Cilantro, Moral Truth, and Justification

Preamble: 

I've been working on this paper for longer than I care to admit but I have to turn it in at some point. I've written about 4 or 5 different versions of it all with different solutions or non-solutions to the puzzle I present. Anyhow, a few notes:

(A) For some reason the footnotes didn't post to the blog so some of my clarificatory points aren't here. Here are two of the important ones. 

     (1) The anti-realist position I'm concerned with is error theory (there are no moral facts and moral propositions have cognitive content). 

     (2) In the last section I talk a lot about "evidence". What counts as evidence in moral arguments would require its own paper so I make some assumptions about what counts: moral judgments, intuitions, principles, and emotions. I'm happy to include other candidates. 

(B) For non-philosophers all you really need to know to understand this paper is what moral anti-realism is. In the simplest terms it's the view that there are no moral facts. Everything is, like, just your opinion, maaaaaaaan!

Cilantro, Moral Truth, and Justification

Appetizers: Anti-realism About Gustatory Facts 

At dinner tables around the world, there is perhaps no issue more divisive than whether cilantro is or is not a delicious garnish. It is the gustatory domain’s own abortion debate. There’s little middle ground and each side views the other as irredeemably wrong. In more reflective moments, most would agree there are no objective facts about the deliciousness of particular foods.  Abe can claim that cilantro is objectively delicious while Bob claims that cilantro is objectively disgusting but the fact of the matter is that there is no fact of the matter! Granting this assumption, is there any way that we could make sense of the idea that either Abe or Bob’s belief is better justified than the other? 

For the moment, I’m going to assume that there isn’t. It seems as though Abe and Bob could offer justifications for why cilantro is subjectively delicious or disgusting but I doubt any of these reasons would convince a third party of cilantro’s objective gustatory properties. Abe and Bob could insist that their arguments and reasons support objective gustatory facts but we’d dismiss their claims as category mistakes—they’re confusing their personal preferences for objective facts about the world. Any argument they give for objective gustatory facts about the world is better interpreted as facts about their subjective gustatory preferences being projected onto the world.

Now consider an analogous moral case and substitute your favorite moral propositions and their opposite for the gustatory ones. For example, Abe claims that it is an objective fact that slavery is a morally good institution while Bob claims the opposite—i.e., that it is an objective fact that slavery is a morally bad institution. If in the cilantro case anti-realism about objective gustatory facts leads us to accept neither competing belief is better justified than the other then it seems that consistency requires that anti-realism about moral facts lead us to also conclude that neither Abe nor Bob’s beliefs regarding slavery is more justified than the other.  Just as there are no objective facts about the deliciousness of cilantro, there are no objective facts about the moral badness or goodness of slavery, and so one position cannot be more justified than the other. Any argument is merely a projection of the interlocutor’s personal preferences or explainable by appeal to facts about their psychology.

There may be some extreme anti-realists out there that are willing to bite the bullet and concede the point.  However, I’m willing to bet that many anti-realists would deny that all moral beliefs are equally well-justified even if moral beliefs can't be objectively true or false. If I'm right then these anti-realists need an account of justification that doesn't depend on the notion of truth. Is this possible?

The framework for this paper is to examine the relationship between moral anti-realism and justification. Suppose we accept that MORAL ANTI-REALISM IS TRUE: There are no objective moral facts. On what bases can we then evaluate competing moral claims? Is justifying objective moral claims analogous to trying to justify objective gustatory claims? That is, since there really are no facts of the matter, one claim is just as well-justified (or unjustified) as the other. The puzzle for the anti-realist is to reconcile the following two assertions: MORAL ANTI-REALISM IS TRUE with NOT ALL MORAL CLAIMS ARE EQUALLY JUSTIFIED.  I will argue that, if we adopt an externalist theory of justification, a Peircian fallibilism offers a potential solution to the puzzle. Before proposing my solution, I will consider and evaluate other attempts to reconcile the two assertions.

A Quick Word on Theories of Justification

Before proceeding we’re going to need to take a brief look at theories of justification and pare down the scope of my inquiry.

Two Theories of Justification

One way to analyze the concept of justification is along internalist/externalist lines. Internalists argue that a belief is justified so long as the believer is able to provide some sort of argument or supporting evidence when challenged. Externalists argue that a belief is justified if it was generated by a reliable belief-forming process, where "reliable" means that the process generates more true than false beliefs in the long run. For example, beliefs formed by visual perception are justified on the externalist view because visual perception generates more true beliefs than false beliefs in the long run.  So, my belief that there is a computer in front of me is justified because it was formed by my seeing it—which is a reliable process. It’s more likely that I’m actually seeing a computer than I am hallucinating it. 

I wish to side-step taking a definitive position on the internalist/externalist debate and suggest that both types of justification are plausible in ethics. We think that a moral belief is justified via normative reasons (internalist) but we also think particular ways of arriving at moral beliefs confer justification. For example, we think that moral judgments arrived at through careful reasoning and/or reflection are more justified than those produced by unreflective emotional knee-jerk reactions. And so it's plausible to think the reliability of the process that produces a moral belief is at least partially relevant to the belief's (relative) justification. For the remainder of this paper, I will grant myself that assumption and constrain the scope of justification to externalist justification. A full treatment of an internalist model in regards to my inquiry requires a paper unto itself—although I suspect internalist theories may have similar problems

Round 1: Externalist Justification Isn't Possible if There Are No Moral Facts

The simple argument against the possibility of an anti-realist account of externalist justification goes something like this. 

P1. Reliability is cashed out in terms of whether a process produces more true than false beliefs in the long run.

P2. Anti-realists deny that moral propositions can be true or false.

C1. So, there's no way to evaluate the reliability of a process when it comes to moral beliefs because the very attributes that we require to measure reliability aren't available.

C2. So, on the anti-realist model all moral beliefs are equally justified (or unjustified).

In short, anti-realists deny the very attribute (truth) required to measure reliability. If we can't know which processes are more reliable than others there is no externalist ground to say one moral belief is better justified than another.  But, again, surely some moral beliefs are better justified than others…but how? 

Reply

Consider the inference rule modus ponens . We know modus ponens to be a reliable belief-forming process from using it in other domains. It is content neutral. Its reliability credentials have been checked out so all we need to do is import it (and similar content-neutral processes) into the domain of ethics. Anti-realists can say that moral conclusions arrived at through modus ponens (or other combinations of formal rules of inference) are more justified than those that aren’t. 

Counter-Reply

Modus ponens and other valid argument structures are contingently reliable processes. That is, if the inputs (i.e., premises) are true then so too will be the outputs. The problem is that the anti-realist has denied the possibility of true inputs in the moral case. If inputs can be be neither true nor false then the conclusions are also neither true nor false. And worse yet, the same argument structure can yield apparently contradictory outputs.

Consider the following examples:

Ethics 1

1E. If you abort a fetus it is wrong. (Neither true nor false).

2E. I aborted a fetus. (Neither true nor false).

3E. Therefore, what I did was wrong. (Neither true nor false)

Ethics 2

1E*. If you abort a fetus it isn't wrong. (Neither true nor false).

2E*. I aborted a fetus. (Neither true nor false).

3E*. What I did wasn't wrong. (Neither true nor false).

