Is Design a Stochastic Process?

Posted on by

https://en.wikipedia.org/wiki/Stochastic

Researchers use the term stochastic systems to describe the physical systems in which the values of parameters, measurements, expected input, and disturbances are uncertain

Would we expect different designers to create different designs? Does the same designer ever design competing solutions? Why is that? What factors inform a design decision and outcome?

Outside of hard determinism (https://en.wikipedia.org/wiki/Hard_determinism) isn’t everything a stohastic process?

196 thoughts on “Is Design a Stochastic Process?”

  1. Richardthughes: Meanwhile, look at all of those stochastic inputs: ” considering the aesthetic, functional, economic, and sociopolitical dimensions of both the design object and design process.”

    Are those stochastic inputs? I would say they were variable but definitely not stochastic. Am I missing something?

    peace

  2. fifthmonarchyman: is variable the same thing as stochastic?

    I’ll let others answer that, but the very nature of “variable” shows it can vary. The values they can vary between and the probability density of those values are not clearly defined, but may be amenable to modeling.

  4. Mung: So if I am developing some software, I need to understand all the trial and error that went into the design of the programming language that I am using, and all the trial and error that went into the design of the computing platform it will run on, etc., etc.?

    No wonder no one understands the design process!

    You don’t need to understand the whole process to program something, to understand the whole process of programming is a different matter

  5. Richardthughes: The values they can vary between and the probability density of those values are not clearly defined, but may be amenable to modeling.

    I’d agree.

    It seems like “may be amenable to modeling” clause would rule out stochastic processes. At least that would be my understanding.

    In fact the area between not clearly defined and random is where I would say that design lives

    peace

  10. dazz,

    I guess so. We’ve had a few free will threads here. I think its known I’m a hard determinist. But as the bigger system is unknown to me, I still feel like I’m making choices (that I was always going to make) ;P

  11. fifthmonarchyman,

    There’s absolutely no way I’m wasting a minute of my life reading something called “A GOD-CENTERED APPROACH TO PROBABILITY AND RANDOM EVENTS”

    Thanks anyway

  12. dazz: There’s absolutely no way I’m wasting a minute of my life reading something called “A GOD-CENTERED APPROACH TO PROBABILITY AND RANDOM EVENTS”

    Your loss.

    If you wanted to understand what smart folks on the other side thought it would be worth your while.

    peace

  13. dazz:
    fifthmonarchyman,

    There’s absolutely no way I’m wasting a minute of my life reading something called “A GOD-CENTERED APPROACH TO PROBABILITY AND RANDOM EVENTS”

    Thanks anyway

    but.. but … diagrams!

  14. Richardthughes: but.. but … diagrams!

    Yes you know how us ignorant stupid fundies need pictures. 😉

    What do you expect from such a uneducated rube.
    I mean it’s only Harvard after all

    peace

  16. fifthmonarchyman: If you wanted to understand what smart folks on the other side thought it would be worth your while.

    I don’t doubt there are smart folks on the other side, it’s just that at this point, I think it’s a pretty safe bet anything you recommend is not going to come from any of them

  17. Richardthughes: Do you think Design (the broader enterprise, not a specific instance) has stochastic components?

    Everything that happens in the real world has stochastic components.

    Sure, mathematics doesn’t (not counting probability theory). But that’s because mathematics is abstract, not real. Or, to say it differently, mathematics is idealization.

    Newton’s laws are not stochastic, because they too are idealizations. An engineer, designing a rocket to the moon, has to take the random component into consideration — else he will miss the target. Reality is stochastic. Our intellectual accounts are often idealizations which miss that stochastic aspect.

  18. dazz: I don’t doubt there are smart folks on the other side, it’s just that at this point, I think it’s a pretty safe bet anything you recommend is not going to come from any of them

    confirmation bias anyone 😉

    Just goes to show the benefits of living in an intellectual bubble.
    Phd from Harvard, Who cares must be stupid if FMM likes him, ;-).

