Scapegoat the Visual

I have a meeting with Harvard Professor Max Bazerman today. As most readers will know, Professor Bazerman is world renowned for his research on negotiations and judgment in managerial decision making. He also studies bounded awareness and ethicality, societal decision making, and want/should conflicts. Along with collaborators Don Moore, Francesca Gino, and Lisa Shu, Professor Bazerman’s recent research dives into several aspects of “Moral Luck” discussed previously. In preparation for the meeting I created the following visual to better explain some of my thinking on the topic. I am happy to share it here on http://www.prospecttheory.net/.

Mechanical Research

If the research community at large is not already well aware of it, I recommend taking a moment to check out Amazon’s Mechanical Turk for its data collection potential. Like the historical chess playing “machine” Turk of its namesake, the site works by harnessing people behind the scenes to do bite sized “Human Intelligence Tasks” or HITs. HITs are small tasks still best performed by human rather than machine intelligence and include tagging photos with the proper labels, eliminating duplicate entries from catalogs, writing reviews, and, most importantly for academic researchers, filling out surveys.

As an example, I just executed a trial run of Mechanical Turk for a study I’m working on regarding curiosity for inherently positive or negative information and the circumstances in which subjects are better able to control the satisfaction of their curiosity. I created a HIT out of the control version of the two different curiosity questionnaires and asked that the survey be completed by up to 100 unique subjects for a reward of 25 cents. In less than four hours from submitting the HIT I have 100 responses at a cost of $27.50 (the site charges a small fee for use). Additionally, Mechanical Turk allows you to reject and not compensate any responders who did not complete their HIT satisfactorily. So, as a quality check I mixed in a question that helped ensure the responders were paying attention. A vast majority of the participants correctly answered a question very similar to the following: “If one hundred thousand and nine is greater than nine thousand enter ‘Q’ otherwise enter ‘T’.”

Though a quick, powerful, and cheap way to collect human subject data, Mechanical Turk does appear to have some major limitations. Most importantly I have yet to figure out a way to bar past respondents from answering subsequent altered versions of surveys used in between subject study designs, though as each respondent has a unique ID it is possible for repeat participants to be eliminated after the fact. Additionally, the baseline demographics of the typical Mechanical Turk worker in the subject pool and the self-selected participant factor may require special statistical treatment. Finally, the interface for creating surveys is rather limited so HTML skills are required.

Even with its limitations, at the very least Mechanical Turk seems like a great vehicle to do pilot studies. I plan to use it next on an outcome bias study to see if there is any merit in pursuing research on an alternative theory to Moral Luck written about previously in this blog.

BITH: The Financial Crisis and Action Ambiguity

Behavior in the Headlines: The stock market is down, credit flow is paralyzed, and Americans are expressing fervent outrage at “Wall Street fat cats and greed.” Constituents are demanding that any “bailout” package include stiff punishment for the financial insiders who “got rich on the path to getting us into this mess.” In the framework of the Moral Luck/Scapegoat discussion, we have a bad outcome along with a strong associated judgment that the actions of financial insiders were unethical and a high willingness to punish. Which of the two theories, Moral Luck or Scapegoat, does a better job of explaining the current situation? Our financial crisis has a unique attribute that may provide insight into these two effects – action ambiguity. In this case at least, it seems that Scapegoat prevails.

While the market was going up (good outcome) Americans had very little interest in the activities of the masters of the universe in high finance. We are all rapidly learning more about our dire economic straights but I believe most still have very little knowledge of the specific actions undertaken by the financial insiders at which Americans are now so angered. This is a case of ethical acts in a black box. With a bad outcome people are willing to judge activity as unethical even before they understand who made what actions. Moral Luck suggests an action will look less ethical in light of a bad outcome. Here we have bad outcomes generating a desire to judge Wall Street actors as unethical before we have even identified what specific actions we are judging.

The financial crisis is a messy real world example and not truly an action black box. Obviously some people, including key thought leaders, know more about the specific questionable actions of Wall Street insiders. However, perhaps a similar black box experimental design could be constructed. Introduce a bad outcome. Next ask subjects if someone should be responsible for the outcome. Then introduce an actor who can be logically associated with the outcome. Minimize the description of the action so that is has very little detail and phrase it in a statement that control subjects would find ethically neutral in a vacuum. “The chef mixed the cake.” I predict that given the right kind of bad outcome (one without culturally predetermined judgments of blame or innocence) subjects will assume that someone must be responsible and further be willing to assign some ethical responsibility to whatever logically connectable subject actor is introduced.