It appears as though importing valid argument structures into ethics doesn’t give a solution to the puzzle of reconciling MORAL ANTI-REALISM IS TRUE with NOT ALL MORAL CLAIMS ARE EQUALLY JUSTIFIED.

Round 2: Content-Generating Processes

Perhaps the problem here is the content neutrality of the above processes. We need processes that justify initial moral premises as well as yield conclusions. We have some familiar plausible candidate processes that might confer justification: reflective equilibrium, rational reflection, rational discourse, coherence with existing beliefs, idealizing what a purely rational agent would want, applying the principle of impartiality, universalization, to name a few.

Notice also that we think that some cognitive processes for ethical judgment don’t confer justification—i.e., are unreliable. For example, if I form a moral judgment when I'm extremely angry I might come to reject that belief once I calm down and employ one of the above methods instead. So, a belief arrived at as a consequence of a temporary acute emotional reaction is not well-justified.  

Moral psychology and social psychology are littered with experiments where either the subject or the environment are manipulated to produce judgments that the subject would reject upon learning of the manipulation. This seems to hint at an answer: Beliefs that have been formed by processes that involve manipulation of the environment and/or the subject’s mood are less reliable/less justified than processes that don't involve any obvious manipulations. 

The Same Problem?

This account implies that some processes are more likely to "get it right" than others. But what is it to “get something right” if there's no target for epistemic rightness (i.e., truth)? This seems to be the same problem from Section 1 all over again. If moral beliefs can neither be true nor false, in what way can we say that one process yields more true outputs than another? Sure, we might systematically reject judgments produced by some processes in favor of others, but why  prefer the outputs of one process over another if we can’t say that judgments from one is more likely to be true than another? The grounds for thinking judgments from one process are more justified than those from another seem to be that the judgments from one process are more likely to be true than another.

Round 3: Analogy with Scientific Justification and Fallibilism as a Possible Solution

Harmon argues that there is an important disanalogy between explaining ethical judgments and scientific judgments. We can’t explain why a physicist believes there is a proton without also postulating that there is something “out there” in the world causing the physicist’s observation, which is in turn interpreted as a proton. A moral judgment, on the other hand, can be explained merely by appeal to a subject’s psychology. We needn’t postulate any thing or moral properties “out there” that cause a subject to have a moral belief that x. 

Let’s accept Harmon’s argument. Despite the fact that the causes of scientific and moral judgments might differ, there may be ways in which justification functions similarly between both.

Scientists habitually couch their arguments and conclusions with language that is fallibilist.  Claims and conclusions are presented as provisional, based on the current available evidence and methods. The history of scientific discovery is one of revised and overturned conclusions in light of new evidence and recursive, self-correcting improvements in the scientific method itself.  

From the point of view of internalism vs externalism about justification, we might consider new data as a kind of internalist justification for a claim because they are reasons to believe one thing rather than another. Research methods, on the other hand, can be viewed as instances of the externalist’s processes that confer justification. The idea that some processes are more reliable than others is a familiar idea in scientific research. Claims that derive from methods (i.e., processes) that avoid possible known biases are more reliable and hence more justified than methods that don’t.

For example, the placebo effect is a well-known occurrence in medical science. If patients think they are receiving treatment—even if they aren’t—patients report subjective measures (e.g., reduced pain/discomfort) significantly more positively than non-treatment (i.e., control) groups. We also know that if the researcher knows which patients are in the treatment group and which aren’t, this can influence both the way the researcher asks the patient questions and how they interpret data (they’ll bias toward a positive interpretation). For these reasons we think that the results from medical research that are placebo-controlled and double-blinded are more reliable than those that aren’t. 

In short, data from a study that employs a more reliable process (e.g.,. double-blind, placebo controlled) is more justified than the data from a study that didn’t do either of these things. The more a process avoids known errors, the more justified its conclusions—despite the fact that the blinded study’s conclusions might also eventually be overturned. There is always the background understanding that new and better methods might come about and generate incommensurate data and/or conclusions but this doesn’t undermine the current relative justification that the current methods confer on the outputs beliefs. 

Analogously, moral and social psychology has produced a vast literature showing all the ways we think our moral thinking can go awry.  We know that a cluttered desk can cause us to pass harsher penalties than we would otherwise, that a temporary heightened emotional state greatly influences our judgments, that the feeling of empathy can lead us awry,  and that implicit biases can play important roles in our judgments—to name a few. In short, there are many ways in which it seems like both our basic hard-wiring and various forms of personal and environmental manipulation can cause us to make judgments that we, upon learning of the manipulation or the bias, would likely reject in favor of a judgment arising from unmanipulated deliberation or one of the familiar gold-standard methods of moral reasoning .

Perhaps an anti-realist can think of the activity of moral thinking not as one that aims at discovering some objective truth, rather one that seeks to avoid known cognitive errors and insulate against manipulation. In so far as our judgments derive from methods that avoid (known) cognitive errors and biases, our moral claims are better-justified. This however doesn’t entirely solve the puzzle. We still need to answer why we should choose the output of one type of process over another.  It’s easy to say that the manipulated judgment is “wrong” or is “mistaken” but how do we say this without appeal to truth? “Error” implies a “right” answer. One might just as easily say that the correct judgment is the manipulated or biased one and we are systematically mistaken to adopt the reflective judgment made in a cool dark room. 

The Main Challenge to The Anti-Realist: The moral anti-realist needs some general criteria to explain why we (ought to) systematically endorse judgments from one process rather than another. That we do isn’t enough. We must give an account of why one confers more justification than another.

Peirce, Methods of Inquiry, and The “Fixing of Opinion”

Before suggesting a possible criteria to answer the challenge, I want to sketch out a Peircian analysis of methods of inquiry since it inspired my suggestion. I’ll fill in other details later as they become relevant. Pierce argues that the “fixing of opinion” or “the settlement of opinion” is the sole object of inquiry (Peirce, p. 2-3).  While we needn’t commit to Peirce’s exclusivity claim regarding the purpose of inquiry, Peirce provides useful insight into why we might think some processes confer greater justification than others. I take him to be proposing two related desiderata for our methods of inquiry: that they produce beliefs that are (a) relatively stable over time and (b) relatively impervious to challenge.  The anti-realist can distinguish between processes’ justificatory status to the degree that they they achieve (a) and (b).

This possible anti-realist response to the challenge of justification is not without difficulties. I will explore two related ones. First, it isn’t clear that stability of beliefs is a condition for justification. As standard examples of recalcitrant and dogmatic Nazis or racists show, in fact, we have reason to believe stability might have little to do with justification. Second, I will need to show that stability marks something we think matters to justification: namely, the absence of cognitive errors and inclusion of all relevant evidence. This is the fallibilist aspect of the proposal: Stability needn’t be a proxy for truth tracking, however, it is a reasonable proxy for believing that we are avoiding errors. This is part of the answer but not all of it. 

Peirce notes, stability can be achieved various ways—not all of which we’d think confer justification. The second part of this fallibilist model of justification has to do with the degree to which a process excludes relevant evidence thereby making it susceptible to challenge. Thus, for Peirce, long-run stability requires that the method of inquiry take into account all relevant sources of evidence. A method that excludes forms of evidence will produce beliefs that are more likely to eventually be overturned or require that people be unable to make inferences from one case to another or from one domain to another.  Piece compares 4 methods of inquiry, which aim to produce stable beliefs. In so doing he illustrate how this second criteria (i.e., imperviousness to challenge) works with the first (i.e., stability) to produce a theory of justification.

