POSTDICTIVE.

When I copyedit dictionaries, I find lots of material for LH. When, as now, I edit a collection of historiographical articles, it’s slim pickings, but I did run across an interesting term that was new to me. In one article, the author used the word “postdictive”; I was all set to write a query suggesting an actual word be used instead, but Google informed me it is in fact a word, though not a common one. As this Wikipedia article explains, retrodiction or postdiction “is the act of making a ‘prediction’ about the past. … One speculates about uncertain events in the more distant past so that the theory would have predicted a known event in the less distant past. … Postdiction, in a slightly different sense, is used to evaluate speculative theories such as those formulated by theoretical physicists. In this case it refers to predicting known (but not necessarily past) events.” An awkward term for a useful concept.
Addendum. See this 2011 post for a nice use of postdiction.

Comments

  1. rootlesscosmo says:

    I’ve seen “retrodict,” though not “postdict,” in writing about paleontology, for example where the existence of an intermediate species in an evolutionary sequence can be hypothesized from its ancestor and descendant species, despite the absence of direct evidence; if the evidence is found, the retrodiction is confirmed. The speciation event was in the past, while the discovery event was in the future when the hypothesis was put forward.

  2. Cairnarvon says:

    “Retrodict” is pretty common in scientific works, but the only time I’ve seen anyone use “postdict” is when they were trying to remember “retrodict” but failed.

  3. “Predicting the past” is a common and necessary exercise for anyone doing computer modeling of complex “real-world” systems– A very common saying among people who do simulation for a living is “It’s hard to make predictions, particularly about the future.”

  4. dearieme says:

    It’s akin to that well known medical instrument, key to successful diagnoses, the retrospectoscope.

  5. A very common saying among people who do simulation for a living is “It’s hard to make predictions, particularly about the future.”
    Are you implying that people who do simulation for a living are fond of banalities ? It wouldn’t surprise me, since banality is easy to reproduce.
    Simulation is a new word for the old art-imitating-life thing, the fight against the cosmic ravages of time and age. A restorative daub of paint, a dab of skin moisturizer, a tweak to the imaging algorithm – these are the concerns of cosmeticians. Some French philosophers study the fractal universe, possibly with a view to giving the curly bits more body, or ironing them flat.

  6. I guess that remarks on “the difficulty of getting it right” are banal… On the other hand, I knew a mathematician who spent a few weeks copying tables of numbers in order to get valid statistics on his own error rates. Is that banal too?
    And I’ll admit that seeing “French philosopher”, “simulation”, and “fractal” all in the same paragraph gave me a moment of blind rage, but I’m OK now. My own view is that a simulation is a kind of clock– where ‘tick’ and ‘tock’ have grown up and gotten a job.

  7. So you don’t agree that the goal of “simulation” is to imitate ?

  8. No, ‘imitate’ isn’t quite right. For example, a simulation of a physical system integrates some law of motion (Newton, Maxwell, Navier-Stokes, Schrodinger), and produces a good-enough approximation of the solution in a particular case. The usual metaphor is that one ‘unwinds’ a general law onto the particular initial conditions and parameters that define a specific case of interest. There’s a shock of recognition (‘this isn’t just mathematics, it’s what happens in the real world’) that marks the boundary between success and failure.

  9. Here’s an example, from the LRB blog:

    Brown was asked about VAT at a press conference. His reply was out of Monty Python. It was in the same league as Clinton’s ‘it depends what the meaning of “is” is.’ Brown said, ‘I can give you an absolute assurance that we have not raised VAT since 1997.’ That’s beyond parody. Pressed on one of the biggest questions facing the economy, Brown reached deep down into his ability to tell uncomfortable truths and gave a backwards pledge that he hadn’t raised VAT in the past. And hey – it’s guaranteed.

    .

