The document discusses the concept of suspense in narratives. It provides definitions of suspense from various scientific theories and compares approaches. The author aims to develop a typology of suspense and model how suspense levels change over the course of a narrative based on predicted conflicts and their resolution. To test his theory, he creates a simple story example and algorithms to generate narratives and calculate suspense scores.
Decarbonising Buildings: Making a net-zero built environment a reality
Modeling Suspense in Narratives
1. Meta-effects of narrative
towards a typology of suspense
Richard Doust
Phd program (part-time)
richard.doust@gmail.com
Supervisors: Richard Power, Paul Piwek
Department of Computing
Open University, UK
Meta-effects of narrative : R. DOUST – p.1
2. What is suspense ?
Just how do narrative structures such as a
Hitchcock film generate the well-known feeling
known as suspense ?
⇒
Meta-effects of narrative : R. DOUST – p.2
3. What is suspense ?
Just how do narrative structures such as a
Hitchcock film generate the well-known feeling
known as suspense ?
⇒
What are the minimal requirements on formal de-
scriptions of narratives such that we can ade-
quately capture these phenomena and generate
new narratives which contain them ?
Meta-effects of narrative : R. DOUST – p.2
4. Why does it matter ?
way to create more ‘lively’ texts from fictional
or technical storylines
Meta-effects of narrative : R. DOUST – p.3
5. Why does it matter ?
way to create more ‘lively’ texts from fictional
or technical storylines
teaching tool for writing and film-making
Meta-effects of narrative : R. DOUST – p.3
6. Why does it matter ?
way to create more ‘lively’ texts from fictional
or technical storylines
teaching tool for writing and film-making
standard measure of suspensefulness over a
wide range of domains
Meta-effects of narrative : R. DOUST – p.3
7. Why does it matter ?
way to create more ‘lively’ texts from fictional
or technical storylines
teaching tool for writing and film-making
standard measure of suspensefulness over a
wide range of domains
essential component of interactive drama
systems
Meta-effects of narrative : R. DOUST – p.3
8. Suspense in scientific literature
Gerrig and Bernardo (1994)
Cheong and Young’s (2006) narrative generation
Cognitive appraisal paradigm (OCC)
Desire-frustration theory (Smuts et al., 2004)
Brewer and Lichtenstein’s approach (1982)
Meta-effects of narrative : R. DOUST – p.4
9. Suspense in scientific literature
Gerrig and Bernardo (1994)
Suspense is greater the lower the number of solutions
to the hero’s current problem that can be found by the
reader.
Cheong and Young’s (2006) narrative generation
Cognitive appraisal paradigm (OCC)
Desire-frustration theory (Smuts et al., 2004)
Brewer and Lichtenstein’s approach (1982)
Meta-effects of narrative : R. DOUST – p.4
10. Suspense in scientific literature
Gerrig and Bernardo (1994)
Cheong and Young’s (2006) narrative generation
Suspense level depends on the number and type of
solutions reader can imagine in order to solve the
problems facing the main character.
Cognitive appraisal paradigm (OCC)
Desire-frustration theory (Smuts et al., 2004)
Brewer and Lichtenstein’s approach (1982)
Meta-effects of narrative : R. DOUST – p.4
11. Suspense in scientific literature
Gerrig and Bernardo (1994)
Cheong and Young’s (2006) narrative generation
Cognitive appraisal paradigm (OCC)
Suspense is conceptualised above all as an emotional
experience that occurs over time: satisfaction or
disappointment for hoped-for positive events and relief
or anguish for previously feared negative events
Desire-frustration theory (Smuts et al., 2004)
Brewer and Lichtenstein’s approach (1982)
Meta-effects of narrative : R. DOUST – p.4
12. Suspense in scientific literature
Gerrig and Bernardo (1994)
Cheong and Young’s (2006) narrative generation
Cognitive appraisal paradigm (OCC)
Desire-frustration theory (Smuts et al., 2004)
Suspense arises when ”we can’t do much about a
future event, when we are powerless over a given
narrative. The classic case of this is when we are
onlookers of a film or a play. We have a desire to
intervene but no possibility to do so.”
