Moral dilemma belong to the situations that make good stories

Modelling Interest in
Fictional Narratives Generation of Scenarios
Author : Antoine Saillenfest supervised by Jean-Louis Dessalles
Scientific and economic stakes
Modelling
narrative
interest and emotional
impact in narratives is
of major importance,
both scientifically and
economically
How can we develop a framework and a simple model for
representing and generating interesting scenarios ?
A variety of potential applications
Modelling the Interest
Previous works
see also : www.simplicitytheory.org
Simplicity Theory
Human cognition is sensitive to Kolmogorov Complexity (i.e. the length of the minimal determination of a situation) [1]
Example 2 : Two running nuns (departure from
Narratology
Artificial Intelligence
Structure of
a scenario Suspense...
Simplicity Theory [2] Scenario generation Kolmogorov
complexity [1][4]...
For a situation s
"I met two nuns who were running on the
jogging trail, not far from the convent. I couldn’t
help telling the event when back home."
Interest
World machine
Standard
emotion
H : the most simple causal story that explains
how s could happen
Unexpectedness
01010111010110001101010010001
Neurosciences
Experts
Cognitive
Psychology
Emotions...
Example 1 : Pachinko
: size of the minimal explanation
Emotions Morality [3]...
f : feature "running" ; r : class of "nuns"
Then :
Here :
If s can be considered unique in its class :
Observer machine
01010111010110001101010010001
Finally :
: size of the minimal description
Moral Dilemma Stories
Moral dilemma belong to the situations that make good stories
Saillenfest, A. & Dessalles, J-L. (2012). Role of Kolmogorov Complexity on Interest in Moral Dilemma Stories [6]
A Complexity-based Model of Moral Judgment and Interest
We consider the case in which Tom sacrifice one person to save the 5
who are initially threatened.
Experimental validation
64 individuals - 38 ♂, 26 ♀ (m. 26.11, sd. 7.72)
”According to you, will the readers of the story approve of Tom’s actions?” (-5: ”Disapprove”, 5: ”Approve”)
”According to you, will the readers of the story find the alternative interesting?” (-5: ”Not Interesting”, 5: ”Interesting”)
Moral judgment is defined as the difference between moral evaluations of
desired outcomes and moral evaluations of undesired outcomes .
Adjusted mean moral
approval ratings
The calculation of the moral evaluation of this action (a) knowing its
consequence (s1) is a direct application of Simplicity Theory :
4
4
3
3
2
2
Adjusted mean
interest ratings
The river [...] was flooding one of the two tunnels of the mine.
Tom knew that there were five people in tunnel A. In tunnel B,
there was only one person. [...] The trapped persons were
going to drown. [...] The current cannot be interrupted in both
tunnels at the same time. [...] He stood at the entrance of
the two tunnels, near a crane and a heavy and voluminous
box.
1
0
-1
1
0
-1
-2
-2
1 – Person 2 – Friend 3 – Cousin 4 – Child
1 – Person 2 – Friend 3 – Cousin 4 – Child
Alternatives
5 persons
die
Main effect of the identity of the victim. Interest and moral approval are
significantly lower in case of undefined person
In our example :
Tom acts
1 person
dies
This model makes the following predictions :
- The smaller the causal chain between an action and a negative
outcome, the less approved the action. In particular, actions with direct
negative effects are the least approved ones.
- Actions causing negative consequences involving relatives or family
members are more interesting but less approved
Revisiting classical notions
scenario 1:
"Jill was allergic to the medicine,
Jack didn't know it and Jill died"
Adjusted mean moral
approval ratings
And the Interest of s1 is :
causal
chain
a-s1
Alternatives
4
4
3
3
2
2
Adjusted mean
interest ratings
causal
Tom doesn't act chain
a-s5
1
0
-2
-2
2 – L2
3 – L3
4 – L4
1 – L1
Alternatives
3 – L3
4 – L4
Significant effect of the length of the causal chain on both morality and interest
scenario 2:
"It was an adult-strength medicine.
Jill died (s2)."
Bibliography
[1] Chater, N., & Vitányi, P. (2003). Simplicity : a unifying principle
in cognitive science ? TRENDS in Cognitive Sciences, 7(1), 19–22.
[2] Dessalles, J.-L. (2008). La pertinence et ses origines
cognitives : nouvelles théories. Hermes-Science Publication.
[3] Greene, J. D., & Haidt, J. (2002). How (and where) does moral
judgment work ? Trends in Cognitive Sciences, 6(12), 517–523.
νJack(a) = νJudge (a)
[4] Li, M. & Vitányi, P. (1997) An introduction to Kolmogorov
complexity and its applications. New York: Springer-Verlag. (2 nd
edition)
νJack(a) > νJudge (a)
[5] Norrick, N. R. (2000). Conversational narrative: Storytelling in
everyday talk. John Benjamins Publishing Company.
if si is desired
"Jill was sick, his father Jack gave her a medicine (a) in order to treat her (s 1)".
=> scenario 1 : something undesired happened by accident
=> scenario 2 : something undesired happened by negligence
2 – L2
Alternatives
We can also define various necessity for an action a regarding its consequences (s i):
Example : Mens Rea - Accident vs. Negligence
0
-1
1 – L1
Variation in the identity of the victim
1 - a person
2 - a friend of Tom
3 - a cousin of Tom
4 - a child
1
-1
Standard emotion Causal responsibility Targetting Inadvertance
if si is not desired,
Variation in the length of the causal chain
1 - It stayed across the current...
2 - It was carried by the current and stopped
by the struts...
3 - It was carried by the current, hit the struts,
some struts got broken and part of theceiling
collapsed...
4. It was carried by the current, hit the struts in
the tunnel; beams fell down from theceiling;
they were carried by the current, they were
stopped by other struts, itformed a new dam...
In scenario 1, judge agrees with agent, s2 could not have been
predicted.
In scenario 2, judge disagrees with agent. Agent should have
anticipated in some way that a could cause s 2.
Jack is more responsible of Jill's death in scenario 2 than in scenario 1
Antoine Saillenfest
[email protected]
[6] Saillenfest, A. & Dessalles, J-L. (2012). Role of Kolmogorov
Complexity on Interest in Moral Dilemma Stories, In N. Miyake, D.
Peebles & R. Cooper (Eds.), Proceedings of the 34th Annual
Conference of the Cognitive Science Society, 947-952. Austin, TX:
Cognitive Science Society
www.asaillenfest.com