To achieve stability of belief in the method of tenacity one dogmatically adheres to his chosen  doctrines by rejecting all contrary evidence and actively avoiding contradictory opinions. Coming into contact with others, however, necessarily exposes him to different and conflicting views. This method can’t accommodate ever-present new data without giving up on rationality and the ability to make inferences. Unless we live like hermits, the “social impulse” against this practice makes it unlikely to succeed. The method of authority is the latter method but practiced at the level of the state and with the addition of social control. This method fails to achieve long term stability in that it cannot control everyone's thoughts on every topic. This isn't a problem if people are unable to make inferences from one domain of inquiry into another, but to the extent that they are, stability will be undone and doubt will emerge.

The above two methods have an important common feature in regards to how they achieve stability. In both cases, when I need to decide between two competing beliefs, the method tells me I ought to pick the one that coheres with what I already believe. In other words, there will be cases where I will have to “quit reason” by rejecting contrary evidence and inferences. The above methods for arriving at beliefs systematically exclude relevant evidence in generating beliefs. 

Now, compare these methods with something like wide reflective equilibrium or rational discourse. With these methods, how do we determine what to believe?  Rather than exclusively referring back to what we or the state/religion already endorse, when these methods confront new evidence they adjust output beliefs accordingly.  In the long run, the supposition is that (for example) reflective equilibrium and rational discourse will lead to more stable beliefs than the above two methods because the outputs include the best available total evidence rather than reject it.

Reply to the First Challenge

Let’s restate the first challenge: Stability on its own doesn’t seem to confer justification. There are many way methods of inquiry by which we might achieve stable beliefs, not all of which confer justification. When we defend stability as a justificatory property, the concern is that we’re begging the question: when a process generates stable beliefs that we approve of, we think stability is good; when it generates beliefs we disagree with, we think stability is bad.

To reply to the challenge, let’s consider, for example, reflective equilibrium. With (wide) reflective equilibrium we suppose that by finding an equilibrium between everyone’s principles and considered judgments, in the long run we arrive at a view that no one could reasonably reject (because it also takes their views into account). Stability, on this method, arises as a consequence of taking everyone’s principles and considered judgments into account; i.e., it doesn’t obviously exclude any evidence that might, in the long run, diminish the stability of the beliefs the method produces. And so the critic of stability has a point that stability on its own might not confer justification. What matters is why the view is stable. It’s assumed that a moral view derived from processes that take into account competing view points are stable for the right reasons: they are less impervious to challenge. And they are less impervious to challenge because they don’t exclude evidence (widely construed).

Reply to the Second Challenge

The second challenge is to give positive reasons of thinking long-term stability confers justification. When we make knee-jerk judgments or manipulated judgments we end up with beliefs inconsistent with our other beliefs, and so we have reason to conclude we’ve made an error somewhere—either in endorsing an inference, a judgement, a general principle. Conversely, with a process like reflective equilibrium or rational discourse we eliminate inconsistencies in the long run. By proxy, we’re also eliminating errors in either our inferences, judgments, or general principles; i.e., things that undermine justification. In the long run, the supposition is that, some methods of inquiry (e.g., reflective equilibrium) will lead to fewer and fewer errors, in turn contributing to more and more stable beliefs.

This also helps answer a problem I raised in the first section of the paper.  A reliabilist account of justification has truth built in and so doesn’t help an anti-realist explain how following a valid argument structure can confer justification. If the inputs don’t have a truth value, then neither will the outputs and so one output is just as justified as another. A fallibilist approach provides a solution. Failure to follow a valid argument scheme undermines justification because it indicates an error in reasoning. And so, deliberative belief-forming process that don’t follow valid schemes are making errors and in that respect their outcomes are less justified than those that do. 

Elimination of errors is part of the answer to why we think beliefs derived from one process are better justified than beliefs derived from another. Long-term stability of output beliefs is partly a proxy for absence of errors and so, to the degree that a process generates stable beliefs we have reason to think those beliefs are less likely to contain or be the product of errors. 

The second part of the answer is similar to the reply to the first challenge: Processes that attempt to exclude classes of evidence or deny certain inferences aren’t going to be stable. There’s an analogy with science: a research method that regularly has its conclusions overturned because it fails to take into account certain sources of evidence is a process that generates beliefs that are less stable in the long run. The beliefs are less stable because important classes of evidence aren’t taken into account or controlled for (for example, various cognitive biases). Similarly, a moral reasoning process that fails to take into account certain sources of evidence (e.g., competing arguments, the fact that we are prone to certain cognitive errors) is also going to generate beliefs that less stable, and by extension, less justified.

Conclusion

The puzzle for the anti-realist is to reconcile commitment to no moral facts with the view that some moral beliefs are more justified than others. If we take an externalist account of justification, a Peircian fallibilism offers a possible solution to the puzzle. Why should we think the the outputs of one process are more justified than another if their output can’t be true? Some processes generate beliefs that are relatively more stable and impervious to challenge in the long run than do other process. Stability, on this model occurs as a consequence of taking into account all the relevant data and avoiding cognitive errors in generating output beliefs. By doing so, outputs are less likely to be overturned by excluded evidence. Stability also is a proxy for absence of error. If a process produces beliefs that are systematically over-ridden, it must be because the outputs are inconsistent with other judgments, beliefs, or inferences. Processes that systematically generate inconsistencies indicate errors, which in turn also undermine stability. 

I’d like to close with the following thought experiment. Suppose both realists and anti-realist agree on which processes confer greater relative justification than others. Would the realism/anti-realism debate matter much? Aren’t comparatively well-justified beliefs (and actions) what we’re really after in ethics?

Realism vs. Antirealism on Theoretical Entities: Maxwell vs van Fraassen

Preamble
Here's a paper I maked for filosofy of sciences class. i maked it myself.
There's an inside joke with my professor Dr. J. Brown that needs some explaining to understand part of this paper.  Every time he uses hypotheticals he always says "I could have come to class in a bear suit".  Anyway, the grad students all pitched in an bought him a bear suit (he's well-liked)


The paper is about the realist vs antirealist debate in science.  That is, does science get at how the world really is or does it just produce coherent theories.  There are lots of strands of realism vs antirealism debates but my essay involves the one over the existence of theoretical entities.  Basically, unobservable things like atoms and electrons and subatomic particles have never been observed directly.  They've only been observed indirectly.  Furthermore, there are limitations on what we actually can observe so it is unlikely they will ever be directly observed.  For that reason they are theoretical entities: we infer their existence from detected properties.   It's kind of like guessing what's in a box by shaking it, feeling its weight, and evaluating how it moves in the box.  We'll never be able to look in the box.  


The realists argument can be boiled down to what's called the "miracle" argument which goes like this: Supposing science isn't right about theoretical entities--it's a miracle that it's wrong yet  we have the coherence of the theories, the explanatory power, the predictive power, and the ability to use the knowledge in applied technology.


The basic anti-realist argument is called "the pessimistic meta-induction".  This goes like this: science has gotten it wrong in the past about theoretical entities (eg. mass, gravity, ether, phlogiston, etc..) why should we think this time is special?  In this paper one of the main arguments is slightly different and it goes like this: epistemic principles don't give us ontological conclusions.  In normal language that means just because a theory has a certain desirable property (coherence, explanatory power, simplicity, predictive power) it doesn't make it True of the world, it only means it has that epistemic property.  And (related to the pessimistic meta-induction) there have been countless theories in the past with those properties, why should we think our current ones are special?

