Week two of my new course, Scientific Inquiry–Theory & Inference, seeks to provide first year PhD students who want to join the scientific community the basic training they require. Week one addresses how to succeed in graduate school. Week two is the first of a two week section on Human Knowledge. I have titled it Ontology & Epistemology, but that title isn’t great. This week provides the foundation that justifies the superiority of understanding science as community practice: a set of norms created to solve collective action problems implied by the material we cover in week two. We structured the course to make this claim. And it take an entire semester to present the case.
Ontology? Scientists Don’t Need no Stinking Ontology!
That is correct. Contributing knowledge claims that achieve high status within a scientific community does not require any reflection about the “real world” and how human’s acquire knowledge about that world. So why chew up valuable time reflecting on such an arcane topic?
Training students so that they can become producers of knowledge about the processes that generate political regularities is our primary goal. Achieving that goal does not require that the students invest effort in constructing a logically coherent model of how why science is effective. The elementary school approach that defines science as a series of steps (a method) pursued by individuals is sufficient. Imagine, however, that you want also to provide training so that some of them can not only contribute knowledge claims, but also play an active role in (re)shaping the norms and institutions that demarcate knowledge claims that (should) achieve status in scientific communities from those that do (should) not.
I argue in this course that the group (human community) is the scale at which we can best construct an explanation of the superiority of science for the production of knowledge claims that explain regularities humans observe (aka “causal theories,” “causal explanations,” and so on). Standard accounts of “science,” “the scientific method,” etc. in these courses assume sufficient correspondence between “reality” and human perception of it that we can treat any disconnect as random noise. That assumption clears the path for a focus on the norms and activities an individual scientist must train herself to adopt and use. Science has functioned for centuries (and political science for decades) with such an individual focus.
Yet the norms and practices are not static. The norms and institutions that regulate all human communities are contested. Sure, we can treat them as static (in equilibirum) over short runs, but they are dynamic as we increase the scale of temporal aggregation we use (covered in week 12). And take a guess at who I will turn to to help me develop a useful model of the function of norms and institutions? Not philosophers! Nor historians! Nor physicists! Nor biologists, chemists, engineers, psychologists, etc. Nope. I turn to social scientists to help me develop such models.
But wait. I have gotten ahead of myself. I have not yet explained why we might find value in a model of science that operates (in part) at the human group (community) scale and focuses attention on norms and institutions. What collective action problem(s) are these norms and institutions helping groups of humans solve?
We can invoke an assumption that the human mind directly perceives reality. That assumption permits a model of science as “a method” that individuals can be taught to execute. Indeed, that is a popular model of science taught in elementary and middle school. Yet an alternative, model of science with much greater theoretical power is available.
What if we conceive of science not as a method practice by individuals, but instead a set of solutions to a variety of collective action problems that standard assumptions about human shortcomings suggest would hobble individual researchers’ ability to execute a method? We might, for instance, assume that researchers seek to maximize their status. Such an assumption will readily produce hypotheses about fraud and other problems that bedevil scientific communities.
We might further assume that scientists recognize that individual human beings’ status seeking motive will undermine any attempt to construct a “method” to which individuals should hew. Each has an incentive to cheat, thereby undermining the collective production of useful knowledge. To solve the individual v collective incentive clash they might form a community defined by norms and institutions that incentivize individuals to self police and hew to the community norms (effectively by threatening ostracism should one be exposed as a fraud).
The “model” of science I am sketching for my students is quite obviously a social scientist’s model of science. It is also entirely consonant with what we might call a post modern ontology and a constructivist epistemology. The primary advantage this model of science offers is direction with respect to thinking about the community’s norms and institutions. Rather than thinking about “file drawer problems,” “causal identification,” or whatever the flavor of the year happens to be, as isolated issues absent a comprehensive model of science, we now have a model of science with which all social scientists can work as we reconstitute our community’s norms and institutions. In this model science is a social contract the community (re)produces that define the criteria knowledge claims that warrant intersubjective agreement must meet. Because there is no “real world” to perceive, describe and understand, but we are capable of producing useful knowledge about social regularities, yet have individual incentives to amass status and advance claims that reinforce our status, we require norms and institutions that will produce that useful knowledge “on average” at the community level.
