This is a paper I wrote in my philosophy of science class. The instructor told us that the class would be "one long argument," and at this late point in the course, our task was to summarize and synthesize the material presented thus far, and evaluate it. The question was, "Can science be objective? And if so, how?" We were given six pages to answer (my bibliography was page seven). This is another paper that I'm pretty proud of, so here it is for your enjoyment. Links to the cited articles are given at the end, except for two which I couldn't track down (I'm not counting Locke, I only used him for a money quote at the end). They're all good reads, so feel free to take a walk on the web if you're unclear on a point and want it in the author's own words.
Much ink has been spilled in the attempt to characterize science, both its process and its products, in a way that both is accurate and preserves at least some of our notions of what science is “supposed” to be about. Logical positivism has been rendered untenable by the ubiquity of underdetermination and the nature of observation as fundamentally theory-laden. Is there any way for science to be objective in this light, or is it just a highly political free-for-all? This depends on what we mean by “objective,” for some definitions of the term shall certainly fail. Yet there is an important way in which science can be seen as objective, and in a satisfyingly progressive way.
The heartbreaking thing about logical positivism is that it works so well on paper, but is so impossible in practice. Logical positivism holds that science is objective in two ways: first, the hypothetico-deductive method is allegedly a purely logical system; second, proper observations ought to be theory-neutral. If it were possible for humans to conform to these standards, then yes, a purely logical system informed solely by theory-neutral observations would indeed be objective. Laudan’s account of Duhemian underdetermination provides the relevant specifics, explaining that our observational predictions are deduced from not only the hypothesis in question and our measurements of initial conditions, but also from a set of auxiliary theories (Laudan, 1998). So even if our measurements are spot-on, we cannot tell with certainty whether the hypothesis in question has been verified or falsified because there might be an error somewhere in our auxiliary theories. In short: stuff we take for granted might be wrong.
What’s more, Hanson points out that our observations can never be theory-neutral because the very act of seeing is contextually organized (Hanson, 1958). If we had access to the physical mechanism of our perception, then we could relate our percepta in a theory-neutral way; but since our brain already organizes our perceptions into “seeings” (Hanson’s term), we are simply physiologically incapable of reporting what we experience in any theory-neutral manner.
Even worse, we cannot evaluate competing theories by the “external” criteria of truth, explanatory power, unification, or empirical adequacy. As Kuhn argues, these criteria are all paradigm-relative (Kuhn, 1998): evaluating the empirical adequacy of theory x against theory y will require external standard z, and any z will implicitly favor either x or y straight out of the box. Yet there is no a priori way to pick one z over any other, and so this stab at objectivity sinks irretrievably into incommensurability.
So much for logical positivism and its conception of objectivity. There are other ways to be objective, however, with varying degrees of viability. Perhaps the simplest is that “objectivity” means “truth,” but this raises the specter of “truth” all over again and gives us problems of second-order knowledge and circular justification. Kuhn’s super-empirical virtues get us almost there, but as he himself points out, these are individually imprecise and collectively conflicting. Harding’s account of objectivism also fails in two apparently opposite but intertwined ways. Harding’s account of strong objectivity, I shall argue, is the nearest thing to objectivity that is attainable by humans – and it had better be good enough, for it’s all we can do.
The naïve conception of objectivity as truth bogs the debate down with knowledge problems – if knowledge is justified true belief, then we know what a belief is and how justification works, but how do we know that we’ve got the “true” part? This amounts to attempting to establish a justified true belief that we have a justified true belief. This problem is twofold: not only are we attempting to determine whether it is true that we have truth, we are attempting to justify our justification, and so the problem has circularity built right into it. This is the problem of second-order knowledge; we cannot know that we know because we cannot verify a justified true belief that our justified beliefs are true (i.e. we cannot know that our justified beliefs are true, even if they are)1 . More generally, it is logically impossible for us to “step outside” our ability to assess our beliefs, and thereby assess our assessments from an outside perspective; we will still be using our own faculties to assess our faculties, in other words (Stewart, 2011).
Kuhn proposes a workaround, attempting to at least get outside the empirical virtues of a theory by positing a set of super-empirical virtues: accuracy, consistency, broadness, simplicity, and fecundity (Kuhn, 1998). While these may be used as a fairly universal external standard for theory comparison, Kuhn points out that they have problems of their own, for they are individually imprecise and collectively conflicting (Kuhn, 1998). Individual imprecision refers to the fact that there is no absolute scale of simplicity or fecundity: theories are simply more or less precise, more or less fecund, than each other. The collective conflict between them is that a theory may sacrifice broadness for accuracy, or fecundity for simplicity, and there is no clear way to convert the “currency” of any one of the super-empirical virtues into any other. So while these are indeed objective standards (insofar as they are external to and transcendent of the empirical virtues of a theory), there is no objective way to measure them against each other. So at the end of the day, Kuhn’s super-empirical virtues succeed at objectivity in one important sense, but fail to attain objectivity in an equally important sense.