Anyway, that's the debate in a nutshell.  For me this in one of those philosophical issues that make you loco.  I changed my conclusion 5 times over the course of writing the paper.  I kept switching sides.  It's not an issue to which I think there will be any resolution soon.   


It's not necessary, but I suggest reading my 2 previous posts on realism and antirealism for background, but if you don't care to, you should be able to follow along anyway.

Thanks for stopping by....

Can Epistemic and Pragmatic Reasons Bear on Ontological Conclusions about Theoretical Entities?

An Investigation into the Debate between Maxwell's Realism and Van Fraassen's Antirealism about Theoretical Entities 

Introduction

In rejecting logical empiricism1, van Fraassen and Maxwell end up on different ends of the realist-antirealist divide. For Maxwell, the observable-unobservable distinction is arbitrary and so is a poor principle from which to infer antirealism about theoretical entities. Conversely, van Fraassen argues that ontological conclusions do not follow from a theory's epistemic merit; therefore, so long as a theory is empirically adequate (i.e., is consistent with what we see in the observable world) there are reasons in favour of using it, but we ought to remain agnostic about the ontology of theoretical entities. I will suggest that both positions fail to give adequate consideration to the implications of differences in epistemic and pragmatic merit between scientific theories2: both accounts imply an all-or-nothing ontological attitude toward theoretical entities as a whole. Instead, perhaps we ought to adjust our ontological attitude toward theoretical entities in relation to the aggregate of a particular theory's epistemic and pragmatic qualities in which an entity is embedded. While van Fraassen is probably right concerning theoretical entities that no particular epistemic quality confers upon them ontological certainty, this does not preclude us from assessing a theory's epistemic and pragmatic qualities and making reasonable inferences in favour of or against the ontological status of various theoretical entities which will ultimately inform our ontological attitude.

In laying the groundwork for my conclusion, my paper will follow a thread of possible replies and counter replies beginning with Maxwell's argument that the observable-unobservable distinction is arbitrary and so ontological claims against theoretical entities cannot rest on this distinction. Specifically, my paper will have the following structure: (1) Discussion of Maxwell's argument for the arbitrariness of the observable-unobservable distinction; (2) possible anti-realist replies; (3) discussion of possible realist counter-replies; (4) possible anti-realist replies; (5) conclusion.




Maxwell's Argument for the Arbitrariness of the Observable-Unobservable Distinction
Maxwell gives us several arguments against anti-realism toward theoretical entities from which he concludes we ought to be realists. The first argument Maxwell gives is in response to Bergman's anti-realist argument which I'll call the “distortion argument”.

Suppose we are looking at a microorganism through a compound microscope. What do we observe? Bergman says, “all I see is a patch of colour which creeps through the field like a shadow over a wall. And a shadow, though real, is certainly not a physical thing” (quoted in Maxwell, p. 1055).

The gist of this argument is that what I'm seeing is a distortion of the (theoretical) entity in the petri dish, so how can I be sure that the properties of the image I see through the microscope are the same as those of the entity in the dish?

Maxwell's replies that it is simply balderdash to deny that what we are looking at is real because the light travels through a lens. He uses an argument ad absurdum to demonstrate his point: The underlying logic of the skeptical argument is that if we are observing distorted light or a projected image of the entity, then we are not looking at the thing itself; so, we cannot make any ontological claims about it. But, if distorted light is grounds for denying something's ontological status this would imply that people who wear glasses cannot make ontological claims about what they see.

We can further demonstrate this absurdity when we apply it to looking through a wet window. Should the fact that the light is being slightly distorted impede us from believing the world on the other side is real? Similarly, when we look at the moon or other planets through a telescope, they certainly appear to be physical objects rather than “patches of colour”. Are we to deny their reality because the lens has distorted the light or projected an image of the object?

Because of the continuity between methods of observing the world on the micro and macro scale, where we draw the line is arbitrary—and the logical extension of the distortion argument is clearly silly. In sum, because the ontological status of entities clearly doesn't vary with our ability to perceive them and where the observable/unobservable line is drawn, we ought to reject the observable/unobservable distinction as a grounds for ascribing or denying ontological status.




Antirealist Reply: Arbitrary Line Doesn't Imply No Line
The antirealist can reply that this argument is a caricature and fails to meet the central objection. While it is certainly true that there is a continuum of observational enhancement techniques it does not follow that there are no important differences between extreme ends of the spectrum. Where we draw the line will to some extent be pragmatic, but there are important differences between extremes. Surely, we can say that there is a difference between looking at a room with glasses on, and observing bleeps and blips on an instrument panel at CERN to infer the existence of subatomic particles.

To illustrate, we can draw an analogy with the classical problem of distinguishing between what is constitutive of baldness and what isn't. While it seems correct to say that there are grey areas along the continuum, this does not preclude the possibility of canonical cases. A head with no hair is a bald head--that much is certain, and Charlie Brown with his one hair is also bald. Hair thick enough so that the scalp is not visible is not a bald head. Of course, there will be cases in between that are unclear but this does not preclude the existence of clear cases on either end of the continuum.

The same might be argued of observation. While Maxwell is correct to point out that there is a continuum and that there is some arbitrariness in where the line is drawn, this does not entail that there are no clear cases of each end of the spectrum. This fact is not diminished by the problems of where to draw the line.

In the case of the observable-unobservable distinction, in what might that difference inhere? I suggest that at least one important difference has to do with the length and epistemic certainty of the chain of inference from observation to conclusions about the properties of a theoretical entity. Consider the “direct observation” end of the spectrum: It does not seem unreasonable to say that my looking at the wall and concluding that “there are walls and they are solid” is a direct form of observation whose conclusions we should consider valid. If this isn't sufficient, to further establish certainty about the properties I'm observing, I can employ my other senses and ask my friends to verify my observations. So, why ought we to consider the conclusions from this type of observation valid? Because the inferential step is minimal and my knowledge comes from a epistemically reliable source: the direct experiences I have of macro objects justify my ontological conclusions about them3.

Now consider the other end of the continuum: how we get images of atoms from a scanning tunnelling microscope (STM):

The STM is based on the concept of quantum tunnelling. When a conducting tip is brought very near to the surface to be examined, a bias (voltage difference) applied between the two can allow electrons to tunnel through the vacuum between them. The resulting tunnelling current is a function of tip position, applied voltage, and the local density of states (LDOS) of the sample. Information is acquired by monitoring the current as the tip's position scans across the surface, and is usually displayed in image form. (Wikipedia4)




In this case, there are several inferential steps: the information from the tunnelling current-- which has been derived from other variables--is run through software that converts it into other electrical signals, which in turn are converted into an image on a monitor. So, are we observing an atom when we view it through an STM microscope? It doesn't seem that way. It seems that we are observing a reconstruction of an atom based on our instruments having detected voltage differences between tip position, applied voltage, and LDOS. In sum, there is a fairly long chain of inference from detection to “observation”. In such cases, perhaps there is warrant for some skepticism that our reconstruction of the differentials in voltage resembles how the world really is. The worry can elucidated through the classic Chinese thought experiment of the three blind students and the sage who teaches in a bear suit:

Having never seen their sage who wears a bear suit (unbeknownst to them) three blind students are curious to know what he is like. With the sage's consent each “observes” by touch a different part of him. The first touches his leg and exclaims “Lo! the sage is like a moss-covered young sapling”; the second holds the sage's tail and exclaims to the first, “you are mistaken, verily our sage is like a snake!”; meanwhile the third who has touched the sages paw can bear it no more, he exclaims “you are both fools! Our sage is like a furry ping pong paddle!”