  10. The only difference I can see between simulation as you describe it, and imitation, is
    1. imitation is often thought of as a static product, or a process leading to such a product
    2. simulation is often thought of as a dynamic process, with no static product
    3. imitation is often thought of as second-best, whereas simulation is cool, high-tech and novelty-producing
    You stress the whiz-kid-blinded-by-science aspects of simulation. I myself think simulation is amusing, but not the most interesting game in town. Can you give a concrete example of some physical simulation project that you find particularly interesting, in which no imitation (you might prefer “dynamic approximation”) is involved ?
    I knew a mathematician who spent a few weeks copying tables of numbers in order to get valid statistics on his own error rates. Is that banal too?
    By “banalities” I mean not activities, but remarks such as “it’s hard to predict the future”.

  11. I think ‘simulate’ and ‘imitate’ are pretty much the same thing, in this context. But simulation is the word scientists have settled on, perhaps because imitation seems to carry more than a hint of fakery (as in cubic zirconia being imitation diamonds).
    Like it or not, simulation is an essential element of many scientific endeavors, necessitated by the nature of the problems. To take a topical example, analysis of results from the LHC requires a great deal of simulation of collisions, in order to infer from the wreckage the nature of the collisions. Physicists would love to have a tidier way to do it, but they haven’t found one.

  12. On the general question of the difference between imitation and simulation– I think your off-hand comment about ‘novelty-producing’ is a non-trivial clue. Since I’m already persuaded that the laws of nature are true, I want to go from the basic laws to new information about the properties of specific cases– simulations are useful tools for finding new information.
    A specific and common technical use of simulation is in development of error models. Given, e.g., various distributions of initial conditions, what is the distribution of final states? And then, given a sensor model, what is the distribution of measured quantities? This sort of analysis doesn’t seem to me to be ‘imitation’ of anything.
    FWIW, mechanical engineers take this sort of analysis to great lengths. And there’s something to be said for being able to evaluate the stability and probable failure modes of structures in orbit around the Earth without having to build them and send them up on rockets.

  13. John Emerson says:

    The people who say “It’s hard to predict the future” are making a joke, Stu, not expressive a profound truth.
    One use of “retrodict” is to describe “just so stories”, when people take something that’s already happened, concoct a causal explanation, and then claim that they’ve discovered a law valid for the future. This use of the word is pejorative, and it applies to a lot of past reasoning in social science and history.
    There’s a similar method done with an awareness of the weaknesses of that one, which consists of taking a present state, whatever data there is about past states, and then brainstorming the various ways the transition A to B might have been done. It sounds like the same thing, but the difference is in knowing that this is a speculative activity, allowing that there might be more than one path from A to B, allowing that A might have led somewhere else than B, and allowing that B might be reached from some other starting point than A. This method is really the destruction of simple-minded historical explanations and the construction of a historical description more respectful of contingence and more aware of the deficiencies in the data.
    A simulation sets up an original state and runs multiple tests to see what outcome are likely to emerge from it, given contingency and chance. The description of the original initial state can be changed and a new set of tests run to see what the possibilities of the new initial state are. These are fictions, but they can be very illuminating, for example in revealing unexpected possibilities (which might be disaster or might be benefits in various contexts).
    This is really all to the good. Historians have not always been careful about using causal language, and if you read history you run into a lot of crappy causal explanations. My favorite is Toynbee’s “The nomad invasions were caused by climate change”, which is utter nonsense, but there are a lot of them. You also get them in pop ev psych, pop psych, pop sociology, and a lot of the thinking about social problems.

  14. John Emerson says:

    This method is really the destruction of simple-minded historical explanations and ALLOWS the construction of a historical description more respectful of contingency and more aware of the deficiencies in the data.

  15. marie-lucie says:

    Put in simple terms, I guess imitation is copying something that already exists (whether static or dynamic), while simulation is trying to set up a model of a situation and studying the potential consequences of it, both as it is and after slightly tweaking some of the elements or parameters. Therefore, unlike plain imitation, simulation is a creative process, indispensable in research of many kinds.

