[“Adaptive Thinking: Rationality in the Real World”, Gerd Gigerenzer, Oxford University Press, 2000]
In Part II, chapters 4-6, Professor Girgenzer provides a few arguments towards a view of what he labels ‘ecological rationality’ which he defines as follows:
Ecological rationality refers to the study of how cognitive strategies exploit the representation and structure of information in the environment to make reasonable judgments and decisions. (page 57)
While this is reasonable, it doesn’t go quite far enough. He seems to endorse a greater truth when he quotes Roger Shepard:
We may look into that window [on the mind] as through a glass darkly, but what we are beginning to discern there looks very much like a reflection of the world. [“Mind Sights”, R.N. Shepard, Freeman, 1990]
Allowing for differences in the meaning assigned to the term world, it’s nevertheless clear that some psychologists are starting to see what St. Thomas Aquinas saw 750 years ago: the human mind is shaped in response to its environments. But this shaping process is actually two-fold. The process of the shaping of an individual human mind to its immediate environment is possible because the human species has been shaped to specific capabilities, shaped by the responses of our ancestors to their environments and shaped to be able to respond to those sorts of environments or to environments ‘close enough’ in some sense. Professor Gigerenzer is, in a sense, trying to define what ‘close enough’ means in the specific case of analyzing uncertainty.
This is a difficult task to show the mismatch we can create between our natural ways of thought and our presentation of data which is mostly based on models useful for highly-trained academics — even when that data is presented in a daily newspaper. It’s this mismatch that he sees in the results of many psychologists and other researchers who seem to have found solid evidence that human beings don’t think in well-formed rational ways, at least not when it comes to numbers, and specifically that human beings don’t think in Bayesian ways. For those who don’t know much about formal statistics, Bayesian analysis allows us to answer a question like: what is the probability that a woman has breast cancer given a positive mammogram, which might be due to the presence of cancer or to a false test result.
On page 122, Gigerenzer notes that “Like his fellow Enlightenment probabilists, [Bayes] blurred the distinction between warranted degrees of belief and objective frequencies by trying to combine the two.” While this is true, many activities in the world don’t have clean statistics as do many medical applications, and even there, initial probabilities for a new technology, or a new disease, might well require such a blending. For example, an insurance underwriter might be able to figure out the size of loss distribution for a new venture, such as commercial research labs in space, but he’ll have no data for the frequency of various sorts of loss, total or partial — none have yet occurred for commercial labs in space. Underwriters in that situation would consult engineers, perhaps surveying them for their best guesses. Some would also bias estimates to produce prices which might draw accounts if they feel that’s a good business to be in for the long run. Others might bias estimates to a conservative level to protect financial assets allocated to this new venture.
In any case, Gigerenzer takes on the medical cases in a clear way, providing us with a frightening picture of the confused information and advice sometimes provided by counselors for HIV infection tests and also breast cancer tests. This is partly because the information about “chances of having breast cancer given a positive mammogram”, “chances of producing a positive mammogram if she doesn’t have breast cancer”, and so forth are given in the abstract form of probability ratios or percentages. Too often, counselors and then clients are given a set of abstract numbers which seem to say, quite wrongly, that a positive test result indicates breast cancer is almost certianly present. In fact, a positive test result often indicates the need for one or more tests, perhaps a different test and perhaps an independent repeat of the same test.
Gigerenzer produces information showing that even doctors sometimes don’t understand and usually can’t communicate what these ratios mean. This is understandable. I took several courses on applied statistics and the theory of probability in college and I found my skills for reading abstract statistical information to be rusty when I was reading these articles.
I certainly support Gigerenzer’s contention that we we deal best, at least most easily, with information in a form presented to our ancestors in the environments in which human beings evolved. Basically, this refers to the fact that apes can count, though even simple arithmetic skills of the sort needed in bookkeeping or filling out tax forms is not really so natural to us. When we’re tempted to be too pessimistic about the ability of the human being to adjust to new conditions in time-spans far shorter than those of evolutionary biology, we should realize that most of us can carry out complicated and ‘unnatural’ tasks such as filling out confusing forms though those sorts of tasks were well beyond most highly competent and well-educated men just a few centuries ago. I’ll leave that discussion for the future.
In the breast-cancer example, counselors and clients alike can understand the situation far better if they’re given not abstract percentages but rather information in the form:
- Out of 1,000 women, 10 will have breast cancer and 8 of those will have positive mammograms; and
- Out of the remaining 990 women, 95 will have positive mammograms.
With information presented thus, most medical personnel can reason and explain that the percentage of women who have cancer given a positive mammogram is 8/(8 + 95). Or, more directly, 103 of the 1,000 women will have positive mammograms and 8 of those will actually have breast cancer: 7.8%. This is apparently a realistic example for actual breast cancer tests and a typical population of women being tested. The odds of having breast cancer given a positive result might be much higher for women from high-risk groups, such as some family lines or ethnic groups which carry genes predisposing the women to breast cancer. This can be seen easily by just increasing the 8 to, say, 25. The chances of actually having breast cancer given a positive mammogram rises from 7.8% to 20.8%.
When the data was given in the less natural form of ratios and conditional probabilities, even doctors estimated that the chances of a woman having breast cancer given a positive mammogram would be about 70% rather than 7.8%. There is a very similar situation for tests for HIV infections when the test population is at low-risk for the infection: such as white, heterosexual males. There’s reason to believe that some not even infected with HIV may have committed suicide after positive test results when they weren’t even infected with HIV. This is not a conclusive argument against having the tests, because they generally screen those in low-risk populations who need further testing. When it comes to high-risk populations, a positive result for HIV usually means that the client is infected. A similar statement might be appropriate for tests of breast cancer in high-risk populations: those women carrying genes predisposing them to breast cancer. The final results of a good testing program should be pretty accurate.
As I noted in my book, ( To See a World in a Grain of Sand): our minds are shaped in response to our environments but that shaping has occurred both in the evolutionary history of our race and also in the development of a single human being. I’m probably far more optimistic than most evolutionary theorists about the possibility that the human mind has become an entity that can transcend its evolutionary past so that the entire universe can become its true home — at least in principle, though no particular human mind can actually achieve this state, at least not on this side of the grave. In any case, we have to work hard to develop our own minds properly in a complex world and we have to work smart to develop minds which evolved to meet the needs of apish creatures during the New Stone Age. If there is truth in Gigerenzer’s research programs on the human mind and also in my philosophical exploration of the human mind, then our thinking is a matter of interaction with our environments or with the universe or world as a whole. We’re not computing machines somehow independent of the world and taking in data for that world, data which we process while isolated from that world, data which leads to understandings of the world which imply there is some textbook to be written which explains that same world.
In any case, we’re creating physical technologies, social structures, and mindscapes which are different from anything we are naturally equipped to handle. And, yet, few there are, even among those anxious to be leaders, who will take the trouble to understand the great possibilities and dangers we’re creating for ourselves and for future generations.