A question about causality versus control in making judgments

[From Dick Robertson,200913.09.1108CDT]

Daniel Kahneman in his intro to his (Nobel winning) book, Judgment Under Uncertainty, gives an example where Ss ignored the unbalanced frequency distribution of engineers and lawyers when stereotypical personality descriptions were given and Ss asked to guess the probability of a given S being one or the other. They took frequency correctly into account when that was the only datum.
This and many more “findings” are offered to support the cognitive-causality view of judgments. My question is, how would you apply a PCT analysis to the question?

Best,

Dick R

[From Rick Marken (2009.05.13.1120)]

Dick Robertson (200913.09.1108CDT) –

Daniel Kahneman in his intro to his (Nobel winning) book, Judgment Under Uncertainty, gives an example where Ss ignored the unbalanced frequency distribution of engineers and lawyers when stereotypical personality descriptions were given and Ss asked to guess the probability of a given S being one or the other. They took frequency correctly into account when that was the only datum.

This and many more “findings” are offered to support the cognitive-causality view of judgments. My question is, how would you apply a PCT analysis to the question?

I’d have to get a bit more detail on this but, in general, I think the first step in a PCT analysis would be to try to figure out what each subject is controlling for when he or she guess the probability. If the data represent averages over subjects then there might be a problem; some subjects might be controlling for providing a probability that is a match to the value based on doing probability calculations that take the prior probabilities (which I believe is what the “unbalanced frequency distribution of engineers and lawyers” are); others might be controlling for matching the probability to the stereotype, or whatever. I think the main goal of a PCT analysis would be to design tests that would make it possible to determine what a subject is controlling (or trying to control) in these situations. The methods used to do this would probably involve allowing subjects to adjust the probability estimate (DV) continuously to compensate for changes in the prior probability distribution (the disturbance or IV).

Maybe you could post an example of one study, the results obtained and the interpretation of the results. It would be nice to see what people are getting Nobel Prizes for these days;-)

Best

Rick

···


Richard S. Marken PhD
rsmarken@gmail.com

[From Bill Powers (2009.05.13.1313 MDT)]

Dick Robertson,200913.09.1108CDT --

Daniel Kahneman in his intro to his (Nobel winning) book, Judgment Under Uncertainty, gives an example where Ss ignored the unbalanced frequency distribution of engineers and lawyers when stereotypical personality descriptions were given and Ss asked to guess the probability of a given S being one or the other. They took frequency correctly into account when that was the only datum.
This and many more "findings" are offered to support the cognitive-causality view of judgments. My question is, how would you apply a PCT analysis to the question?

Are there more lawyers than engineers? Is "ignore" the right word when you don't know something? Or did he tell the Ss how many of each there are?

I don't know of any basis for making probability judgments other than frequency, but if they're the kind that most people make, I'm sure they don't involve any calculations, even if you tell them what the numbers are. In my experience, it's most common for people to decide first what they think will happen and then look for evidence to show how right they were.

Best,

Bill P.

[From Dick Robertson,2009.05.13.1806CDT]

What he said in the book: "One of the factors that have no effect on representativesness but should have a major effect on probability is the prior probability, or baseratefrequency, of the outcomes In the case of Steve [a made-up person with a made-up personality description of, ‘very shy, withdrawn, invariably helpful, but with little interest in people, or in the world of reality. A meek and tidy soul, he has a need for order and structure, and a passion for detail.’] How do people assess the probability that Steve is engaged [as a] farmer, salesman, airline pilot, librarian or physician? In the representativeness heuristic the probability that Steve is a librarian, for example, is assessed by the degreee to which he is representative of…the stereotype of a librarian…people [note that word, RJR] order the occupations by probability and by similarity in exactly the same way…[leading to serious errors, because similarity …is not influenced by several factors that should affect judgments [like, the proportions of librarians to e.g. those of the various other groups in a simulated population.]

Further down he makes the point that if the only information was of proportions in a population, then Ss would take that into account in guessing what occupation a made-up target would have.

I am going to try to offer a paper to this symposium contrasting the different methodologies implicit in causal verus control studies of the presumed topic of interest – economic psychology – simply as one more attempt to find that one in ten or twenty people who would want to learn about PCT. I can use all the help I can get. The first thing I plan to say is about that tendency to talk about what “people” do based upon median or mean scores or counts in samples of n size – versus a “testing specimens” approach. There I have a little solid ground in the “Testing the Self” paper. After that I can only report what you other guys have worked out about economics and so on. I don’t even know how to compare the apples and oranges of decision making (like, Should I buy this stock or leave it alone? as the kind of question I presume the symposium members want to study to compare a purely rational method against ‘subjective and emotional biases’ in deciding) – against how you would do it from the point of view of starting with whatever a person is controlling for. Please forgive that tortured paragraph; I couldn’t figure out how to draw the contrast that I want to make in a simpler way.