The Blame Game: Desperately Seeking Scapegoat

Before moving on to other phenomena, let’s take the topic of moral luck for another spin. To begin with, please observe that there is an inherent assumption in the standard framing of the effect. The assumption is that subjects are judging the moral act itself. As demonstrated in experiments with this framing in mind, subject judges non-normatively rate an act as less moral when there is a negative outcome and more moral when there is a positive outcome -- as compared to control groups that judged the act “without an outcome.” Like an optical illusion, juxtaposing a moral act with different outcomes can make the act itself look different.

As a mental exercise, let us see if we can turn the standard framing on its head. What if the effect is manifest not from subjects judging moral acts in the light of outcomes but instead from subjects compelled to assign blame for bad outcomes? Colloquially, this concept is found in the term “scapegoat” and in the phrase “someone is going to have to take the fall.” In this alternative “outcome as driver view,” the energy/motivation to blame/punish/label amoral is generated by the bad outcome not an aversion to the revisited morality of the precursor act.

Consider bad outcomes in a vacuum. A sweet little old lady with no family or friends loses her poorly diversified retirement savings in the stock market and can no longer support herself. Or alternatively, a flood victim is left trapped on his roof for days and finally, succumbing to the elements and starvation, dies. With these outcomes, one of the very first questions we are compelled to ask ourselves is “who is to blame?” This seems a natural and productive response. We want to know how this unjust situation could possibly have been allowed to happen or even if someone purposefully caused it to happen. The reason we want to know who is to blame is so that we may be better able address a pressing need for support in the case of the old lady (who is responsible for her now?) or respond to similar situations in the future in the case of the flood. This motivation to assign blame exists even though, unlike in the moral luck experiments, there is no preceding moral antagonist identified. If we were to identify a possible antagonist (stockbroker, FEMA official) the motivation to find someone or something responsible would compel subjects to rate an available antagonist negatively to restore a sense of justice.

This scapegoat framing could explain why subjects rate the immorality of actors when there are bad outcomes as more immoral when compared to the control which has “no outcome” but, on the surface, the theory does not explain why situations with positive outcomes are rated as less immoral than the control. Since this is a mental exercise and we are questioning assumptions, let us take things a step further and challenge the idea that the control scenarios truly represents no outcome. Perhaps there is an outcome and a negative one at that – uncertainty.

Most people very much dislike uncertainty. Moral luck experimental control stimuli leave subjects with unresolved scenarios in which subjects can easily envision bad outcomes resulting. This uncertainty and threat of a bad outcome is in itself a negative outcome. An uncertainty outcome is likely less saliently negative than the certain loss of retirement funding or death from starvation in the earlier examples, however, it is still a negative outcome that may generate desire to assign blame.

Finally, if we quickly assume that positive outcomes may generate motivation to assign positive credit or at least (and perhaps more likely) they do not generate motivation to assign blame, then the scapegoat theory would indeed explain the outcomes seen in the various moral luck experiments. The certain negative outcomes scenarios would have its antagonists rated the worst, the uncertainty outcomes rated negatively and next to worse, and the certain positive outcomes would rate best either as a positive act or at least a neutral one.

For this scapegoat theory to have any value there should be additional hypothesis that could be generated which would predict results that differ from what might be predicted using the standard moral luck theory. Here are a few possible ones that pop to mind (some more merited than others):

Negative Outcomes:

* When faced with a bad outcome in the absence of a moral antagonist, subjects will be willing and able to self generate a generic antagonist and assign blame. The worse the outcome the worse will be the morality rating.

* When an antagonist is introduced, even one with weak ties of responsibility, subjects will assign near full blame to this antagonist (similar rating to the subject’s invented generic antagonist).

* Subsequently, when a more clearly responsible second antagonist is introduced in the presence of first, subjects will reassign most of the blame to the second antagonist, improving the morality ratings of the first.

* If morality is associated with an actor, each actor should generate their own rating independent of other actors. If blame is a fixed quantity based on the negativity of the outcome, introducing more antagonist actors will diffuse assigned blame across antagonists.