  16. Except, of course, as regards that little word “model”. That’s the imitative or approximation part, the rest is experiment and application. Parametrization can be regarded as tooling used to model a model, as in the paper linked by MattF.
    Imitation and modeling are not necessarily trivial pursuits. Remember, for example, the discovery of perspective, those depictions by the artists themselves of the wooden frames they set up to figure out how three-dimensional bodies can be projected onto two dimensions. A more modern example is economic modeling and simulation. Much of the “new information” gained from all that turns out to have been worthless, as repeated economic crises have demonstrated.
    Or take the model of brains as computers. Many people knew by experimental results even 50 years ago that this model is worthless, and this knowledge is now a commonplace among neuroscientists. Nevertheless brain-as-computer is still central to popular presentations. Another example is DNA-as-genetic-code, and the genome as computer program. Well, it turns out that these things are not usefully understood as programs, since they modify themselves and there is no CPU.
    So how closely the model approximates (imitates) the world is still a crucial issue, and always will be. My insistence on the word “imitation” was just a home-made experiment to bring out how obsessed so many people are by “novelty”. I did not introduce “novelty” as an off-hand comment.

  17. Did they have computers 50 years ago?

  18. Hmmm. Apparently yes, although Wikipedia says: “no single device can be identified as the earliest computer, partly because of the inconsistent application of that term”.
    What I meant, of course, was: Was the mind being compared to a “computer” 50 years ago, and if so, was the device it was being compared to worthy of the comparison? Just curious.

  19. They did, we didn’t. See the WiPe article on “computer”.

  20. Was the mind being compared to a “computer” 50 years ago
    See the WiPe article on John von Neumann:

    In 1955, von Neumann was diagnosed with what was either bone or pancreatic cancer.[10] While he was in the hospital he wrote a short monograph, The Computer and the Brain, observing that the basic computing hardware of the brain indicated a different methodology than the one used in developing the computer.

  21. Perhaps I should have written above “how closely the model (whatever that is) approximates the world (whatever that is) is still a crucial issue”. Many people have thought usefully about what is involved when we do things like science, other people have done useful science without thinking about what they are doing. Thinking without acting, and acting without thinking, are not good or bad things in themselves. But all the boastful. jumping-the-gun, statement-of-intent chortling by scientists about “novelty” can be rather tiresome.

  22. Bathrobe says:

    Stu, I just wasted 40 minutes looking at web pages about “The brain isn’t a computer”. Well, maybe not wasted — it was fascinating. But I’ll pay for those moments of intellectual pleasure…

  23. Bathrobe, your moniker and the German Bad for bath just reminded me of the name of a store for high-tone bathroom fittings here in Cologne: Bad Design. I don’t think they will ever penetrate the international market.

  24. John Emerson says:

    Much of the “new information” gained from all that turns out to have been worthless, as repeated economic crises have demonstrated.
    Too broad. The nature of the problem is still under debate. I think, and I think that there’s increasing agreement about this, that the equilibrium models introduced by Samuelson and Friedman and dominant for decades were just plain wrong, but this has nothing to do with the question of computational economics (running multiple trials of models.) I can’t be sure, but I suspect that the simulators were more likely to point to the weak spots of that theory than to support it.
    The big crash seems to have been based on too much confidence in the equilibrium models (“The Great Moderation”, ha ha ha) together with the development of new financial instruments that few people understood.

  25. I think that there’s increasing agreement about this, that the equilibrium models introduced by Samuelson and Friedman and dominant for decades were just plain wrong, but this has nothing to do with the question of computational economics (running multiple trials of models.)
    It has everything to do with it. When you run multiple trials of a model that is “just plain wrong”, then you get a lot of information that is worthless. But how do you determine that a model is “just plain wrong” if you don’t generate predictions from it and track what actually happens against those predictions ? Do you mean “theoretically wrong” ?
    I can’t be sure, but I suspect that the simulators were more likely to point to the weak spots of that theory than to support it. … too much confidence in the equilibrium models … development of new financial instruments that few people understood
    If PRACTICING economists didn’t validate their models aggressively, nor even understand them, then do they deserve to be called economists ? If THEORETICAL economists didn’t do this either, do they deserve to be called economists, rather than (say) arcane mathematicians ? Your remarks are close in spirit to the stalling excuses used, for example, over decades by communists to justify the poor performance of state economies such as the Soviet Union, Cuba, the GDR – “we just need more corrections and more patience. There is nothing fundamentally wrong with our model.” One might just as well say: “it’s bigger than both of us”, and light a votive candle.
    All I can be sure of is that mathematical economics may have produced a lot of cool mathematics, but no economics.