Harding’s characterization of objectivism is somewhat better, but not quite there yet. She describes objectivism as the position that objectivity consists in being value-free, impartial, universal, and dispassionate (Harding, 1991). While this certainly sounds objective, in practice things become much murkier. Harding goes on to clarify that this account of objectivity is both too broad and too narrow: too broad, for some of the biases and values ruled out by such an analysis might actually be good (e.g. valuing honesty over dishonesty); too narrow, for our “objectivity” is now only intersubjectivity (while individual biases and values might be eliminated, the invisible institutional biases and values are preserved) (Harding, 1991).
Harding calls this “weak objectivity,” and proposes her own “strong objectivity” to replace it. For Harding, strong objectivity consists in being unaffected by personal bias and adequately supported by appropriate evidence (Harding, 1991). This, again, sounds objective; and, again, raises the question as to how we can know whether our biases are affecting our science or not, and whether our evidence is adequate or appropriate. After all, no account of objectivity can escape the need to demonstrate, in some fashion or other, that its standard has been met. But Harding’s strong objectivity accepts the fact of individual bias – while we may not ever be able to eliminate every bias in any individual, we can limit the influence of biases upon the scientific process (Harding, 1991). And since science is a group effort, taking the group perspective is perfectly acceptable. Because we are no longer concerned with whether an individual is objective, but rather whether the group effort is objective, we may accordingly assess the group effort as individuals and check for the following four features which Harding says help make scientific social knowledge be more objective (Harding, 1991):
1. There must be recognized avenues of criticism, such as journals and conferences, to constitute the “arena” in which discourse may take place.
2. There must be shared standards for what counts as adequate or appropriate evidence, to lay the ground rules for discourse in the aforementioned arena.
3. There must be avenues for community response, which may take diverse forms such as publicly controlled grant funds and popular science publications.
4. There must be equality of intellectual authority among qualified practitioners, so that any member of any class is set on equal footing based on scientific acumen alone.
Harding points out that we currently have features 1, 2, and 3 pretty well nailed down, but feature 4 is fairly lacking as we live in a racist and sexist culture.
There are two important points here. First is that these features are not only beyond empirical virtues, they are beyond science itself. These are not features of any given laboratory, but features of the society in which laboratories may be situated; and they accordingly govern not this or that scientific practice, but rather set ground rules for how scientific discourse ought to be structured. The second point to note is that this “fills out” the notion of objectivity, doing away with the all-or-nothing conception whereby you either have objectivity or you don’t. On Harding’s analysis, objectivity is a spectrum, and the best we can do is to try to become more objective over time. This is both a more robust account of objectivity than any previously offered, and also a more achievable one.
And so we see that objectivity is achievable by humans, but only by degrees, and it does not necessarily look like what we might think it ought to at first blush. The objectivity of logical positivism is simply impossible, since theory choice is logically underdetermined and all observations are theory-laden from the moment we make them. Objectivity cannot simply be truth, for this analysis merely places us back at square one. Nor can it consist in objectivism, for this misses the mark altogether. For science to be objective, and for that objectivity to be achievable by humans, we must conceive of objectivity as a spectrum and limit our ambitions to becoming more objective over time (as opposed to entirely objective all at once). Harding’s account of strong objectivity supports this goal and gives clear direction for how to work towards it. While there are those who may say that this sort of objectivity is not “good enough,” they amount to so many incarnations of the universal skeptic that Locke had in mind when he wrote, “If we will disbelieve every thing, because we cannot certainly know all things; we shall do much-what as wisely as he, who would not use his Legs, but sit still and perish, because he had no Wings to fly” (Locke, 1975).
1.It bears mentioning that this only applies to a correspondence theory of truth, for in principle, at least, a belief system can be known with certainty to internally cohere and thereby be true under a coherence theory of truth. But coherence doesn’t matter much when we want to know how the world is; I could concoct a perfectly coherent story about Heaven where everybody rides unicorns on rainbows and plays in piles of puppies all day, but none of this is worth a damn if the belief doesn’t correspond to reality (i.e. if there is no actual Heaven to go to in the first place).
Hanson, N.R. (1958). On Observation. Chapter 1 in Patterns of Discovery: An Inquiry into the Conceptual Foundations of Science. (pp. 4-25) Cambridge: Cambridge University Press.
Harding, S. (1991). Strong Objectivity and Socially Situated Knowledge. Chapter 6 in Whose Science? Whose Knowledge? (pp. 138-163) Ithaca, NY: Cornell University Press.
Kuhn, T. (1998). The Nature and Necessity of Scientific Revolutions (it's chapter 9 - D). In M. Curd and J. A. Cover (Eds.) Philosophy of Science: The Central Issues. (pp. 86-101) New York: W. W. Norton and Co
Kuhn, T. (1998). Objectivity, Value Judgment, and Theory Choice. In M. Curd and J. A. Cover (Eds.) Philosophy of Science: The Central Issues. (pp. 102-118) New York: W. W. Norton and Co.
Laudan, L. (1998). Demystifying Underdetermination. In M. Curd and J. A. Cover (Eds.) Philosophy of Science: The Central Issues. (pp. 320-352) New York: W. W. Norton and Co.
Locke, J. (1975). Introduction. Chapter 1 in Peter Nidditch (Ed.) An Essay Concerning Human Understanding. (p. 44-48) Oxford: Clarendon Press.
Stewart, T. (2011). When Is a Belief Formed in a Circular Manner? Unpublished manuscript.