Saturday, October 18, 2014

Stories, Models, and Miracles

This one is for Andrew Gelman, with whom I've been having a number of interesting conversations about the sense in which models are to statistics what stories are to narratives. In the social sciences, we might say it depends on how we construe "the people": either as a "population" to be interpreted statistically, or as a "history" to be interpreted narratively. In the first case we build models and test them, in the second we tell stories and ... well, see how they go over. Models are beholden to probabilities, stories to plausibility. That sort of thing.

When I read about Michael Shermer's "anomalous" wedding music, I thought of Andrew right away. After all, he's been consistently critical of Daryl Bem's ESP research but he likes a good story as much as the next guy. So let me tell the story, and then explain what I think it has to do with our scientific models.

Shermer got married this summer. His fiancé, now wife, Jennifer Graf, had shipped her things to Beverley Hills from Köln, including a number of heirlooms from her grandfather, who had been very important to her as a child. Among the items was a transistor radio that Shermer tried and failed to get to work. Three months later, in the days leading up the wedding, "Jennifer was feeling amiss and lonely. She wished her grandfather were there to give her away." Then, on the day of the wedding, a strange thing happened: the radio suddenly started working, "a romantic love song wafted" from it. The next day, it stopped again, and it hasn't worked since.

The moral of this story, of course, is that any old radio will do. No, I'm kidding. It is, of course, as Shermer tells us, that Graf's grandfather was there with them after all, had in effect given her away. The moral, indeed, is, as she is to have said, "I am not alone." And, because Michael Shermer is a world-famous "skeptic" and champion of science, this story has additional significance. "[T]he eerie conjunction of these deeply evocative events gave [Jennifer] the distinct feeling that her grandfather was there and that the music was his gift of approval. I have to admit, it rocked me back on my heels and shook my skepticism to its core as well." The moral of the story, that is, is that, modern science and skepticism notwithstanding, it is possible for the dead to communicate their opinions to the living.

Charles Eisenstein, writing for the new-age magazine Reality Sandwich (which was founded by Daniel Pinchbeck, by the way), has declared Shermer's conversion a "miracle". (He is more impressed with the fact that this happened to Michael Shermer than that it happened at all, presumably because he believes this sort of thing happens all the time.) Perhaps, he suggests, this "portend[s] a fundamental transition. Perhaps it signals the unraveling of the epistemologic hegemony of science." Shermer is indeed among the more prominent ideologues of science, so maybe it does, but there's another sense in which this story, even when it happens to ordinary people might shake our scientific models "to the core".

Consider the common reaction in comment fields and skeptics' forums. Actually, consider Shermer's own remark (which his critics do little more than echo):

Had it happened to someone else I might suggest a chance electrical anomaly and the law of large numbers as an explanation—with billions of people having billions of experiences every day, there's bound to be a handful of extremely unlikely events that stand out in their timing and meaning. In any case, such anecdotes do not constitute scientific evidence that the dead survive or that they can communicate with us via electronic equipment.

This is the point I want to emphasise. We have no working model of the causal mechanism here, but we have a story that gets its significance from the (unproven) premise that the dead can communicate with the living—that they care about us, and that they approve and disapprove of our choices. That's a powerful narrative with deep moral implications.

If a "belief in science" means that we must reject this interpretation of the story, impoverishing Michael and Jennifer's memory of their wedding, and diminishing the significance of their marriage vows, then surely science is a somewhat petty business. Indeed, it comes to function as a rather mean and spiteful god, preventing us from enjoying the wonders that befall us, robbing our lives of meaning. But if science will not step in here and rule out the spiritistic interpretation, then where does that leave us? Aren't we then always entitled to go with whatever interpretation feels right? The obvious counter-example that comes to mind here is vaccination, the case against which is based on coincidences between getting the shots and developing autism. If we believe the models that show us that vaccines do not cause autism, aren't Michael Shermer and Jennifer Graf likewise condemned to live in a world where their beautiful wedding story just ain't true? I, for one, don't have the heart to tell them.

23 comments:

Andrew Gelman said...

Hey, I have an even better story. It happened to me nearly 30 years ago. I went into a cafe in Harvard Square and I saw this guy, the back of his head looked just like Michael Keaton. And I'm a big Michael Keaton fan. I don't know what this all means, but it's shaken my faith in science to the core. I'm thinking maybe Heisenberg's uncertainty principle isn't true, but I'm not sure. The problem might be with the germ theory of disease, or the idea of evolution by natural selection. But there's definitely something otherworldly going on.