Here, the anti-realist might concede that—as the blind students can about the sage--we can detect properties of the world; but how those properties are “assembled” into entities is unknowable. For this reason we should be antirealist about theoretical entities.


Realist Counter-Reply: Epistemic Qualities But it is not only the anti-realist that can appeal to the three blind students and the bear-suit- wearing sage analogy to support his position. He can propose that there's no reason to suppose that the blind students had to stop their observations of the sage when they did. They could have continued to each walk around the sage, communicated, and replicated each other's observations. There is a parallel in science. It's not as though the only line of evidence we have for atoms comes from one type of observation—e.g., an STM microscope. There are multiple lines of converging evidence that point to the existence of atoms. Just as if the three blind men had continued their observations they might eventually concluded that all three properties inhere in the same object, the same goes for observing the different properties of atoms. So, the realist can say that the atomic theory has the epistemic quality of (1) support from multiple independent lines of evidence.

The realist can further argue that theoretical entity “atoms” have several other important epistemic qualities: (2) They offer the best explanatory model for the properties and phenomena that are detected/observed. (3) Their postulation provides a reliable model from which we can predict phenomena. (4) We can apply the atomic model successfully in technology, even technology that is recursive—i.e., it helps to make further discoveries about atoms and even other theoretical entities such as subatomic particles.


Anti-Realist Counter-Counter Reply: Epistemic Qualities Don't Necessarily Entail Ontological Consequences  Van Fraassen's response to the the various arguments from epistemic qualities is that as far as theoretical entities go the epistemic qualities of the T of which they are part don't necessarily entail ontological conclusions one way or the other (p. 1066). The waste bin of history is filled with theories that had one of or some combination of (1), (2), (3), and (4) which were eventually rejected. Instead we should say that so long as T adequately explains all (directly) observable phenomena, we can say T is empirically adequate—but that is all (Ibid, pp. 1066 and 1131). Furthermore, if--as some realists suppose--scientific theories cumulatively aim at approximations of truth, realist can't explain how previous theories were “approximately true” about entities “whose central terms have evidently not referred” (Ibid, p. 1132). For example, phlogiston was proposed to explain heat and ether to explain how light travels, despite the epistemic qualities their supporting Ts at the time, they failed to refer.

We should work within the theory that, compared to its rivals, best meets the standard of empirical adequacy, but given the absence of any necessary relation from epistemic qualities to ontological consequences and historical precedent, we ought to remain agnostic about the ontology of theoretical entities.


Conclusion
For the most part I agree with van Fraassen that we ought to remain agnostic about theoretical entities (where ever the demarkation line happens to be drawn); however I disagree that we should be absolutists in our agnosticism. Just as there is a continuum between observable and unobservable entities, I suggest there's also a continuum of epistemic reasons in favour of a theory's postulated entity's likelihood of resembling the world. There will never be 100% certainty or Truth, but this doesn't preclude our talking in degrees. Also, just as it will be a combination of pragmatic considerations that we will consider when drawing the line on the observable-unobservable continuum, so too will there be pragmatic considerations regarding where we draw a the line between strength of epistemic reasons in favour of agnosticism or ontological belief. Finally, in the observable world we (sometime unconsciously) evaluate epistemic and pragmatic reasons in inferring ontological conclusions about the world, so, I suggest we carefully extend this allowance to some theoretical entities.

Consider how the sage in the bear suit might come to ontological conclusions about the observable world. It seems unlikely that he would/does not permit himself certain basic ontological beliefs about the observable world just because in his wisdom he knows there is no necessary entailment from epistemic reasons to ontological conclusions. Why might he make ontological conclusions? I hypothesize that, amongst other beliefs, he would have a belief in the general epistemic reliability of his senses; that is, they reliability produce true beliefs about the world. Also, perhaps partly because the coherence, simplicity, and predictive and explanatory power of his beliefs/theory about the observable world. Finally, there are pragmatic reasons for the sage to conclude from his epistemic reasons that the wall he is approaching is solid. It seems odd that he would hold separate his epistemic and pragmatic reasons from his ontological conclusions about the world. In short, he has strong epistemic and pragmatic reasons to infer ontological consequences.
If we can accept this much--that in the observable world most realists and antirealists alike draw ontological conclusions from epistemological and pragmatic reasons—then, with some qualifications, there doesn't seem to be any necessary prohibition against doing the same in the theoretical realm. It does not follow from this that all theoretical entities should be granted ontological status. Science is not a unified body of knowledge and this suggests that each theory should be assessed individually because, obviously, not all theories and their postulated entities are on equal epistemic footing and/or have equally strong practical reasons in their favour.

Again, just as the antirealist can point out to the realist that the difficulty with drawing a observable-unobservable distinction doesn't entail there isn't one; the realist can reply that just because there's a continuum of strength of epistemic and practical reasons in favour of believing in ontological conclusions doesn't entail there are no difference between the extreme ends. In the theoretical realm (just as in the observational realm), there are some cases where it seems the epistemological and pragmatic reasons in favour of accepting ontological implications overwhelm those against, it would seem odd not to acknowledge their cumulative weight. Perhaps, electrons are one such entity.

Finally, the antirealist might yet respond that our beliefs and desires simply do not an ontological Truth make. And they are correct. He might further point out that even in the observable world he has good epistemic and pragmatic reasons to act as though walls are solid, but he still withholds judgment as to their ultimate reality. I propose that a some point this ontological agnosticism exceeds its utility. What do we gain from this almost pedantic ontological tiptoeing? At some point the practical considerations seem to outweigh the philosophical and--just as we do in the observable world--we ought to move from acting as though wall are solid to simply believing they are.

As a philosopher it pains me to say it, but there is a point at which we have to be willing to admit that the philosopher doth protest too much. There's a parallel with the Cartesian skeptic. He'll never be satisfied, but there is a point at which skepticism overstays its philosophical welcome and exceeds its utility5. I do not , however, purport to know where on the continuum that demarkation line ought to be drawn. Nevertheless, so long as we accept that antirealists avoid walking through walls because they have good epistemic and practical reasons for not doing so, then we can accept that at some point on the continuum, epistemic and practical reasons can tip the balance in favour of ontological conclusions about some theoretical entities.


Footnotes


1Although there are many “brands” of logical empiricism it can be generally defined as the view that if theoretical language can't be cashed out in observational language then it has no ontological or metaphysical meaning.


2My point is not that they don't consider it, but my position is that there are possible implications that are overlooked.


3At this point I'm going to assume that conclusions from epistemically reliable observation methods are valid grounds for ontological conclusions. Later in the paper I will address anti-realist realist rejection of the inference from epistemic qualities of a theory to ontological assertions about the world.