Brewer and Lichtenstein’s approach (1982)
Meta-effects of narrative : R. DOUST – p.4
13. Is suspense a unified concept ?
...
Meta-effects of narrative : R. DOUST – p.5
14. Is suspense a unified concept ?
Our working assumption: Yes.
Meta-effects of narrative : R. DOUST – p.5
15. Is suspense a unified concept ?
Our working assumption: Yes.
Our starting point = list of verbally resumed
story events. Recent studies by Speer et al.
(2007) claim that people break narratives
down into digestible chunks in this way.
Can we discover (or uncover) narrative
commonalities across different media ?
Meta-effects of narrative : R. DOUST – p.5
16. Brewer and Lichtenstein’s approach
Psychological theory of narrative
3 major structures account for story
enjoyment: surprise, curiosity and suspense.
Suspense: an initiating event which could
lead to significant consequences for a
character in the narrative.
‘additional discourse material is placed
between the initiating event and the outcome
event, to encourage the build-up of suspense’
Meta-effects of narrative : R. DOUST – p.6
17. The link to existing research
Can we make the Brewer’s notions of narrative
understanding robust ?
How does placing additional material between
initiating and outcome events increase the
suspense felt in a given narrative ?
Meta-effects of narrative : R. DOUST – p.7
18. The link to existing research
Can we make the Brewer’s notions of narrative
understanding robust ?
How does placing additional material between
initiating and outcome events increase the
suspense felt in a given narrative ?
Can curiosity and surprise be formally linked
to suspense to provide a theory of narrative
enjoyment ?
Meta-effects of narrative : R. DOUST – p.7
19. Some theoretical contributions
Definition of a Storybase - used for
generating highly similar stories
An argument in favour of the
‘conflicting predictions’ idea of suspense
First steps towards a general typology of
suspense phenomena
First phase : to model intuitive notions of sus-
pensefulness over a few basic stories, verifying
the results on a sample of readers.
Meta-effects of narrative : R. DOUST – p.8
20. System components
STORYBASE generates a particular story
INFERENCE a set of inferences triggered at each
story step
SUSPENSE HEURISTIC at each story step, we
calculate the new suspense level
Meta-effects of narrative : R. DOUST – p.9
21. Storybase
Starting and Stopping points - all the events allowed
to play the role of first or last event in the
story.
- possible next steps for each
Event links
character in the story
- constrain the choice of events
Causal constraints
given by the event links - can be positive or
negative.
Meta-effects of narrative : R. DOUST – p.10
22. A short example: The Park scene
Meta-effects of narrative : R. DOUST – p.11
23. A short example: The Park scene
start,
Meta-effects of narrative : R. DOUST – p.12
24. A short example: The Park scene
start,
"walker goes to pathA",
Meta-effects of narrative : R. DOUST – p.12
25. A short example: The Park scene
start,
"walker goes to pathA",
"strangeman sees walker at pathA",
Meta-effects of narrative : R. DOUST – p.12
26. A short example: The Park scene
start,
"walker goes to pathA",
"strangeman sees walker at pathA",
"strangeman hides behind bush",
Meta-effects of narrative : R. DOUST – p.12
27. A short example: The Park scene
start,
"walker goes to pathA",
"strangeman sees walker at pathA",
"strangeman hides behind bush",
"walker goes to pathB",
Meta-effects of narrative : R. DOUST – p.12
28. A short example: The Park scene
start,
"walker goes to pathA",
"strangeman sees walker at pathA",
"strangeman hides behind bush",
"walker goes to pathB",
"walker goes to pathC",
Meta-effects of narrative : R. DOUST – p.12
29. A short example: The Park scene
start,
"walker goes to pathA",
"strangeman sees walker at pathA",
"strangeman hides behind bush",
"walker goes to pathB",
"walker goes to pathC",
"strangeman prepares to jump out of
bush"...