Jonathan said...

Surely "science" tells us that people do not have strong emotional reactions, and almost never draw meaning from seemingly random events.

Presskorn said...

I am much more symphathetic to the Shermer piece. The piece clearly states that he told himself the explanations that T.B. writes he doesn't have the heart to tell him.

And Shermer is not without right to say that he "savours the experience more than the explanation". All of us do that most of the time (when falling in love, when listing to music, when drinking wine, when reflecting on ours pasts, on our children etc.).

What he would wrong to be to say is that this experience has "shaken his faith in science to the core" as Gelman writes. But notice that Shermer doesn't quite say that (even if his phrases are flirtingly close to that). He would wrong to something like that because it would seem imply that the assignment of emotional significance to events contradicts causal accounts of the same events.

But he goes on to say that is exactly *not* his point: "The emotional interpretations of such anomalous events grant them significance regardless of their causal account."

And it's not quite right to say with Thomas Basbøll that this leaves us in a situation where we are "entitled to go with whatever interpretation [that] feels right". Rather, it is a logical and epistemological fact that the same set of events may have differing descriptions that do NOT contradict each other. So the right thing to say is that we simply do not have to choose between an emotional and a causal description.

This is not a case of having your cake and eating it too. There are two cakes to be had.

PS: The vaccine case is NOT any sort of counter-example either. The autism hypothesis is merely an example of a plainly wrong scientific hypothesis concerning vaccnies. But it does not follow from the incorrectness of that hypothesis that parents have no right assign all sorts of emotion (postive or negative) to vaccination of their children.

PSS: I just read the post again, and perhaps I'm having trouble figuring out exactly what you are arguing, Thomas. But on my first reading, it's clearly a non sequitur...

Thomas said...

Andrew and Jonathan's (let's say) absurdist responses made me doubt myself too. I didn't intend my post to be a non sequitur.

I'm not sure I can see the difference between A and B here:

Story A: My grandfather's radio starting playing on my wedding day.
Model A: The dead can communicate with the living.

Story B: I vaccinated my child and she developed autism.
Model B: Vaccines can cause autism.

(Note the "can" in both models. I.e., they model possibility, not causality.) The scientific evidence, of course, belies both models, though both stories are true (i.e., such things do happen).

I think Shermer is saying his story A includes the model. The story is: "My wife's dead grandfather communicated his approval to us on our wedding day." Anti-vax people, likewise, tell the story with the model's interpretation, as it were, "built it": "My child's autism was caused by the vaccination."

That is, as I read it, Shermer's story depends on a plainly wrong hypothesis about spiritistic media.

"Science", that is, tells us not to connect those two dots, and if we don't the emotion we might feel is different. I don't really buy the idea that we're entitled to feel whatever we like and think whatever we like besides. What we feel depends, in part, on what we think happened.

And my basic intuition is that science should leave Jennifer and Michael alone. I don't feel the same way about the vaccination case. But I'm not sure why.

Presskorn said...

I not exactly sure that Shermer cuts the cake right. Especially since I am insisting that there are two cakes. Your two A and B examples make it much easier though, since there is a clear distinction between them. You’re at perfectly right in saying: “And my basic intuition is that science should leave Jennifer and Michael alone. I don't feel the same way about the vaccination case. But I'm not sure why.” But Model A and Model B makes the “why” perfectly explicit – apparently without you knowing it (it has e.g. very little to do with causality versus possibility). Here’s the difference:

We know exactly what sort of evidence that would count as confirming "Model B: Vaccines can cause autism". If we go looking for such evidence, however, we find none. Perhaps we even find falsifying empirical evidence. Model B is scientifically irrational.

On other hand hand, we have NO idea what would even count of as empirical confirmation of “Model A: The dead can communicate with the living.” We might have some idea what would the purpose of saying something like that in context.* But we have no idea where to go look for evidence that would confirm that as a scientific hypothesis. In fact, it’s just not scientific at all. It’s not scientifically irrational like Model B. It’s arational. Or perhaps even better: abrational. But that also means that science has no grounds on which to contradict it (contra Dawkins and the rest of the anti-religious choir).