4http://en.wikipedia.org/wiki/Scanning_tunneling_microscope


5My suggestion is not that utility on its own is the arbiter of Truth, but in the observable world, even the most skeptical of philosophers considers it, along with epistemic reasons, in how they live their daily lives. Why not, in some cases, extend this to the theoretical realm?

Realism vs. Anti-Realism 2: Arguments Concerning Scientific Realism by Bas Van Fraassen

Notes and Thoughts on Arguments Concerning Scientific Realism

Summarizing quote: "The rigour of science requires that we distinguish well the undraped figure of nature itself from the gay-coloured (heh!heh!he said "gay") vesture with which we clothe it at our pleasure."
--Heinrich Hertz quoted by Ludwig Boltzmann, in turn quoted by Van Fraassen, in turn quoted by me.

Preamble
In the previous post, I introduced the realist-antirealist debate in philosophy of science.  Again, although there are many species of each, we can loosely say that the debate is over whether unobservable theoretical entities, like subatomic particles, refer to something real or are purely theoretical and we should not think of them as corresponding to some real "thing".

Historically, logical positivism was the first contemporary position in philosophy of science (the only statements that have meaning are verifiable and refer to objects we can directly perceive).  Scientific realism mainly grew out of criticism of this anti-realist position.  In this article, van Fraassen agrees with realists that we ought to reject the positivist philosophy, but disagrees that this should entail realism.  Instead, he proposes an anti-realism that he calls "constructive empiricism".  The idea is that we can accept scientific theories but remain agnostic about their truth; we require only that they be "empirically adequate" (i.e., true about observables). 

Scientific Realism and Constructive Empiricism

Scientific Realism: Vas ist das?  An unsophisticated definition is that the picture of the world that science gives us is a true one, "faithful in its details, and the entities postulated in science really exist: the advances in science are discoveries, not inventions."  There are a couple of caveats.  Scientific realists aren't necessarily committed to the view that all current scientific theories are True or that the scientific enterprise is going to finish any time soon.  But this is the gist of it.  A philosophical theory about science must answer two questions: what is a scientific theory, and what does a scientific theory do.  Realists say scientific theories are about what there really is, and science is an activity of discovery not invention.

Summarizing statement of scientific realism:  "I understand scientific realism to be the view that the theoretical statements of science are, or purport to be, true generalized descriptions of reality." (Ellis)

The advantage of this statement is that it avoids any commitment to a particular current theory being true, only that they "purport" to be.

Another formulation:  A realist (with respect to a given theory  or discourse) holds that (1) the sentences of that theory are true of false; and (2) that what makes them true or false is something external--that is to say, it is not (in general) our sense date, actual or potential, or the structure of our minds, or our language, etc...(Dummett and Putnam)

Yet another formulation:  That terms in mature scientific theories typically refer, that the theories accepted in a mature science are typically approximately true, that the same term can refer to the same thing even when it occurs in different theories--these statements are viewed by the scientific realist. . . as part of any adequate scientific description of science and its relations to its objects. (Boyd)

Final formulation:  Science aims to give us, in its theories, a literally true story of what the world is like; and acceptance of a scientific theory involves the belief that it is true. (van Fraassen's summary of all the views)

8 Quick Arguments against Realism, How it Can't Explain the Success of Science, and NKTB
In these arguments I will primarily be referring to scientific theories about  (directly) unobservable theoretical entities like atoms and subatomic particles.

Step 1.  We can have lots of fun!  Just because a theory's central terms (e.g. subatomic particles) refer to something doesn't entail that the reference will be successful.  For example, if you see the shadow of a bunny rabbit on the wall you might infer there's a bunny.  But the shadow might have been made by someone's fingers you can't see.  So, your belief that "there's a bunny" doesn't refer to the entity to which you think it does.   

Also, just because a theory is successful, doesn't mean that all or most of its central terms (entities) refer to something.  For example, in early chemistry heat was thought to be some sort of invisible fluid "phlogiston".  But it turns out that term doesn't refer to anything real.  Same goes for "ether" that was postulated to explain how light travels.  Given historical precedent, there's no reason to suppose our theories are any better off. 

Step 2.  I can 'splain it to you! The notion of "approximate truth" is too vague to permit us to judge whether a theory comprised of laws that are all approximately true would be empirically successful; that is, would it adequately describe the observed phenomena.  It seems that a theory can be empirically successful even if it is not approximately true.  Take for example Newtonian physics.  It gets the predictions right (at low relative speeds) but its theoretical explanations of time and space are not approximately true.

Step 3.  Read it with me! Related to the previous point is that realists don't have an explanation of how some past theories had theoretical terms that turned out not to refer to anything (eg. phlogiston, ether) yet were successful predictors of empirical phenomena.  That is, the theory worked (for the known phenomena and testing methods) yet it was totally wrong.  There was no phlogiston or ether.   Who's to say the same won't happen with today's theories? 

Step 4.  Tell me some more!  Realists who give a convergentist (science converges on the truth by building on previous theories) account of scientific progress can't give a good explanation of how this works.  The realist model is that the new better theory preserves some of the laws that were in its predecessor because they were approximately true.  But, obviously they weren't approximately true enough to get it right, otherwise why would the theory have been revised and replaced?  So, what does "approximately true" mean if something that we say is approximately true has to be modified?  Where's the line between a modification, a change, and a new law?  "Approximate truth" is too vague to mean anything.  How did the preceding theory's law refer to something approximately true about the world if it was getting it wrong?

Step 5.  Bees in a hive!  Realism presupposes its truth (Problem of circularity).  Realism assumes that just because a referring theory or an approximately true theory are explanatorily successful that they are true.  But this presupposes that explanatory success and truth are one and the same, but they aren't.  History is full of examples of theories that were explanatorily successful, yet false.

Step 6.  Reeeee mix! (wika wika).  Just cuz a theory does a better explanatory job than its predecessor, this doesn't imply that the new theory can explain why the others succeeded or failed.  For example, Einsteinian physics doesn't explain why Newtonian physics was approximately true.  So, this idea of convergence on truth is weakened.  In other words, if E physics can't explain why N physics got it approximately true, how can we say that N physics did get it approximately true or the E physics was built on N physics?  It might have gotten its predictions right for all the wrong reasons.  Nukin' pu nub fo all the wong weasons...

Step 7.  I know a guy named Devon.  He built his own convenience store.  He calls it "Devon-eleven".  Again, against the convergent view of science:  If a new theory's predecessor has been falsified then it is not possible for the new theory to contain either all of the predecessor's content or all of its confirmed consequences or all of its theoretical mechanisms.  The realist might reply that this is not a necessary condition on a new theory.  But the anti-realist counter is that if the realist believes scientific theories aim at referring to real things and are approximately true then he has to admit that the previous theory didn't do this if some of the postulated entities or laws have to be rejected.  And again, why assume that the new one gets it right?

Step 8.  Isn't this great?  Realists haven't given any argument for why anti-realists won't be able to explain the the success of science.  The only "argument" the realists make against antirealists is that otherwise it would be a gosh-darned miracle that our theories--which have allowed so much prediction, explanation, and beneficial application--didn't refer to real things in the world.  But that's not an argument, that's the fallacy of personal incredulity. 

