"One ought, every day at least, to hear a little song, read a good poem, see a fine picture
and, if possible, speak a few reasonable words. " ~Goethe

~ also, if possible, to dwell in "a house where all's accustomed, ceremonious." ~Yeats

Monday, April 13, 2015

Causality: King Then Queen


"The king died and then the queen died" is a story.
"The king died and then the queen died of grief" is a plot.

~ E. M. Forster ~
Forster "also points out that the difference between the two is causality. Theorists have debated the validity of the distinction since Forster proposed it in the 1927, arguing, for example, that the very temporality of "and then" entails causality (or at least invites the reader to supply it) so that the only difference between the two versions is the explicit naming of the cause in the second. The debate also includes objections to defining plot solely in terms of causality, since many narrative artists build plots on other principles. Nevertheless the debate itself shows that Forster identified four elements of narrative—character (or agent), event, temporality, and causality . . . . Because narrative spells out the specific relations among agents, events, time, and causality, it is capable of explaining phenomena that escape more abstract analyses such as those based on science-oriented ideas of general laws."
"Narrative - E. M. Forster's King And Queen
And Narrative Across The Disciplines

What always strikes me about the "grief" in Forster's second sentence is not only that it introduces causality but that it reveals the queen's emotional state. Narrative requires conflict, and the queen is a conflicted character. She is grieving; and we know what that means: denial, anger, bargaining, depression. She is in conflict with herself, and with forces larger than herself, such as Nature, God and Death. Now we have a plot.

Forster's brief analysis of royal death and grief always makes me think of that classic exercise in cause - effect analysis: the Drawbridge Problem:
As he left for a visit to his outlying district, the jealous Baron warned his pretty wife: "Do not leave the castle while I am gone, or I will punish you severely when I return!"

But as the hours passed, the young Baroness grew lonely, and despite her husband's warning, decided to visit her lover who lived in the countryside nearby. The castle was located on an island in a wide fast flowing river with a drawbridge linking the island and the land at the narrowest point in the river. "Surely my husband will not return before dawn," she thought, and ordered her servants to lower the drawbridge and leave it down until she returned.

After spending several pleasant hours with her lover, the Baroness returned to the drawbridge, only to be blocked by a gateman wildly waving a long, cruel knife. "Do not attempt to cross this bridge, Baroness, or I will kill you," he raved. Fearing for her life, the Baroness returned to her lover and asked him to help. "Our relationship is only a romantic one," he said, "I will not help."

The Baroness then sought out a boatman on the river, explained her plight to him, and asked him to take her across the river in his boat. "I will do it, but only if you pay me my fee of five Marks." "But I have no money with me!" the Baroness protested. "That is too bad. No money, no ride," the boatman said flatly.

Her fear growing, the Baroness ran crying to the home of a friend, and after again explaining the situation, begged for enough money to pay the boatman his free. If you had not disobeyed your husband, this would not have happened," the friend said. "I will give you no money."

With dawn approaching and her last resource exhausted, the Baroness returned to the bridge in desperation, attempted to cross to the castle, and was slain by the gateman.
Drawbridge from the Liebig Collection

A variety of discussion guides are available for studying the motivation behind each character's behavior. Years ago, when teaching "The Drawbridge" as part of a unit on the short story, I asked the class to consider why the Baroness would have risked a visit to this unfeeling lover. One of my students, who was having a tough semester and dealing with a death in the family, shook his head in resignation and answered, "Maybe she thought he loved her." I continue to value his conclusion as one of the best commentaries of all on these conflicted characters. Looking for love in all the wrong places -- sigh -- as is so often the case.


Speaking of causal analysis, a few weeks ago, my older son Ben, wrote to recommend a "decent John Green video on who started World War 1, because it reminded me that I was going to talk to you guys about causality!"

My younger son Sam added: "Good video. I could go check more primary sources and learn more about WWI, but I'm chill with just blaming Russia," inspiring me to suggest that we could blame Queen Victoria.

The Roots of World War I,
Tangled in the Web of Queen Victoria's Royal Offspring

But Ben was ready to present us with some philosophically challenging material -- no glib conclusions! So here goes! Interestingly enough, Ben's narrative begins with a somewhat modified version of the problematic Drawbridge Exercise:

Trying to disentangle correlation from causation is what econ has been all about for the last twenty or so years (formally, this is called identification). It's been dying a little bit recently. People became increasingly focused on super small but well identified problems. The pendulum is swinging back to tackling and thinking about much bigger, but trickier problems. Either way, though, it's good to have a nice framework for thinking about causality. Without understanding what causes what and the mechanisms through which it causes it, we can't hope to design decent policies.

The setting: king says if princess runs away to be with peasant, he'll have her executed. peasant asks her to run away, she does, guy at moat lets down moat. King sends knight to retrieve princess. King tells executioner to cut off her head. He does. Princess dies.