  26. John Emerson says:

    Stu, you’re talking through your hat. Did I say that they were running multiple trials of those same models? And I didn’t quite say that multiple trials were used to test models to see how they worked (since experiments on human societies are not possible and since “natural experiments” are rare.) Multiple trials can be used to refute models, and they can also be sued to develop models which more closely approximate what actually seems to have happened.
    I’m not even sure what you’re saying now, and I doubt that you know what I’m saying. One of the results of computational economics (multiple trials of simulations) has been to weaken the old, unrealistic, purely mathematical models. But you seem to want to reject everything that uses mathematics or computers, or seemingly, any attempt to understand economic data.
    Economics falls into about ten schools, of which about three are orthodox or mainstream and the rest are ignored outcasts barely hanging on by their fingernails. The orthodoxy has been discredited and the field is in turmoil, but I’m not going to say that there’s nothing in any of the ten schools. The stuff we’re talking about, or I am anyway, computational economics, sometimes is called a school but it really is more like a method and can be used by any school.

  27. Of course I’m talking thought my hat, John. Listening to myself talking, I find a hat adds a pleasingly subdued quality to what I say, like a mute for a trombone. I’ve never found a hat big enough to fit on my head, so I might as well talk through one.
    It appears that I didn’t understand what you meant by “computational economics”. I had taken it to mean “theoretical economics involving running mathematical models on computers”.
    God knows I have nothing against mathematics. But I am simply not impressed by hand-waving policy clowns who appeal to supposedly difficult matters of any kind – not even, and especially not, mathematics and statistics – in order to save their butts. For instance to distract attention from reponsibility – and be it the responsibility of a nation – for driving an economy yet again to the brink of disaster. The problem is not economics, whatever that may be, but hubris.
    Economics falls into about ten schools, you say !? That makes it sound more like psychoanalysis than a hard science. I didn’t know things were quite that bad.

  28. John Emerson says:

    Grumbly, I’ve been ranting against economics for a decade or more. I’m the official economics nihilist over at Brad deLong’s site. I just disagree with parts of your own rant.
    A lot of the old mathematical economics used a lot of simplifying assumptions to make the math manageable. As a result, they came up with idealized, unrealistic models which worked much of the time, but not enough. Computational economics allows you to run calculations thousands or millions of times while tweaking the functions that were simplified out of the original unrealistic equation.
    My bottom line about econ and also philosophy is that a tendency within the field has come to dominate via academic politics. More here.
    I’ve argued at DeLongs that econ should be regarded as a bag of tricks or rules of thumb like the ones used by builders and animal breeders, etc., instead of as a systematic formal science. There are people within the profession who accept something like that, but they’re a minority.

  29. I’ve argued at DeLongs that econ should be regarded as a bag of tricks or rules of thumb like the ones used by builders and animal breeders, etc., instead of as a systematic formal science.
    Now you’re talkin’ ! I couldn’t agree more. Do you have a disciple club I could join ? Or a task force to get results ? I can blow my own horn through my hat, so I wouldn’t need any extra kit. And quantilla prudentia is my middle name.

  30. dearieme says:

    “a bag of tricks or rules of thumb”: aye, but they’re mainly negative rules – “That won’t work the way you say because…..”.