One could of course make a whole of series of more muddy examples, but I think your two models make the distinction come out clearly. It is somewhat analogous the Kitzmiller versus Dover Area School District case concerning ID that Fuller testified in. Concerning that case, my philosophical (not juridical) opinion would be the following: If you teach ID in the biology class, you cannot teach it in religion (the good people of Dower apparently did both). And if you teach creationism in the religion class (which they no doubt did!), you cannot teach it in the biology class.

My stance is basically: It’s all good fun, but don’t mix beer and red wine….

* We can also legitimately make empirical second-order descriptions concerning such beliefs. For instance, concerning their relative prevalence, e.g the statement “The wifes of Norwegian fisherman working under dangerous conditions are more likely to believe in the notion of fate than the average Norwegian male living in Oslo” is a perfectly good scientific and verifiable empirical statement.

Andrew Gelman said...

Another complication is that a lot of people are superstitious. I know superstitious people myself!

We tend to assume that since Shermer labels himself as a skeptic, that he is not superstitious and should "know better" than to believe in ghosts. But maybe that's a misguided view of Shermer. Perhaps he has strong superstitious feelings, but has been convinced that those feelings are unscientific hence his career as a skeptical journalist, but the superstition is still there and bubbles up from time to time.

That's just a story, of course, but my point is that it's natural to interpret the Shermer story in terms of some frame or another. Perhaps the frame "prominent skeptic is stunned by a coincidence" isn't the best way to interpret this story.

I do think all this discussion is far more interesting than Shermer's presentation of the story. Yet another case of stone soup and another victory for the blogosphere. It just makes me sad that this blog has so few readers . . .

Thomas said...

@Presskorn: The whole point of Bem's research is to make that sort of hypothesis look scientifically testable. (I once spent some time hanging out with parapsychologists in Freiburg, at a center endowed, as I recall, by money from an industrialist who died in the 1930s and wanted to improve the odds that his wife would be able to communicate with him beyond the grave. In a sense, an R&D project in mediumship. Interestingly, they were very honest about what they thought the data on ESP showed. Nothing.) Mediumship, like ESP, like telekinesis, is empirically false. But I've seen a lot of good, perfectly scientific experiments, that demonstrate this.

@Andrew: Fortunately, the stats for this blog are marked, periodically, by what I call the Gelman Bump.

Presskorn said...

@Thomas: If I consistently hear the voices of dead people, it counts as confirmation of a psychopathological condition, not as confirmation of the communicative abilities of dead people (among things because the phrase "the communicative abilities of dead people" lacks not only proper reference but also clear sense). And if I look at my coffee mug now, merely saying to myself that I would like it move, and it doesn’t, then I don’t take myself to have made an experiment falsifying telekinesis. I don’t take myself to have carried out any sort of “experiment” at all.

I am sure that the good people at Freiburg have tried to devise more refined set-ups, but I’ll gladly that dare the a priori statement, that the interpretation of any such “experiment” will remain highly ambiguous as to what its results confirm. In fact, if they insist on calling them proper “experiments”, I will also gladly denounce them as scientifically irrational. But Shermer hasn’t suggested anything along these lines either, right?

(cf. Wittgenstein’s and Kierkegaard’s contempt at the seldom but occurring Catholic practice of trying to empirically confirm miracles. Such practices make a mockery of religion to the extent that I would revoke their status as specifically religious practices. They are rather half flawed science and half flawed religion. And makes them flawed is not merely their poor rate of successful empirical confirmation. They are conceptually confused.)

Thomas said...

Can we not imagine an experiment in which thousands of people are given a number between one and ten to communicate to a loved one from beyond the grave. (To ensure that no fraud occurs, the number can be assigned AFTER the death of the sender. The sender just has to be told before dying where to pick up their assignment--presumably the ghost just goes and has a look, right?) After a year, the "receiver" is asked to "guess" what number their grandfather had been assigned.

In Freiburg, an experiment in precognition was described to me that worked just a like memory test in reverse. That is, a series of numbers was shown on the screen, one at a time. Where in the memory test you would have been asked whether this is the same number you saw, say, ten numbers ago, in the precognition experiment you're asked whether this is same number you will see ten numbers up ahead.

You get thousands of people to do this for thousands of numbers and that gives you a data set to analyse. What's so a priori unscientific about that?

