dimension scores (which span roughly −0.5 to +0.5 and can't directly be ternary coordinates) are first min-max scaled per dimension to [0, 1], then row-normalised so each entity's three values sum to 1.

This preserves relative ordering within each dimension while producing valid barycentric coordinates. Entities that score strongly on one dimension cluster near its vertex; entities the model found structurally ambiguous cluster near the centre.

Look therefore for contrasts between the centre and peripheries.

The Map of This Talk

(click to expand)

“A million unknowing actors have moved, unknowing, through the pathways of story.

It is now impossible for the third and youngest son of any king, if he should embark on a quest which has so far claimed his older brothers, not to succeed.

Stories don't care who takes part in them. All that matters is that the story gets told, that the story repeats.”

— Terry Pratchet, Witches Abroad

this slide left intentionally blank

(...but you'll be glad to know we're now in Act 2)

Coopers Hill Cheese Roll 2024. Photo Molly Darlington / Reuters

basin of attraction

forward pass

The model is made of basins and attractors...

...but reinforcement from human feedback constrains your way through.

Generated images of 'Vimy Ridge'

https://electricarchaeology.ca/2024/11/15/what-does-canadian-history-look-like-to-the-machine/

https://www.cyberneticforests.com/news/how-to-read-an-ai-image

this slide left intentionally blank

Adam Kucharski, Real signals or artificial stereotypes?

https://kucharski.substack.com/p/real-signals-or-artificial-stereotypes

Myra Cheng et al.

Sycophantic AI decreases prosocial intentions and promotes dependence.Science391,eaec8352(2026).DOI:10.1126/science.aec8352

Sycophancy

Strong Attractors

a recipe for disaster

let the

ll the

Greed's a helluva thing, eh?

...how to sell an AI system to upper management...

computation is not neutral but reflects the world of its training
the world of its training creates attractors that pull towards particular outcomes
we've only gone and plumbed these systems into everything
which creates a hyperobject whose disparate workings we only dimly perceive unconnected across the world

[link]

Slop is a deliberate choice meant to foreclose the imagining of alternatives.

this slide left intentionally blank

“It takes a special kind of person to fight back, and become the bicarbonate of history.”

— Terry Pratchet, Witches Abroad

(You. You are the bicarbonate of history. And now we're into Act 3).

Drew Breunig's 'Spec Driven Development'

https://www.dbreunig.com/2026/03/04/the-spec-driven-development-triangle.html

Hermeneutic Circle

Worlds, not Stories

see Morgan, Colleen. ‘Archaeology as Worldbuilding’. Cambridge Archaeological Journal, vol. 36, no. 1, Feb. 2026, pp. 92–101. DOI.org (Crossref), https://doi.org/10.1017/S0959774325100164.