Conclusion
The conclusion is that realism has not shown how it explains scientific success beyond presupposing that explanation=truth.  Throughout the history of science there have been many scientific laws that approximately explained all or most phenomena at the time, but were later shewn to be false.  Realism cannot explain their success.  On their model, those theories and the entities which they postulated were approximately true of the world.  But as we've seen they weren't.  Realism is the wrong model for interpreting what science does and what theoretical entities are (they are not approximate references to real things in the world).

On the other hand, we shouldn't go too far in the other direction.  We shouldn't say that a theories predictive and explanatory power are meaningless.  The difference might be between wanting to believe something is true and having good reasons for believing it.  That sounds right.  But I'm not clear on what other possible reasons beyond predictive and explanatory power there might be.

Realism vs Anti-Realism: Ontological Status of Theoretical Entities by Maxwell

Notes and Thoughts on "The Ontological Status of Theoretical Entities" by Grover Maxwell

Preamble
One of the big debates in philosophy of science is between realists and anti-realist.  Although there are many types of each, generally speaking we can say that realists think that our theories describe real entities while the anti-realist don't.   Here's an example to illustrate.  Consider "electrons".  When I use this word does it refer to an actual physical thing?  Or it it referring to a theoretical account of a set of observations? Or is it just a "useful fiction" to facilitate scientists' thinking but should not be thought of corresponding to any real physical entity?

A problem is that in modern science no one has ever directly observed any of atomic or subatomic particles, although we talk about them as though they're real physical things.  Consequently, our knowledge of them is only inferred from bleeps and blips on instrument panels.  "Electrons" and such are merely theoretical postulates to give a coherent account of the bleeps and blips that we count as observations.  So, the anti-realist argument goes, all these theoretical entities (atoms, subatomic particles, forces) are just that, they're theoretical, not real entities.

Maxwell's essay is a response to several anti-realist positions.  His basic argument is that the divide between observable and unobservable entities which anti-realists rely on is itself only a theoretical distinction. For example, before microscopes, germs were theoretical entities.  But with the invention of microscopes they became observable.  So, why should we suppose that just because some entities are unobservable now that they will continue to be.  Therefore, they can't depend on the unobservable/observable distinction to support their theoretical/real distinction.

Quick Definition:  "Ontological Status" is a fancy way of talking about the degree to which something exists.  For some people, the only things that have ontological status are physical things, for others ontology extends to concepts like numbers and (maybe) ideas and mental states.  There are a great many views on what sorts of things exist.  But (almost) all agree that physical things exist; that is, they have ontological status.

The Observable-Theoretical Dichotomy:  R U 4 Realz?

Vs What we see through a microscope isn't real: 
 I know this will sound kinda wack to the peeps out there who haven't studied philosophy.  But wackiness abounds in philosophy.  Lets carry on shall we?

Suppose, pre-microscopes, a scientist infers the existence of germs.  His theory appeals to unobservable "bugs" that cause infections.  Application of his theory yields great results when people start washing their hands and such, lending further support to the theory.  A few years later the microscope is invented and he's vindicated--these "bugs" are observed and given the name "microbes".

A philosopher of science, upon hearing this, denies that the microbes were ever observed under the microscope.  But why?  He argues that "when I look through a microscope, all I see is a patch of colour which creeps through the field like a shadow over a wall.  And a shadow, though real, is certainly not a physical thing."  Basically, when you look through a microscope you are not directly seeing the thing in the plate, rather you are seeing the light which bounces off the microbe which is then amplified (i.e., distorted), then you see the distorted light coming out of the lens.

What's the basic argument?  It's that the microscope distorts/amplifies the light which reflects off of the microbe, so we're not really seeing the microbe, but a distortion.  But there seems to be some arbitrariness going on.  If I look through a slightly convex window does that mean that I can't infer that what's on the other side isn't real (i.e. physical)?  What about people who wear glasses?  Are they not allowed to infer that what they're seeing is real just because the lens distorts the light?  That seems kinda wack.

Of course this is not to say that we can't draw any line.  Certainly there's a difference between inferring the physical reality of a room through a pair of glasses and inferring the Higgs-Bozon particle from the instrument panel and computers at CERN.  What are the implications to our ontology?  Do we ascribe different ontological status to things depending upon how they were observed?  Why should how they are observed affect whether they're 4 realz?

There is also the problem of people and organisms having different natural levels of resolution in their visual faculty.  Is the world less real to the coke-bottle-glasses-wearing person than it is for me with my 20/20 vision because I can see at a higher resolution?  That seems wack.  When someone gets lasik, does the world become more real for them?  There's certainly a silliness to this distinction which points to its arbitrariness. 

There's another problem about the continuum of existence.  Is an very small molecule, only visible under an electron microscope less real than a larger molecule visible under an optical microscope?  What about some salts and polymers that can have single molecules so large that they are visible to the naked eye?  Does each one have a different level of reality?

Vs Only Entities that Are in Principle Observable Are For Realz
So, this is the view that only entities that are in principle observable can be considered real.   Conversely, if an entity isn't in principle observable then it remains only a theoretical rather than real entity.  So on this view we will appeal to facts like the physiology of the eye and the properties of the postulated theoretical entities.  The "in principle" part also refers to the fact that some theories entail that the entities to which they refer are about are unobservable.

Here Maxwell makes what I think is a poor argument.  He suggests a thought experiment whereby some drug makes it possible for us to directly perceive electrons (which are theoretical entities). The mechanism is that the drug alters our perceptual apparatus such that we can perceive the electrons with some latent sense faculty--not necessarily visually.   I don't even understand what it would mean to perceive an electron with some sense beyond the ones I have.  Maybe my imagination isn't good enough.

Then he sort of dismisses this example, probably because er'body's gonna think the same thing that I did, and proposes another.  Suppose a mutant is born who is able to "observe" ultraviolet radiation, or even X-rays, in the same way we observe visible light.  But this counterexample (I don't think) does the work he needs of it.  I don't think anyone is saying that X-rays and ultraviolet radiation are in principle unobservable; in fact we know several animals and insects "perceive" ultra-violet light.  

I think there's a difference between trying to observe subatomic theoretical particles (i.e., entities) and parts of the spectrum.  I'm just not sure what it is...Maybe it's that there's a difference between the ontology of parts of the spectrum and the postulated ontology of particles.  What's the ontological status of light?  I'm totally confused by this.  When we say "that gives off red light" do we mean there's something floating through the air that causes the appearance of redness?  I don't think so.  We mean (scientifically) that the surface of the object in question absorbs all wavelength of light except the one we perceive--but there's still no red thing floating into my eye.  But when we talk about subatomic particles, we're implying reference to a physical-ish thing.  I don't know.  There's some kind of difference there.

Maybe that is it.  When we observe light, we don't give it ontological status of a physical thing, but with sub-atomic particles we do.  So, why does that matter?  I guess because his thought experiment is all about showing that the difference between things that are observable "in principle" and things that are mere theoretical entities is an illusion.  I'm gonna disagree for now.

Next, he says the matter at hand is not to speculate which theoretical entities are or aren't in principle observable.  It is the theory within which the theoretical entity is postulated that stipulates this.  So, in a  way saying "un/observable in principle" is superfluous because this fact is part of the theory.  This is the heart of the matter because it seems there is no pre-theoretical way to determine what is or isn't observable.  It's like this y'all:  peeps were rappin' 'bout 'lectrons 'til there was a theory that postulated them.  That theory wasn't someone's wacky dream, it was based on a crap-load of observations.  So, as far as theoretical entities go, it is the theory itself that defines whether the entity is observable or not.  And that's kinda cheating.  