Note, also, this important implication available from this approach (which is not available to the science as individual practice approach): progress will occur, though the temporal scale for progress is entirely unclear. No individual study contains “scientific.” Science is, instead, the norms an institutions we contract collectively.
This week’s readings make a case for why we might adopt a post modern ontology and a constructivist epistemology, reviewing both some philosophy of science (at an undergraduate text level) and human perception (at a popular science level). Onward to the reading.
The Assigned Reading
I assign brief video presentation of Wittgenstein’s Beetle in a Box, and then have them read about 2015’s “color of the dress” controversy, and a brief piece about the human tendency to find patterns in meaningless noise. I also assign Jonathan Haidt‘s discussion of confirmation bias (pp. 91-7) and Steven Pinker‘s discussion of the perpetrator/victim narrative (pp. 488-92).
These works show us that models which assume that our brain’s filter the information we collect from the world, and that the processes are biased by evolutionary selection that makes our species “fit” into the natural world such that we have yet to become extinct, can explain quite a bit. Put in more familiar terms, models of human knowledge production that assume that the human brain is a biased instrument vis-a-vis “reality” can explain more of what we agree we are regularities than models that do not. Those seem like pretty large parts of social life to have no account for.
As for Wittgenstein’s thought experiment, such models of human knowledge production provide no rival account for human language. The concept “facts” is nothing more than a convention that humans have agreed upon. That is true of all human language: all human knowledge is constructed, including the models of science we have created.
I also have them read a brief unpublished essay of mine that sketches the differences across pre-modern, modern and post-modern ontological positions, and AF Chalmers (pp. 1–26) and Earl Babbie’s The Practice of Social Research provide more detailed, yet very accessible, accounts (pp. 14-37 and 51-66).
Steven Jay Gould describes the limits the human tendency to focus attention on typical and bizarre outcomes, ignoring the variation between, places on theory building (pp. 44-56 and 77-79). Arez Aiden & Jean-Baptiste Michel tell a very accessible story about how Zipf chose to look at words differently and the role his conceptual shift plays in providing part of the foundation upon which Big Data is constructed (pp. 26-8, 32-7 and 49-50). Cohen & Nagel. 1934. “Facts and the Scientific Method,” An Introduction to Logic and Scientific Method pp. 391–2 observe that “nature” does not “imprint” facts on our brains–facts are theory dependent.
All of the reading lays the foundation for establishing the centrality of “intersubjective agreement.” There are no “facts,” only shared agreement about our understanding of our experiences. Language like that sends many scientists scurrying for bedrock. So let me perfectly clear: a model of the success of science that begins by embracing post-modern, constructivist accounts ontology and epistemology is superior to the prevalent (modern) models of science for training PhD students.
Finally, I have tacked on Healy & Moody’s data visualization primer because it kind of fits here, and I want them to be aware that visualization is a topic in its own right. In seminar I tell them that I believe visualization is not given proper attention in PhD training, and urge them to pursue “self education.”
But They Ain’t Know Nuthin Yet
Yup. We cover that in week one, and the students are, in fact, first years. They are learning to bang out “Twinkle, Twinkle Little Star.” And that makes the reading quite a bit to swallow.
That is why I assign so much undergraduate level reading (Chalmers, my essay, Cohen & Nagel) as well as popular accounts of the issue. Second, these courses inevitably ask students to “drink from a fire hose.” Sure, they’ll slake their thirst, but they are gonna get hella wet, and most of the info is splashes to the floor and later evaporates. But my view is that one wrestles with these issues throughout graduate school, and for many of us, throughout our careers. If I knew how to do this course in multiple years, and the curriculum committee would approve the courses, I would do it. But I don’t, and they won’t.
 Nate Monroe and I developed this course together, though our final syllabi are not precise replicas of one another.
 We do not discuss scale until the fourth meeting in the Theory section of the course (Week 12). Welcome to the Gordian Knot that these courses are.
 We might further assume that income and other valuable things are a a strong positive function of status.