Thomas said...

I agree with you BTW about miracles. You can't test them scientifically and show that they're "true". But the very idea of a such a test already tells us that there are no miracles, only coincidences ... or as yet unknown laws of nature.

"Psi" researchers aren't religious (at least not essentially). They really believe in a (perhaps underdeveloped) sense (and action) modality that can be studied and even improved by training. It's subtle, but not miraculous.

This is one of the reasons they're often offended by James Randi's Million Dollar Challenge. The ability they want to demonstrate is not, ultimately, "paranormal", and does not violate (they think) the laws of nature.

Presskorn said...

The brief answer is that what makes it a priori unscientific is the belief that anything like that could follow from a data set like that. We simply have no proper concept of the cause that the results of the data set are supposedly correlated with. Nor do we have any concept of the supposed type of causal relation that would hold between cause and effect. I.e. we have unclear concepts of both the relation and its relata. Unlike the vaccine/autism case, where we have clear concepts of both, can independently verify of both and have a good working idea of the relation that might (but in fact doesn’t) hold between them.

PS: In Kuhnian, we might say that such experiments are devoid of sense (taken *as* experiments), since they lack a paradigm within which to be sensible, i.e. a framework within which where they can be taken *as* experiments.

PSS: In Wittgensteinian, we could say that science requires a higher degree of conceptual clarity than the level required to wonder at what happen at a weeding or to reflect on our lost ones. Science belongs to the regimented streets in the outskirts of the city of words.

PSSS: Also, while I admit that your description of the precognition “experiment” fares slightly better, your description of the first experiments borders on being self-defeating: “To ensure that no fraud occurs, the number can be assigned AFTER the death of the sender. The sender just has to be told before dying where to pick up their assignment--presumably the ghost just goes and has a look, right?”… Really?!

Thomas said...

You're going to force me to write a science fiction story about this!

Quick question for reference: would say that SETI is unscientific in the same sense? I.e., we're looking for a signal from "someone" with whom we've never communicated before.

Thomas said...

"Pseudoscience fiction"

Thomas said...

"psi-fi"

Presskorn said...

Your other comment touches on the political aspects (questions of recognition and so forth) and here I admit that my position runs the risk sounding "dogmatic". But there must be a way of avoiding blind doctrine - i.e. to remain open to testing, say, conspiracy theories of 9/11 - and yet to leave other sorts of "experiments" simply out of the question. There are, after all, vast differences between the two.

Presskorn said...

Totally off the cough on SETI before returning the Wittgenstein entry: No, SETI seems alright: We know who we are, we have a rough idea of what another intelligent life-form might be like (i.e. a life-form which is in relevant aspects like us… and alive!) and we have a working idea (which may however not actually work) on how we might communicate, e.g. through the visible signs on the golden plates onboard Voyager or through digital or analog (radio) communication...

Thomas said...

"There are, after all, vast differences between the two."

This begs the question.

I agree with you that there is a difference. I'm just not at all sure what it is.

Consider:

"There is no evidence for ESP."

"There is no evidence for ETI."

In neither case is it for lack of trying, i.e., lack of "data". There's lots of data. There's just no signal in it, at least none that we've yet found.

Why do you want to say that one of these claims is about the current state of a "scientific" project, while the other one isn't?

Presskorn said...

It is only begs the question, since you seemingly don't accept my account of the difference. But all my previous responses to your post have been trying to spell out this difference between, say:

"There is no evidence that the KGB had any involvement in 9/11"

and

"There is no evidence that my dead grandfather is talking to me."

That there is a difference is, I think, is "in plain view" (to speak like Wittgenstein). And as you've just stated, you agree that there is indeed a difference. But in that case, why would you want to stick an account that ostensibly cannot makes sense of the difference? Your admission of difference is really a reductio of the account.

And I don't get the account either; it seems overly empiricist in its focus on data and data alone (Unlike my appeal which includes semantic and Kuhnian considerations, while accepting that experience is the tribunal of truth - considerations broadly alike to McDowell's in Mind and World). But even an extreme empiricist like Quine accepted that even if, as he held, all of our scientific theories are revisable, they still confront experience as a unified whole. It's a web of theories that confronts experience and data; not singular claims.

