I know my wife loves me

For the second time, the religious trope that, "One cannot know—in a scientific sense—that your wife loves you" has appeared. Since people do in fact believe they know their spouses love them, they must be employing a non-scientific method of acquiring this knowledge. Furthermore, since the belief is presumably more often accurate than chance would predict, this non-scientific method is demonstrably effective.

I'm having a "spirited" discussion with Kenneth, a Catholic intellectual at Time Immortal: Faith, Reason and Epistemology which grew out of my comments to George Jonas on Christopher Hitchens.

His comments on love:

[T]here are many things that most people, by and large, accept without substantiative empirical proof of their existence. Love is an easy example, because there’s no way I can actually ask a person to prove, with evidence, that they love me. I either accept it on faith that they do, or I don’t believe it. Or I say that there is no such thing as love…but why be silly?

The operative word here is "proof", and the conflation of proof with empirical[1] knowledge. We can see his focus on proof in his further comments:

[I]ndifference and hostility are not always hallmarks of a lack of love. Indeed, depending on the person, they can be brought out by stress, hormonal shifts, or worry. More to the point, many medications (prednisone comes readily to mind) can stimulate those exact attitudes in a person — I know from first-hand experience, precisely because someone I love dearly was once on the aforementioned drug, and would often fly into psychosis-induced rages at the drop of a hat.

The real confusing factor in many of the above criteria is that the shift is not sudden, and therefore easily identifiable as being the result of some untoward cause. Instead, the transformation is gradual — it very much looks and feels like the person is becoming steadily and consistently angrier and more resentful, and very much looks and feels like they are “falling out of love” with you. This was certainly my own experience with this young woman of whom I speak.

But she had not ceased to love me — indeed, we’re getting married in October! — but the point is that based on most normal categories, I could have reasoned that she had, in fact, ceased to love me. I would have been wrong in drawing that conclusion, though. The human being is a very complex creature on a number of levels, and sometimes seemingly unrelated changes or issues can produce effects that will be falsely identified as evidence of a wholly different thing.

So while such things as love can sometimes be falsified, the criteria we might use to determine its falseness are not ideal, nor particularly suited to all situations — they are not universal. And so we cannot conclude, in any sweeping sense, that love, as a concept, is completely falsifiable.

Kenneth (with considerable skill) makes four arguments here. The first is the pedestrian observation that finite evidence cannot constitute absolute proof. The second is the principled objection to scientific epistemology, that it is not possible to decide between specific hypotheses in isolation on the basis of evidence. The third is the assertion that people employ a degree of faith (in the strict sense of truth-apt belief contrary to the evidence or without sufficient evidence) . All three of these points are sound but irrelevant.

The fourth argument would be an attempt substantiate his original assertion that love is generally accepted "without substantiative empirical proof" of its existence, and such acceptance is justified. My responses to his other arguments will show that this assertion is, if not unacceptably vague, actually false.

The objections that Kenneth raises are sound, but they relate to the metaphysical interpretation of empiricism, not specifically to whether people do or do not in fact draw conclusions about love based on an empirical basis, or, to the extent that they do draw non-empirical conclusions, they are any more accurate than any other non-empirical conclusions.

Empirical conclusions are indeed not "proof" in the sense of rigorously deductive certainties. Empiricism is evidentiary, not deductive. As I elaborate in my series on The Scientific Method, a scientific theory is (ideally) a logical formal system from which one derives the evidence obtained; it is not a set of conclusions logically derived from the evidence. Empiricism instead is a rational decision-making procedure: Not which theory is proven, but which theory is better than another.

Likewise one cannot draw empirical conclusions only on the basis of agreement with evidence, especially regarding a single hypothesis in "isolation". As the Dead Parrot sketch so amusingly illustrates, one can always adjust the subsidiary hypotheses of a theoretical framework to maintain agreement with the evidence for a specific hypothesis: "He's pining for the fjords!" For this reason, we must employ additional criteria to make empiricism a rational, deterministic decision procedure.

Two additional criteria do the job: scope and parsimony. A theory that accounts for a large body of evidence compares favorably to a theory that accounts for only for a small subset of that evidence. Likewise a more compact theory, a theory with (depending on the mode of analysis) fewer irreducible premises or ontological objects compares favorably to a theory with more premises or objects. Everything else being equal, a theory with greater scope or that is more parsimonious is preferable to a theory with lesser scope or more extravagance.

Kenneth's third point, that people do in fact employ reasoning that is not scientifically rigorous regarding love, is not controversial. But this trivial fact is beside the point. Two questions drive more directly to the heart of Kenneth's point. Do people employ reasoning that bears little or no relationship with empirical reasoning? And, more importantly, to the greater or lesser extent that people employ reasoning that would draw conclusions different from rigorous scientific reason, are they justified in doing so?

We can see that we must answer "no" to both questions.