I want to rejectify this a little.  It's true that theoretical entities like subatomic particles have their un/observability quality stipulated by that same theory.  But, that theory operates within a larger theoretical framework whose principles are that objects below a certain size are unobservable for reasons x, y, z.  Theories don't exist in isolation, and they don't emerge independent from meta and pre-theoretical contexts; these other considerations shape and constrain what types of new theories and hence new theoretical entities are possible.  So, it's not entirely circular.

Of course, Maxwell can reply that it still is circular because the parallel and meta-theoretical considerations are already defining what types of entities are in principle observable.  That maybe so, but any theory of observation, including that there are real macro physical objects instead me being plugged into the matrix, requires some pre-theoretical stipulations about what is un/observable.  That we have background theories doesn't make it a free-for-all that anything in principle observable.  At most, Maxwell can say that we should be agnostic about what types of theoretical entities are un/observable. 

He can also reemphasize the point that an entities observability has no necessary logical bearing on its existence.  That seems right.  But, (I repeating myself) it cuts both ways: an entities' unobservabilitization doesn't make it's existence any more likely either.  If anything it's a strike against it, albeit a small one.  Maybe unobservable.

Vs.  Observation Language
Background:  So, in philosophy one way we distinguish between different kinds of sentences is between observation sentences/language and theory sentences/language.  Observation language refers to statements about our observations.  Eg. "there is a red spot on the wall", or "the gauge moved from left to right", or "the cat weighs 8kg".  There are various limits on what can be included in an observation language, for the sake of argument Maxwell address a general one that allows physical object terms (chair, table) and observable predicates (big, brown, heavy, etc...).  

The main point is that all descriptive terms refer to the thing that has been observed or are reducible to something physical.  Any term in my observational language has to reference some real thing.  So, on a strict version of this, if I have never actually observed x, then the term is not observational language for me because it doesn't refer to anything I've observed.  For example, I've never seen a liger, so in an extremely restrictive observational language "liger" wouldn't be an observational term for me.  Of course this strict of a language is kind of silly.  Just because I haven't seen a liger myself shouldn't have any bearing on its ontological status.  

So, we might loosen our observational language and say that observational terms must be members of a kind, some of whose members have been observed or instances of a property some of whose instances have been observed.  In this case, I can use 'liger' because it's a kind of cat, and I've seen cats before!  Or maybe we can say that ligers have properties that I'm familiar with, so for that reason I can use 'liger' in my observational language because it refers to sets of properties that I've observed.  The main point is that our observational terms have to refer to something real.

But this formulation also runs afoul.  Sometimes an entity can be one of a kind.  Or sometimes our language is clearly observational, yet doesn't refer to anything; eg. ligers over 14' long.  There is no instance of property or kind of 'Ligers over 14'' long that could be the referent of the expression.  The way out is to make our properties more general.  We can say, well, we are familiar with cats, with the colour blue, and with things that are 14' long, so our language does reduce to properties and kinds.  

The problem is that once we loosen the restrictions this much on observational language, it's hard to see how we're going to distinguish between theoretical and observational entities.  For example, in this loose language I can say that unicorns are part of my observational language because unicorns have properties that I'm familiar with.  The same goes for sub-atomic particles--they're going to have general properties that we've experienced in real entities.  Eg. spin, velocity, mass.   So, once I've allowed properties and kinds to be very general in my observation language, on what grounds can I distinguish theoretical entities from observable ones?

What Do?
Clearly, it's important to have some kind of distinction betwixt observational and theoretical language.  We need observation sentences to either confirm theories or to refer to entities that are currently unobservable.  If there is no distinction then er'thing's just cirkalur.   So, what do?  Maxwell says that instead of basing our observations in terms that must be cashed out in physical kinds or properties, we ought to use as our base the "quickly decidable sentence".  K.  Just a little forewarning.  Shit's about to get a little loco.  Don't worry about it too much.  Keep going.  Dju cang do eet!

A quickly decidable sentence is "a singular, non-analytic sentence such that a reliable, reasonably sophisticated language user can very quickly decide whether to assert it or deny it when he is reporting on an occurrent situation."  Say wut?  Non-analytic means that the predicate isn't contained in the subject (i.e., as opposed to analytic like "a bachelor is an unmarried man").  As for the rest, I'm guessing he just means that the descriptive terms aren't too mumbo-jumboish and that so long as a person isn't a complete rube they'll be able to say of the observation that the descriptor applies or it doesn't.  

For example, if someone axes me "Do you wanna do karate in the garage?"  That's a quickly decidable sentence.  Or suppose I put on a lab coat on and do a titration in a beaker being warmed by the bunsen burner.  I have test tubes too, but they're just for effect.  Anyway, a quickly decidable sentence would be something like "the solution is a base because it turned purple and I'm wearing a lab coat".  I could quickly assent to that.  I'll be honest, I'm not too sure if this is exactly what he means.  I'll have to do a bit of research.  

Lets add one more thing to muddle our minds.  An observation term will be defined as "a descriptive term which may occur in a quickly decidable sentence".  Ok, so that kind of works with what I just said about the contents of the quickly decidable sentences not being full of mumbo-jumbo descriptors; just simple stuff like, it's purple, it's a base, etc...

We have two pieces that fit together--quickly decidable sentences and observation terms--to give us a new, non-reference based version of the observation sentence.   So, an observation sentence is a quickly decidable sentence whose only descriptive terms are observation terms.  Eg. Is that apple crunchy?  Yup.   Or "that Delorian requires 1.21 jigga watts"  Assent or dissent?  Assent in no time!

One more example.  Suppose you're driving down the freeway and you see someone mowing their lawn.  You could probably assent that they were mowing the lawn, but maybe not what colour his shirt was or the square footage of the lawn.  So, "he was mowing the lawn" would be a quickly decidable observation sentence while, "he was wearing a navy shirt with brown pants" wouldn't be.  Perhaps if you were walking by and could make more careful observations "he was wearing a navy shirt..." would become a quickly decidable sentence.  I'm not sure what Maxwell would say.

I have a problem with this.  He proposes this new theory of observational sentences but doesn't really tell us under what conditions they apply.  The closest he comes is acknowledging that determining how quickly we can evaluate the truth or falsity of an observation sentence is unknown so far.  Well, so how the crap do I know what qualifies as a legitimate observation that I can evaluate and what doesn't?  It's a totally useless criteria without this information.  It's like saying, "you're right when you're right".  Thanks, but how do I know when I'm right?

Conclusion
What was the whole point of this ruckus?  Maxwell is a realist arguing against the anti-realists who say that non-observable entities like sub-atomic particles are merely theoretical and don't exist in any real way.  The anti-realists say that atomic and subatomic particles are only theoretical entities constructed to explain indirect observations.  Maxwell's argument is that the anti-realist assumptions that the observable-unobservable line is objective and that it bears on the observable-theoretical line are false.  He showed that the line between observable and unobservable is contingent and a function of our physiology, current state of knowledge and the instruments we happen to have available.  Because where the line between observable and unobservable objects is drawn is arbitrary, it is of no relevance to an entities ontological status.  To summarize in a sentence: jus' cuz we can't see wiph our eyes don't mean it ain't fur realz.