Meta-effects of narrative : R. DOUST – p.12
30. Inference
1. Inference of basic events from sensory input.
(A perceived action in the narrative together with an ‘event classifier module’
produces a list of ordered events)
2. Inferences about the current state of the story (or
deductions)
3. Inferences about the future state of the story (or
predictions)
Meta-effects of narrative : R. DOUST – p.13
32. ‘Time-distributed salience’
Salience of a prediction-conflict is ‘spread
over’ or distributed over the relevant
predictions that lead up to it.
All events which could have a bearing on any
of the predictions in the chain become for this
reason subject to extra attention.
Meta-effects of narrative : R. DOUST – p.15
33. The Park scene: Suspense= 0
start
walker in park
walker goes to pathA
strangeman in park
strangeman watches walker at pathA
strangeman hides behind bush
strangeman prepares to jump out of bush
walker goes to pathB
walker goes to pathC
strangeman jumps out and attacks walker at pathC
stop
Meta-effects of narrative : R. DOUST – p.16
34. The Park scene: Suspense=10
start
walker in park
walker goes to pathA
strangeman in park
strangeman watches walker at pathA
strangeman hides behind bush
strangeman prepares to jump out of bush
walker goes to pathB
walker goes to pathC
the bush is near pathC
strangeman jumps out and attacks walker at pathC
stop
Meta-effects of narrative : R. DOUST – p.17
35. The Park scene: Suspense=14
start
walker in park
walker goes to pathA
strangeman in park
strangeman watches walker at pathA
strangeman hides behind bush
strangeman prepares to jump out of bush
the bush is near pathC
walker goes to pathB
walker goes to pathC
strangeman jumps out and attacks walker at pathC
stop
Meta-effects of narrative : R. DOUST – p.18
36. The Park scene: Suspense=14
start
walker in park
walker goes to pathA
strangeman in park
strangeman watches walker at pathA
strangeman hides behind bush
the bush is near pathC
walker goes to pathB
walker goes to pathC
strangeman prepares to jump out of bush
strangeman jumps out and attacks walker at pathC
stop
Meta-effects of narrative : R. DOUST – p.19
37. The Park scene: Suspense=16
start
strangeman in park
walker in park
the bush is near pathC
walker goes to pathA
strangeman watches walker at pathA
strangeman hides behind bush
strangeman prepares to jump out of bush
walker goes to pathB
walker goes to pathC
strangeman jumps out and attacks walker at pathC
stop
Meta-effects of narrative : R. DOUST – p.20
38. Definition of suspense
The suspense level of a narrative depends on
the salience of predicted conflicts between two or
more possible outcomes and on the amount of
story time that these predicted conflicts remain
unresolved and ‘active’.
Meta-effects of narrative : R. DOUST – p.21
39. Two verifiable predictions
1. Suspense will increase as the number of
predictions leading up to the conflict waiting
to be confirmed decreases
2. A narrative can maximise suspense by
keeping ‘active’ predicted incompatibilities
with high salience over several story steps.
Meta-effects of narrative : R. DOUST – p.22
40. Suspenseful narratives
Suspenseful narratives engineer a spreading of
the salience of key moments backwards in time,
thus maintaining a kind of tension over sufficiently
long periods for emotional effects to build up in the
spectator.
Meta-effects of narrative : R. DOUST – p.23
41. Towards a typology of suspense
A mugger waiting for a walker
First-order suspense
to reach a certain point on a path in a park
A plain-clothes policeman
First-order suspense
who is looking for someone in a park
Will the policeman find the
Second-order suspense
mugger before the mugger attacks the walker
?!
Most suspenseful films or novels rely heavily on
a large number of possible predicted outcomes
being ’active’ at any one time. . .