The identification question: what/who killed the princess? For example, did the princess running away from the castle cause her to have her head chopped off?

The policy question: how to save the princess?

Necessary and Sufficient
To say that A is necessary for B means that A must be in order for B to be. Or that B cannot be unless A also is. Or that if A is not, then B is not. Or that whenever B is, A must also be.

(a) A necessary condition for getting a good grade is handing in a term paper.
(b) A necessary condition for being a bachelor is being unmarried.
(c) A necessary condition for thunder is lightning.

To say that A is sufficient for B means that if A is, then B definitely is.

(a) A sufficient condition for getting an A is getting an A on every assignment.
(b) A sufficient condition for being male is being a bachelor.
(c) A sufficient condition for thunder is lightning.

More colloquially, necessity means it can't happen without it, and sufficiency means that with it, we're definitely doing to have it.

We think of an order of causality. We'll start with the strongest form of A causes B. That is, A causes B if:

(1) A is necessary and sufficient for B.
(2) A is non-redundant and sufficient for B.
(3) A is a non-redundant part of a sufficient chain of events for B.
(4) A is an insufficient but non-redundant part of a necessary chain of events for B.
(5) A is an insufficient but non-redundant part of an unnecessary but sufficient chain of events for B.

We call condition (5) the INUS condition, and it's where much of social science is concerned. There aren't a lot of necessary and sufficient causal factors for being employed, or having a successful IPO, or getting a line of credit extension, or graduating from college.

An Example
A good way to become familiar with this is to think through some examples. Did the princess running away from the castle cause her death?

Is the princess running away from the castle necessary for her head being chopped off? No. The King is obviously crazy; he might have had her executed when he found out that she wasn't his daughter because the queen had been sleeping with one of the minstrels because she thought the King was a prick and didn't want to marry him in the first place.

Is the princess running away from the castle sufficient for her death? No,the knight could have refused to go get her and bring her back to the castle. The executioner might have refused to kill her.

So, the princess running away from the castle didn't cause her execution? Well that's not quite right. It doesn't cause it according to the definition of causality.

Is her running away non-redundant? Yeah, see previous example of queenly infidelity.

We might actually say that her running away did cause her execution according to the third definition. Actually, yes. I'm going to say that that's where we get to. Turns out the princess running away did cause her death.

We leave the causal nature of the other story's actors as exercises for the reader.

Designing policy, though, is very tricky. First, it involves figuring out how much each actor caused her death. Something made very, very tricky without huge sample sizes. In this case, there's only one occurrence, so we can, empirically, say very little. Then there's all the logistics of the policy and its unintended consequences. And of those predictable consequences some valuation of their effects has to be made. Also often impossible to do satisfactorily.

New York Times
Kevin Mumford, and many professors since, have recommended going to nyt.com at any time and picking an article. True story.

Here's one about "Refugees in Afghanistan": "Such experiences have become increasingly common for Afghans living in Pakistan after the terrorist attack on a school in Peshawar in December. Though the attack was claimed by the Pakistani Taliban, the Afghan refugees say it fueled a new wave of resentment against them."

Did the attack on the school cause increased resentment towards the Afghan refugees? It's not immediately clear this is the case. Which kind of causality is it, if it is indeed causal? It's clearly not sufficient. The school being blown up by a terrorist group is not definitely correlated with resentment against refugees. What non-redundant link does it serve in the chain? It could be that the refugees have come, consumed resources previously going to Pakistanis, and this hit a threshold that fueled a new wave of resentment. The school attack is redundant in this case. Is the chain of events necessary? As before, nope.

We could precisely identify the causal effects of the attack on the school by comparing the levels of resentment faced by the refugees in the event there is an attack, and again in the event that there's not an attack (this is called the counterfactual). We often estimate the counterfactual by looking at a control group -- we can identify the effects of the drug by comparing what happens to the cholesterol of the treated group as compared to the control group. We can't do this with school attacks for quite a few reasons - expensive, unethical. And even if we could, it wouldn't be a perfect experiment, because it would happen on a different day and the circumstances wouldn't be identical. This is why economics is tricky snacks. Ultimately, we can't prove the attack at the school caused increased levels of resentment.

But we can use a smell test: would the levels of resentment been higher, the same, or lower if the school had not been attacked. That will shed some light on its causal nature.

Thanks to Guest Blogger Ben McCartney!


Next Fortnightly Post
Tuesday, April 28th

Between now and then, read
my shorter, almost daily blog posts

Looking for a good book? Try
my running list of recent reading

My friend Diane always sends the best Valentines!
The King & Queen of Hearts above is from Pier 1 Imports (2006)

No comments:

Post a Comment