Positive Outcomes:

* When faced with a good outcome in the absence of a moral (pro/an)tagonist, subjects will have more difficultly self generating a moral actor and assigning credit in the form of a favorable morality rating.

* When a moral actor is introduced, one with weak ties of responsibility, subjects will assign relatively neutral to positive morality ratings for this actor (similar rating to the subjects invented actor if they were able to generate one).


I believe the desire to assign blame in the case of bad outcomes is powerful, so powerful that people will even sometimes personify the natural world to have “a someone” to blame. However, I am under no real illusions. Remember that this is merely a thought exercise and that the simplest explanation is usually the best. The moral luck framing of this phenomenon has been the stuff of philosophy for a long time and the basis for experiments by some of most admired researchers in the field. I had to go through a lot of logic gymnastics to challenge the moral luck assumptions and in the process generated many new assumptions of my own to lay out the scapegoat theory. There are possibly some big holes in the theory and the predictions are still pretty loose. Additionally, there are very likely results generated by actual moral luck experiments that the scapegoat theory does not explain as I’ve only looked at the most basic findings here. Crafting an alternative explanation is enjoyable and it may be interesting to run a few experiments to test some of the new predictions. However, I predict that we will want to stick with the findings of the papers noted in the previous posting.

Moral Luck

In philosophy, the term moral luck is used to describe the morality of an action judged in the light of its uncontrollable and unforeseeable consequences instead of in isolation. Such judgments are not normative. The morality of an action should not be different because of a lucky good outcome or an unfortunate bad end. Questions of moral luck are not new to philosophy but they seemed like fertile ground for experimental behavioral research. In fact I have been working on a rough experimental design to demonstrate the moral luck effect in collaboration with Columbia’s Leonard Lee.

Here is the draft set up:

--------------------------------------------------------

A doctor is visited by a patient complaining of a stomach ache and other vague symptoms. The doctor has a “gut feeling” that the patient may be suffering from Disease X. Disease X is a serious condition and, left untreated, it can reduce the expected lifespan of a sufferer by up to 5 years. The majority of medical experts estimate there is only a 1 in 10,000 chance that a randomly selected person in the population will have Disease X. None of the symptoms of which the patient is complaining are associated with Disease X and two other doctors have already examined the patient and ruled out the disease. These other two doctors believe the patient has a mild form of a flu virus that should resolve itself in a few days.

There is a test for Disease X that is 100% accurate in its diagnosis. Diagnosed early the disease can be cheaply treated with outstanding success; however, the test costs $5,000 to administer and in 2% of cases the test itself results in a serious infection, which also has negative effects on expected lifespan.

The doctor decided to run the test based on his own judgment. [GOOD OUTCOME: The lab results from the test show that the patient does have Disease X which can now be cheaply and effectively treated. The patient may or may not have an infection resulting from the test (2% chance of infection). BAD OUTCOME: The lab results from the test show that the patient does not have Disease X. The patient may or may not have an infection resulting from the test (2% chance of infection).] On a scale of 1 to 5, how moral was the doctor’s decision to run the test?

Very Immoral (1 to 7 scale) Highly Moral


Should an experienced doctor be allowed to make such a decision even if the statistical odds are not in favor of his or her decision?

Yes / No


If you believe the patient’s family is justified in suing the doctor for medical malpractice, what is a reasonable dollar amount that the doctor’s insurance should be expected to pay in compensation? The average malpractice payment at the doctor’s hospital is $50,000.

$__________________


--------------------------------------------------------

There are a number of improvements that should be made to this set up before running it with real subjects but that may not be necessary. In researching the project I ran across two new papers on the subject that already provide quite solid evidence of a moral luck like effect.

Francesca Gino, Don A. Moore, and Max H. Bazerman, “No harm, no foul: The outcome bias in ethical judgments,” HBS Working Paper Number: 08-080, February 2008

Gino, Francesca, Lisa Lixin Shu, and Max H. Bazerman. "Nameless + Harmless = Blameless: When Seemingly Irrelevant Factors Influence Judgment of (Un)ethical Behavior." Harvard Business School Working Paper, No. 09-020, August 2008.

So it looks like we do not get to be the experimental moral luck vanguard. On the plus side I was at least lucky enough to have a wonderful coffee conversation with one of the authors yesterday, Lisa Shu. I look forward to reading more of her work.