  31. John Emerson says:

    That’s not really true. But it would be very interesting to see the relationships between the ways banks, businesses, and governments make decisions, and the various things that formal economics says they should do (and a lot of trained economists do work in business and finance.) There’s known to be a lot of discrepancy, but are the businessmen wise or stupid?
    For example, you have one set of ideas about deficits, debt, and inflation (bad) and another set of ideas about growth, credit, investment, and interest rates, and another set of ideas about currency and balance of trade. And they all are interwoven, but there’s still disagreement about how they fit together. And until the most recent crash, one school (anti-keynesians) thought it had won the argument, but now everything is back to zero.
    My theory is what I said, that you have partial systems that are pretty well understood, except that they’re not independent of each other and haven’t been fitted together neatly so far.
    But then it comes down to how much to keep and how much to throw out. More disagreements.

  32. Trond Engen says:

    I read your position as not so much against maths in economics as against wild extrapolation beyond the validity range. A Grand Unified Theory will take more maths, not less. And a humble attitude to the universality of results.
    So why are there different schools rather than different valid approaches that should be unified? Why can’t economists accept that their opponents, too, base their results on some aspect of reality? Hell, why are they opponents at all, rather than interesting c contributors? Why won’t economists tell when something in a situation makes their preferred approach invalid? Why don’t they do more research on these limitations, when one would think those were the most interesting parts? Well, obviously, because the subject is so politicised. There’s little room for a critical appoach to one’s own results when they’re less judged on scientific merit than on political usefulness.

  33. I read your position as not so much against maths in economics as against wild extrapolation beyond the validity range.
    Eggs-actly.
    a bag of tricks or rules of thumb like the ones used by builders and animal breeders
    Builders tend to be prosecuted when they erect buildings with their thumbs. Animal breeders appear to be quite au fait on genetics, vaccines, ecobehavioral interdependencies – to judge by the programs I see on German TV.
    Economists already have bags of tricks and rules of thumbs. These consist in deploying mathematics, statistics, and schools of thought for political purposes, as Trond says.
    But it would be very interesting to see the relationships between the ways banks, businesses, and governments make decisions, and the various things that formal economics says they should do (and a lot of trained economists do work in business and finance.) There’s known to be a lot of discrepancy, but are the businessmen wise or stupid?
    John, that’s an interesting question about decision-makers and economists. I have had the impression that the names of “schools of thought” often crop up, if only briefly, in TV discussion involving government-employed economists and governmental economic policy. But when the Chief Economist of the Deutsche Bank speaks, one doesn’t hear about his school tie right off the bat. This is just an impression I have gotten. Many contributors at nakedcapitalism use elaborate school-of-thought ratings for everybody they talk about. They sound like bookies for wheeler-dealers instead of horses.

  34. In reference to John’s question again: we really need a science to study the preaching and practice of economics. I imagine that it should be interdisciplinary, but devoid of mathematics. Say with contributions from political science, sociology, zoology, criminology and the world of fashion.

  35. John Emerson says:

    “Politicized” is wrong. It assumed that if we only had pure apolitical experts doing this, we’d be OK. But one of the problems is that the knowledge isn’t there. We do not have a good theory of economics the way we have a good theory of physics, and for various reasons (complexity, historical change, etc.) we probably never will. So there probably always will be a rule-of-thumb guesswork aspect.
    The second reason it’s politicized is that there are real-world consequences. It’s not like studying ancient history or toad evolution or the microphysics of basalt where there’s no pressure and pure truth can rule. And these pressures would impact anyone in the field, regardless of their purity and good intentions.
    it’s been my experience that in all applied fields, while the experts know the theory, in practice they have all kinds of rules of thumb and general philosophies and bags of tricks required to actually do anything. And that’s the way it is in economics, and I’ve conjectured that that’s how it always will be. So the only criticism of economics, but a big one, is overreach. (OK, two. Overreach and corruption. Three, ideology. But ideology can be a good thing).
    The study of economics is part of “Science studies”. Philip Mirowski and Deirdre McCloskey both have done good things.

  36. Trond Engen says:

    This talk by Paul Krugman for the Association of American Geographers contains a warm defence for unrealistic mathematical models.

  37. John Emerson says:

    krugman is a good guy visavis George W. Bush, but not necessarily otherwise.

Speak Your Mind

*