So the Quinian answer (which is not my answer given in above responses) would be: One of the claims is about the present state of a scientific project, since it is consistent with a good number of other reliable theories. The other is not, because it is inconsistent with innumerable other theories. Perhaps that account could give you some peace of mind?

PS: I am not exactly sure how it relates, but our discussion made me think of a little exchange that occured the seminar with McDowell in CPH last fall. One of the participants said concerning some troubling aspect of an advanced theory about children's language acquisition: "I am not embarrased to say that I don't know how to explain this." McDowell responded in dry tone of voice: "You ought to be embarrased, if you're invoking magic."...

Thomas said...

Here's another experiment. (This is fun to think about.)

Two random number generators are set up by "skeptical" researchers, i.e., people who do not believe in the transmission of signals from beyond the grave.

Each RNG generates a stream of 0s and 1s dependent on a quantum fluctuation of some kind, i.e., a completely indeterministic process that can only be affected by, well, "magic".

Every minute a computer averages the values generated by each RNG and rounds to the nearest whole number, yielding a 0 or a 1.

On a screen the following data is displayed:

1. RNG A's most recent average:
-----
2. RNG A's average the minute prior to the most recent
3. RNG B's most recent average
------
4. The correlation between 2. and 3. over the last 24 hours.

The nul hypothesis is that there will no correlation between 2. and 3. (both processes are completely random and independent of each other.)

A select group of ageing believers in spiritistic communication (with an expressed desire to prove that it is possible) is shown the device, running and confirming the nul hypothesis. They are challenged to try to increase the correlation between 2. and 3. *after they die* by using the information in 1 as guide to their manipulation of the quantum fluctuation in RNG B.

What they need to do, that is, is produce a "significant" effect on 4. at some point after they die.

Since the machine is establishing a baseline of presumably very rare events, "signals", at 4 (the result of either chance or random glitches) in the years before anyone dies, the hypothesis to be tested is whether the “anomalies” increase notably after people begin to die and, presumably, start hacking into the quantum fluctuation in RNG B. The more people who die, the stronger the signal.

This sort of set up is pretty orthodox (or at least once was) in the Psi research community—it’s “paradigmatic”. The results of such experiments (last I checked) confirm the null, i.e., no effect. Though people like Dean Radin claim otherwise, based on rather obvious cherry-picking in the data stream.

Long comment. But my view is that under those circumstances we'd have essentially set up a SETI-type situation. No positive results. Lots of data.

Presskorn said...

It's very ingenious example, I have to admit. But raw data is not enough for science; you also need a series of plausible possible inferences that can make the data "speak". And here my point would still be (as suggested earlier) that it is totally unclear exactly what inferences that would follow, even if the contrafactual of a significant effect on 4., presumably against everyone's expection, would obtain.

If the good people in Freiburg now want to say that according to Occams Razor, the best hypothesis consistent with a significant effect on 4. is that "Dead people stay alive in some sense even after they die, but with the significant difference that they now acquire powers that can effect quantum mechanics", well, then, I wouldn't know what to say...

Thomas said...

As I understand it, parapsychologists believe that quantum mechanics is the study of the "the dreams that stuff is made of" and they believe, citing scientists themselves that this is a mainstream scientific view. Just as aliens would "probably" use certain frequencies of radio waves to communicate with us, so too would ghosts probably communicate through quantum indeterminacy. How else you gonna explain Shermer's radio?

For the record: I think both ideas are absurd, but I actually think it more likely that we will ultimately communicate with aliens through the "subspace" (quantum) ether (of some future physics) than that we'll pick up the radio waves that some idiotic pre-civilised species like ourselves sent out into space thousands of years ago.

So, yes, there is a "theory". And if there ever were a positive "contact" I'm sure that theory would have to be modified to capture what actually happened. (Again, I'm sure the aliens would have to explain that our original ideas about how all this communication works is a bit more sophisticated than the ham radios we're imagining them with.)

Thomas said...

PS: I'm not sure it's true that Shermer "savours the experience more than the explanation". The experience essentially includes the explanation. The story isn't interesting if it is interpreted merely factually, i.e., as a coincidence. It has to be a strange coincidence, and the strangeness depends on the possibility that the dead communicate with the living.

Lily Wilson said...

I don't see anything wrong about accepting both science and miracles in our lives. Science is what makes miracles so very special: if it was not for countless rules and laws of nature, we would never really see any miracles.