First, it is not an objection that people make real-world decisions on the basis of their perceptual reasoning. The normative value of empiricism does not depend on the establishment of absolute proof; it requires only that the chosen theory be better than the alternative. That people do in fact actually decide that others love them is not itself evidence that they are employing non-empirical reasoning.

I don't know about Kenneth, but most people I know draw conclusions about love at least in a context of perceptual evidence: The things that the other person says and does, evidence that could, in principle, be verified with a video camera and a tape recorder. People do not typically believe that someone with whom they have had no interaction at all actually loves them, and the phrase for people with a belief disconnected from perception is at best "deluded" and at worst "psycho stalker" (cough John Hinckley, Jr.).

Secondly, they are employing the evidence in a falsifiable manner. It is beside the point that a specific hypothesis cannot be falsified in isolation; the point is that the hypothesis that "my wife loves me" exists in a theoretical context: my notions about my wife's stateof mind, and love and human psychology in general. Evidence that contradicts my whole theory requires some adjustment to the theory rather than simply ignoring the evidence to the extent of actually denying the evidence. Again, we label people who simply deny the verifiable evidence of their senses as seriously mentally ill, not as exemplars of some alternative epistemology.

We can see such an adjustment in Kenneth's example above: "[D]epending on the person, [indifference and hostility] can be brought out by stress, hormonal shifts, or worry." Here we see that one must updating his theory, his model of the world, because his original simplistic theory was false-to-fact. It is that a falsifiable theory must be updated when falsified that makes the other criteria—scope and parsimony—meaningful.

How far do people adjust their subsidiary hypotheses or arbitrarily restrict the scope of their theories to save the fundamental hypothesis? Any ordinary person will observe that people do in fact go pretty far, especially ignoring parsimony. The battered wife who insists that, deep down, her husband really does love her—indeed anyone with a consistently and thoroughly indifferent or hostile spouse—is ignoring parsimony. And yet again, such people are not held up as exemplars of an alternative epistemology: We see them as deluded or just plain stupid.

Human psychology is complicated and its universals not well known; there are many alternative theories which can account for a range of behavior. But there is a point where such alternatives collapse of their own weight—the principles of parsimony and scope are operative. Contradiction of the canons of empirical reasoning—logic, falsifiability, scope and parsimony—does not intuitively appear to generate better success. Quite the contrary: we view such contradiction as delusion, mental illness or stupidity.

The fact that empirical reasoning does not supply absolute, certain proof, as well as the fact that human psychology is complicated and our understanding is limited, does raise the important question of sufficiency of knowledge: Absent absolute proof, how do we know when we have enough information to justify a decision?

To answer this question, we have to turn to probabilistic game theory. We have to judge the "plausibility" that our theory, which includes the hypothesis that "my wife loves me", is accurate or mistaken—i.e. whether all further evidence will support the theory or will later compel an alternative theory containing the hypothesis that "my wife does not love me" by virtue of scope or parsimony. In principle we have estimates of the prior plausibility (based on the reports of other people who have accurate or mistaken beliefs about love) so we can employ Bayes Theorem to good effect.

We also have to consider subjective value. Whether my wife does or does not love me is not an abstract proposition about which I am emotionally neutral, in the sense that I'm emotionally neutral about the hypothesis that "electrons, i.e. point particles with negative electrical charge and unit mass, exist". Having the accurate belief that my wife loves me entails considerable positive subjective value. Having the inaccurate belief that she loves me will entail considerable negative subjective value. (For the sake of simplicity, we will ignore the case that she does indeed love me now but might change her mind in the future.)

I don't want to get into all the ins and outs of game theory; in essence we multiply each value by the corresponding plausibility; if the results are equivocal, we search for more evidence; if one result dominates, we choose it.

All this sounds reductionist and a little cold-hearted, but it isn't really. Most of this reasoning actually happens, just not quite so explicitly. And people do in fact make decisions about love that can only be called "bad". Furthermore, the decisions we call bad are precisely those which are the least supported by empirical reasoning. Sometimes people get lucky and make good decisions for the wrong reasons, but people never make bad (in a probabilistic sense) decisions for the right reasons.

The prior probability (unconditioned by direct evidence) that someone might actually love one is high enough, and our intuitive beliefs about empiricism and psychology are usually accurate enough, that we don't have to be paranoid or formally rigorous about the decision to believe that others love us. On the other hand, there are people out there who will try to fool us, who will pretend love for the opportunity to exploit or from pure self-delusion; we're justified in exercising reasonable, moderate and sensible caution. And people are in fact often fooled, so it makes sense to examine the issue in a neutral, philosophical and scientific manner to learn how to make better decisions.


[1] I mean small-e empiricism, in the sense of connected in an unspecified way to perceptual experience, not big-E Empiricism, the (deprecated) philosophical position that experience can form a specifically axiomatic foundation for the logical deduction of universal truth.