[From Bill Powers (2009.05.13.1313 MDT)]

Dick Robertson,200913.09.1108CDT –

Daniel Kahneman in his intro to his (Nobel winning) book,
Judgment
Under Uncertainty, gives an example where Ss ignored the
unbalanced
frequency distribution of engineers and lawyers when
stereotypical
personality descriptions were given and Ss asked to guess the
probability of a given S being one or the other. They took
frequency
correctly into account when that was the only datum.
This and many more “findings” are offered to support the
cognitive-causality view of judgments. My question is, how
would you
apply a PCT analysis to the question?
BP
Are there more lawyers than engineers? Is “ignore” the right
word when you don’t know something? Or did he tell the Ss how many of
each there are?

See above

I don’t know of any basis for making probability judgments other
than > frequency, but if they’re the kind that most people make, I’m
sure > they don’t involve any calculations, even if you tell them what
the > numbers are. In my experience, it’s most common for people to
decide first what they think will happen and then look for evidence to
show how right they were.

So how describe what a person would be controlling for in any of the various versions of Kahneman’s experiments?

Best,

Dick R

···

Best,

Bill P.

[From Dick Robertson,2009.05.13.1838CDT]

[From Rick Marken (2009.05.13.1120)]

Dick Robertson (200913.09.1108CDT) –

Daniel Kahneman in his intro to his (Nobel winning) book, Judgment Under Uncertainty, ETC.
RIM
I’d have to get a bit more detail on this but, in general, I think the first step in a PCT analysis would be to try to figure out what each subject is controlling for when he or she guess the probability. If the data represent averages over subjects then there might be a problem; some subjects might be controlling for providing a probability that is a match to the value based on doing probability calculations that take the prior probabilities (which I believe is what the “unbalanced frequency distribution of engineers and lawyers” are); others might be controlling for matching the probability to the stereotype, or whatever. I think the main goal of a PCT analysis would be to design tests that would make it possible to determine what a subject is controlling (or trying to control) in these situations. The methods used to do this would probably involve allowing subjects to adjust the probability estimate (DV) continuously to compensate for changes in the prior probability distribution (the disturbance or IV).

Maybe you could post an example of one study, the results obtained and the interpretation of the results. It would be nice to see what people are getting Nobel Prizes for these days;-)

Good, I’ll look for a specific example. Maybe I’ll have to dig out original reports, because so far (I’m just into the Intro chapter) Kahneman is summarizing, “people do --so and so – as shown by J xyz, 1999 v n etc.”

Best

Dick R.

···


Richard S. Marken PhD
rsmarken@gmail.com

[From Bill Powers (2009.05.15.0804 MDT)]

Dick Robertson,2009.05.13.1806CDT --

The first thing I plan to say is about that tendency to talk about what "people" do based upon median or mean scores or counts in samples of n size -- versus a "testing specimens" approach. There I have a little solid ground in the "Testing the Self" paper. After that I can only report what you other guys have worked out about economics and so on. I don't even know how to compare the apples and oranges of decision making (like, Should I buy this stock or leave it alone? as the kind of question I presume the symposium members want to study to compare a purely rational method against 'subjective and emotional biases' in deciding) -- against how you would do it from the point of view of starting with whatever a person is controlling for.

In my economic modeling efforts I never even asked how people make decisions. You're talking about gambling, which as we all know now can totally mess up the economic system, but not about the basic structure of an economy and how people carry out their functions to make it work.

PCT economics probably belongs to what is called "agent-based economics." That is, it's about the people who interact within the system rather than about the abstract system itself.

This is from a web page concerned with "Agent-based computational economics" or ACE:

"Thus, agents can range from active data-gathering decision makers with sophisticated learning capabilities to passive world features with no cognitive function. Moreover, agents can be composed of other agents, permitting hierarchical constructions." ACE: A Completely Agent-Based Modeling Approach (Tesfatsion)

I used only actual people as the agents.

As an example, I used a manager in a manufacturing company whose job it is to maintain the inventory of finished goods at some specified level. The means of control I gave the manager was very simple: he could vary the selling price of the goods. If the inventory was decreasing he would raise the prices, and if it was increasing he would lower the prices. This took care of the supply end of the law of supply and demand. On the other end I had consumers who tried to obtain for themselves as many goods as they wanted and needed. If prices were low they could raise their reference levels for discretionary consumption, and so on.

I was working on introducing the source of consumer money as wage and capital income, which was simultaneously a cost of production for the producer. Needless to say this was only a beginning, just enough of a model to show behavior with the various agents interacting with each other.

In this kind of model nobody is making "decisions" and everybody is simply controlling variables. Of course decisions appear to be made -- you can always interpret a control action as resulting from a choice being made to do that action instead of something else. But in the model itself no decisions are made. People vary their actions to correct errors when they occur, and that's all.

Also, my model contained no financial manipulations -- no borrowing, no stocks and bonds, no futures trading, and so on. Not even any banks. All that was going to be introduced later, to show the effect when that sort of superstructure was added. I was out of my depth by that time and looking for collaborators.

The basic difference between PCT economics and existing schools is probably the model of economic man. In PCT people try to satisfy reference conditions, which means there is such a thing as "enough" of something. Economists generally seem to assume that people will consume as much as possible until they explode, if the price is low enough or their income is high enough.

Best,

Bill P.