Meta-effects of narrative : R. DOUST – p.24
42. Film plot summaries
Rad Decision: A Novel of Nuclear Power
In 1970, the Fairview Nuclear Power Station is under construction near Brixton, Indiana.
A young engineer named Steve Borden is there. Meanwhile, a Russian named Vitaly
Kruchinkin leaves his beloved wife and becomes an undercover KGB spy in the United
States. Steve Borden eventually becomes plant manager at Fairview Station. The plant
also employs KGB agent Vitaly Kruchinkin, who is developing sabotage plans.
Meanwhile, a Soviet mole has tipped off the U.S. about a spy working at an American
nuclear plant, and the FBI begins to investigate. Then, soon after the horrific events at
Chernobyl, Vitaly Kruchinkin is ordered to sabotage Fairview Station. He is told to make
it look like an accident. FBI Agent Liz Rezhnitsky has learned of the sabotage orders
and is closing in, but soon Steve Borden finds himself battling a crisis which could lead to
a reactor meltdown.
Meta-effects of narrative : R. DOUST – p.25
43. Film plot summaries
1970 d
time1
time2 RUSSIA
<LATERR KGB <SBB Chernobyl
known by VKwife
time5 VK event SB2 becomes manager of FNPS USA state | KGBGOAL | safety & good image for RUSSIA wife of VK
motivates
causes
time6 action VK3 starts to work for KGB state | SB7 | is manager at FNPS FairviewNPS state |SB1 | works at FNPS state |VK2 | has a loving wife state |Wife |left alone
motivates motivates motivates causes
time7 action VK5 goes to USA action SB3 employs VK at FNPS place | FNPS1 | SB works at FNPS FBI employs action |VK4 | leaves his wife
causes
time8 action VK7 starts work at FNPS place | FNPS1 | SB is manager at FNPS state | FBIGOAL | safety & good image for USA LR Mole
motivates motivates
place | FNPS3 | VK works at FNPS action VK8 develops plans to destroy FNPS event Chernobyl nuclear accident state | LR1 | works for FBI state | Mole2 |works secretly for FBI state | Mole4 | works officially for KGB state | KGBZ | has info about agentX? who wants to destroy PowerStationY ?
motivates enables leaks
state | KGB1 | worried about image action Mole3 gets information Z from KGB
motivates action | KGB2 | wants to tarnish image of USA motivates action | Mole5 | passes on Z to FBI
motivates motivates motivates
action | KGB3 | tells VK to destroy FNPS immediately action | KGB4 | tells VK to make destruction of FNPS look like an accident action | FBI1 | receives info Z from Mole
motivates
action | VK9 | plans to destroy FNPS immediately motivates action | FBI2 | tells LR to find information about X and Y in Z
motivates
time9 action | VK10 | sets into motion a process of meltdown at FNPS action | LR1 | starts to investigate X and Y
causes
place FNPS4CRISIS Crisis - start of possible meltdown motivates action | LR2 | finds information about X and Y in Z
action |SB4 | tries to stop meltdown crisis at FNPS action | LR3 | hears about accident at FNPS
causes
action | VK11 | tries to stop SB stopping the crisis action | LR4 | guesses that FNPS has been sabotaged state |SB5 | believes meltdown crisis at FNPS is an accident
motivates
action | LR5 | goes to FNPS
motivates event |FNPSLR | LR arrives at FNPS
FNPSLR
motivates
time10 action | LR6 | tries to apprehend VK
action LR7 neutralises VK
action |SB6 | stops crisis at FNPS
place |FNPS5 | Crisis averted
The end
Meta-effects of narrative : R. DOUST – p.26
44. Arts/Science
Goal: to create a performance piece in a
theatrical setting which...
uses and extends some of the basic ideas of
the theory.
could also result in a published article,
perhaps in two different domains.
Meta-effects of narrative : R. DOUST – p.27