What is this thing...? Lecture 20. Realism Continued * Reading for this week: T&R Chapter 12, Hacking, Laudan. (I can't yet find a good reading for Occam's Razor.) Reading for next week: T&R Chapter 14 (not 13). "The Discovery of HIV as the Cause of AIDS" on Canvas. Also: https://www.sciencemag.org/news/2016/10/gathering-hivaids- pioneers-raw-memories-mix-current-conflicts (PDF also on Canvas, but the online version is better.) __________ 1
Last time: Metaphysical Constructivism Kuhn (maybe, Chapter X), Latour (though he seems to have recanted)*, Nelson Goodman ( Ways of Worldmaking , 1978). * Latour: http://www.sciencemag.org/news/2017/10/latour- qa?utm_campaign=news_daily_2017-10- 10&et_rid=35137912&et_cid=1596541 The world as we know it and interact with it is partly (/entirely?) a human product or construction. The way things are, or "the facts," are dependent on our beliefs, language, theories, and/or paradigms. Why believe this? An old and entirely general form of argument: it makes no sense to think there is a world of that kind. 2
A line of argument I think has had more importance: Attempt to reject a different family of views. But the rejection gets out of hand, and leads to a reversal of the view to be rejected. That reversal does not really make sense, but in some fields and discussions, it is seen as important that the rejection of the alternative picture be as definite and forthright as possible. Start from: the view that people think should be rejected: The world "stamps itself" on the mind of an observer or investigator. Shapin: reality acts on scientific belief with “unmediated compulsory force” (1982). A passive conception of scientific knowledge. Especially after Kuhn, this was rejected, in increasingly emphatic ways. * Note a connection to the 'symmetry principle' of sociology of science. 3
Woolgar (the person who wrote Laboratory Life with Latour), in Science: The Very Idea (1988): The argument is not just that social networks mediate between the object and observational work done by participants. Rather, the social network constitutes the object (or lack of it). (Woolgar, 1988, p. 65). The implication for our main argument is the inversion of the presumed relationship between representation and object; the representation gives rise to the object. (p. 65). Part of the problem here: a false dichotomy. Shapin and Schaffer, in Leviathan . Quoted in T&R: “It is ourselves and not reality that is responsible for what we know” (1985, 344). You don't have to choose between two simplified, cartoon-like views. 4
Starting again: there are two sets of questions here. (i) One set of questions concerns rival general views of how thought, language, and theories relate to the rest of the world. What gets called a "realist" option: the world we study exists in a mind- independent way. (ii) A set of questions concern how to interpret what science seems to be telling us, and whether we should believe it. What gets called a "realist" option: most mature scientific theories (those that have survived a lot testing) are at least approximately true. One source of the difficulties in the standard literature: one thing science itself can inform us about is (i). (Quantum mechanics, in some 5
interpretations.) So this is partly a question internal to science.. if we take the messages of science seriously (which brings in (ii). Another: different scientific fields might require very different treatments in relation to (ii). (This I do discuss in T&R ch 12.) Optimism and pessimism 'Scientific realism' is often now seen as partly a debate about whether we can be confident that we are getting things right. Devitt: "Strong Scientific Realism" (SSR). Most of the essential unobservables of well-established current scientific theories exist mind- independently and mostly have the properties attributed to them by science. (Devitt, 2011 - reference to come. See also Realism and Truth ) 6
Laudan: "confutation" paper (CC). We almost never have reason to think our theoretical terms refer to real objects. I think this sort of question has to be handled in a field-by-field way. What applies to physics might not apply to biology or economics. Models, Fictions, Idealizations What relationship do theories and theoretical language have to the world, when things go well? Simple view: they describe it. A theory contains terms, like "electon," "ion," "gene," "species" that refer (whennthings go well) to real things -- real parts of the world. And the theory will attribute properties of various kinds to these objects. 7
"Carbon atoms have 4 outer-shell electrons, and that is why carbon, in chemical interactions, has valence of 4." Questions: How does the term "carbon" refer to carbon, the stuff? A hard problem. But also: not all cases seem to work like this. Model-building in science The Hawk-Dove model in behavioral biology. (Maynard Smith and Price 1971, "Evolution and the Theory of Games." See also Krebs and Davies, Behavioral Ecology. Assume a population contains two types of individuals, Hawks and Doves. They meet in pairs and engage in contests over resources. When a Hawk meets a Dove, the Hawk wins (with no injuries). When a Dove meets a Dove, it's random who wins (and no injuries). When a Hawk meets a Hawk, it's random who wins and the loser is injured. 8
Assume a process of natural selection operating in the population. The population will tend to reach a stable mixture of Hawks and Doves. OK, but: no population actually works like this, with exactly two kinds of individuals who all behave the same way within their type, meet always in pairs, etc.... Some populations approximate situations like this. How should we understand approximation? In a case like this what we seem to do is: specify a structure, an arrangement; see how it will behave; then draw conclusions about the actual world, based on likely similarities. This is model-building. Is all science like this? Are all theories models? No. Consider Darwin. 9
--> Several of styles or strategies of theoretical work. See my 'Strategy of model-based science,' 2006. Also Weisberg's book Simulation and Similarity. One option: model-building is always mathematical. We specify and explore a mathematical structure, and then work out how that mathematical structure relates to the physical world. This does not solve all the problems (how do mathematical structures relate to the real world?) but it might help. However: it seems that not all model-building is mathematical. What about the Schelling model of racial segregation? Thomas Schelling (1969) "Models of segregation", American Economic Review 59: 488–493. And follow-up papers. "Dynamic models of segregation," 1971, etc. 10
Assume people live on a line, or checkerboard. Initially random distribution of two groups, red and blue. But people will move if they are in too small a minority in their neighborhood. Segregation results from a slight aversion for being in a local minority. 11
This is not a model of how racial segregation actually works – it ignores economic factors, inequality, history of settlement, etc. Can be seen as a "how possibly" model, or as a model of one causal factor at work. It does not (in its original form) have equations, as the Hawk-Dove model usually does. How the model looks: a fiction, that can be compared to reality. A simpler analogue of the real world, that is informative about the real world. Giere 1988, Explaining Science (modified by me). 12
His example: Ideal pendulum in physics. Others: Movement over frictionless planes Infinitely large populations in biology Ecologies with only two species interacting Perfectly rational agents within economic markets 13
Neural networks where the neurons are influenced only by (a few) synaptic connections. Many climate models. Deliberate simplification is very common. In extreme cases, it looks like what we are doing is constructing a fictional scenario from scratch, and comparing it to the real world. Giere in later papers: don't say "fiction"! ("Why Scientific Models Should not be Regarded as Works of Fiction," 2009) Setting aside any political unease, this analysis does indeed raise problems. The role of similarity is puzzling. 14
Does it really make sense to see similarity as a relation between real and fictional scenarios? Maybe the 'fiction' side is just colorful talk? But how do we understand the Schelling model? It is a simple relative of various computer models. People say: 'the population is inside the computer.' I say: the computer operates as an aid to the scientific imagination. We specify a situation and want to work out what would happen. The computer helps with that. Weisberg in Simulation and Similarity : some models are computational structures. A special kind of mathematical structure. This includes the Schelling model – these need not be done with a computer. 15
The aim is to specify a set of procedural or algorithmic dependencies, and claim that these are similar to some causal dependencies in the real world. A computer program is a construction made out of many dependency relations that, when combined, give a complex output of some sort. Schelling model is a very simple case that does not an actual computer. 16
Recommend
More recommend
