Presentation on theme: "Introduction to higher order knowledge: The Omission/Commission Distinction."— Presentation transcript:
Introduction to higher order knowledge: The Omission/Commission Distinction
We will explain a hand full of puzzles using one simple model: -symbolic gestures -omission commission -chemical norms -innuendos All using the concepts of “higher order beliefs” and “common knowledge”
Let’s start with the omission commission distinction.
First, let’s remember what the omission/commission distinction is Recall Batman…
“I won’t kill you…but I don’t have to save you”
Batman commits an act of omission because doesn’t save bad guy when he has the chance. If Batman would have killed him, this would have been an act of commission Notice: Batman’s intention is the same (to kill bad guy) The outcome is the same (bad guy dies) But Batman (and presumably the viewer) thinks omission less bad
Not just fiction. Let’s look at some evidence hat we all make this distinction
Subjects brought into the lab and shown two scenarios that differ only in whether someone commits act of omission or commission In both cases intentions and consequences are clear Here is the setup of one such scenario…
Act of omission…
Notice, you know David’s intentions
Versus an act of commission…
Notice David’s intentions are held constant
Subjects consistently rate acts of omission as less bad E.g., Spranca, Minsk, and Baron (1991) find that approximately 2/3 of subjects rate omission as less bad, and almost no one rates commission as less bad
Not just fiction or the lab. We distinguish between omission and commission in law…
It is illegal for a physician to assist a patient in committing suicide
On September 17, 1998, Dr. Kevorkian administered Thomas Youk a lethal injection Youk's family described the lethal injection as humane, not murder But… On March 26, 1999, Kevorkian was charged with second-degree murder and found guilty Source:
But it isn’t illegal for a physician to turn off life support…
Again, same outcome, but act of omission is treated as (much) less bad than act of commission
The omission / commission distinction shows up in important historical/political events, too…
The Struma, 1941
Decrepit ship with over 700 Jewish refugees, en route to British controlled Palestine, where Jewish migration was restricted. Had to make an emergency stop in Turkey The British pressured the Turks not to let the Jews go ashore … after 10 weeks of negotiating between the Brits and Turks … the (non-functioning) boat was towed to sea and (to no one’s surprise)…
Would the British have torpedoed the ship themselves?
Another historical example…
Many argue Sharon “allowed” the massacre. Would he have ordered his troops to commit the massacre? Was he judged as severely as if he had?
In short, the omission/commission distinction is ubiquitous And plays significant role in decisions as serious as those that affect life and death
Why is omission viewed differently from commission? When is omission viewed differently from commission? Is this distinction something we should legally respect or overcome?
To answer this question, let’s consider the following simple game theory model:
First, David chooses whether to switch the track to save his car and hurt construction worker (i.e. transgression of commission). You and Mark can easily see if the tracks are switched
Then David, Mark, and you play a variation of the indirect reciprocity game, e.g.: 1.Each period, players are randomly matched. 2.When matched, can pay cost c to help by b 3.Discount rate δ
Consider the following “ethical system”: David should not switch the tracks. Start by thinking of everyone as “good” If observe David switching the tracks, consider him “bad” If observe anyone defecting against anyone good, consider him “bad.” Cooperate against everyone who is good. Notice that this “ethical system” is a Nash Equilibrium (for sufficiently high delta and b>c) I.e. if you think others abide by this ethical system, so should you
What if we change this from commission to omission? How should we model this? First, we want to make sure that you are still aware the intentions were bad (i.e. you see David looking back and forth at his keys). But…and this is the crucial assumption…let’s assume that Mark is unlikely to observe this. (that is we presume that acts of omission, EVEN IF the intentions are obvious to you, that’s only because you know some key contextual information like seeing him look back and forth, but its safe to assume that others probably are not always privy to such information).
What if we don’t have “perfect information” about others’ past actions? For example, if Mark can’t see that David was there…
Let’s model this…
First, David chooses whether to keep the track to save his car and hurt construction worker (i.e. transgression of commission). Let’s assume that 90% of the time David is not paying attention. If the train comes and David does nothing, you can see whether David intended to do nothing after hesitating or was just not paying attention, but Mark cannot distinguish. (Notice that when the tracks are switched, both you and Mark can perfectly tell the intention, since the track cannot be unintentionally switched)
Now is there a Nash Equilibrium where you defect against David if he (intentionally) doesn’t flip the switch (i.e. transgresses by omission)? No! (provided 90% is sufficiently high relative relative to b,c,δ) Because Mark will presume that David was likely asleep and you are just defecting to save c…
I.e. knowing that David’s intention were bad isn’t sufficient to cause punishment in repeated games…rather, you must also know that Mark knows David’s intentions were bad. This is called a second order knowledge
In fact one can show that second order knowledge aren’t all that matter…third order knowledge matter too…and fourth order and fifth order and… This is called common knowledge
Notice our logic depended on: -High prior probability of “paying attention” -Punishment is caused by repeated game -The fact that omission requires contextual information to know intentions (You will show all this more rigorously in homework!)
Alright, let’s show the evidence…
First study… Showed M-Turkers the trolley scenario (for now just the omission scenario) Varied whether Mark could see David as well as you (i.e. manipulated “second order knowledge”)
Then asked subjects a series of questions
First, asked two questions intended to elicit moral judgments…
Incredibly, subjects judged David’s action as more wrong when Mark could see David. Even though what they could see hadn’t changed! This shows moral judgments depend on higher order knowledge!
Second study… Always assume only you can see whether David hesitates… Varied whether transgression is omission or commission.
Asked subjects own judgments of David’s Intention, guess of Mark’s judgment of David’s intention and also wrongness.
As required by the theory: Subjects thought Mark would be as certain as self of David’s intention in commission but not omission. This shows that actions of omission and commission differ in higher order beliefs
We also showed that this difference in higher order beliefs can explain part of the omission commission distinction (Statistically, the difference in moral judgments between omission and commission became smaller once we took into account differences in higher order beliefs)
How else can one demonstrate the relevance of higher order knowledge on the omission commission distinction?
When will it matter?
Should it matter?
Where we’re going with this…
In general, when players condition their action E.g., whether to punish someone on some past event E.g., whether that person committed a bad act They can’t condition on any old event, but only on events that create common knowledge, i.e. a certain degree of confidence in others’ beliefs at all orders E.g., won’t penalize acts of omissions but will penalize acts of commission
Next class we will formalize this…
And this will help us explain not just the omission/commission distinction, but a host of fascinating puzzles