    Certainly would not want to waste a minute finding out

    I think our patron saint Cromwell is weeping

    peace

  19. Neil Rickert: Our intellectual accounts are often idealizations which miss that stochastic aspect.

    That stupid guy with the book calls the “random” aspect an opportunity for creativity. It’s where all the action is.

    peace

  20. Frankie: Yes, intentional design so that you don’t get confused with apparent design or optimal design.

    So optimal design can’t be intentional? Nor apparent? Interesting! Making shit up must be fun.

    Frankie: And Richie, it doesn’t matter if you cherry-pick.

    Let’s see if our eagle-eyed readers can spot the cherry-picked definition:

    https://www.merriam-webster.com/dictionary/stochastic

    or

    https://www.stat.cmu.edu/~cshalizi/754/

    Use the Explanatory Filter!

  21. dazz: You don’t need to understand the whole process to program something, to understand the whole process of programming is a different matter

    And understanding the whole process is important to avoid drawing bad conclusions, like “design is not a stochastic process”.

  24. Fair Witness: And understanding the whole process is important to avoid drawing bad conclusions, like “design is not a stochastic process”.

    How is that a bad conclusion seeing that there isn’t any support for saying design is a stochastic process?

  25. Frankie: Why? Show me a design process that does

    Until you show some sort of reasoning behind the claim there is nothing to disprove, but I do have an example.

  26. newton: Until you show some sort of reasoning behind the claim there is nothing to disprove, but I do have an example.

    Designs are planned, random is the antithesis of planned and stochastic processes are not planned.

  27. Richardthughes: Do you think Design (the broader enterprise, not a specific instance) has stochastic components?

    The difficulty here is that I have far too much to say about the matter. In software engineering, it is usually quite a problem even to produce clear-cut requirements to be satisfied in design of the system. The process of requirements specification turns up questions that people had not realized needed answering. People are unsure of how to answer the questions — when you don’t yet have a system, it is difficult to say what it is that you want the system to do — but generate answers because they otherwise will not obtain a system. And when they do have a system, their experience with it usually leads them to want a system that works differently. What they want depends on what they already have — which is to some degree arbitrary. And what modifications they will get depends on the decisions made by designers of the original system. Changes that seem small to users are sometimes impossible to make without major (very expensive) changes to the system. Rather arbitrary decisions made in requirements specification and initial design place strong constraints on what the system will ever evolve to be. (Yes, I know what word I just used.)

    I’m more comfortable saying that there is huge contingency in human design of complex systems than that there is huge stochasticity. (There are some scientists, e.g., Joe Thornton, who say that evolution is contingent.) However, checking the spelling of stochasticity in Wiktionary, I found this definition:

    the quality of lacking any predictable order or plan.

    As for adaptive (learning) systems, there is inescapable uncertainty as to whether the future will be like the past. We have no logical justification for many decisions that we are forced to make.

    Interesting post, Rich. I could get sucked into spending days on it. I’d love to talk, for instance, about the noisy connection weights of the 20 thousand neural nets that I used to forecast annual sunspots counts (with considerably greater accuracy than had previously been achieved). Making the weights noisy essentially limits their precision, and counters overfitting of the data with excessively many parameters in the model. Just for fun, I’ll mention that I learned from Jorma Rissanen’s Stochastic Complexity in Statistical Inquiry that it is not just the number of parameters, but also the precision of the parameters, that matters in modeling.

  31. Not sure what the point of the OP is, but harking all the way back to the days of the explanatory filter, the first way to eliminate design is if the process is deterministic.

    So if we define stochastic as the antithesis of deterministic, then clearly design is stochastic.

  32. Mung: eliminate design is if the process is deterministic

    The first tap of the explanatory filter is regularity, as I recall.

    Do you understand that active information is maximal for deterministic processes?

  33. Tom English: Do you understand that active information is maximal for deterministic processes?

    I do not, nor do I understand the relevance. But I hope you’ll get to it in your latest series. Or at the least I’ll look forward to seeing what the EvoInfo book has to say about active information. 🙂

  34. Tom English: The first tap of the explanatory filter is regularity, as I recall.

    Yes- regularity, law, necessity- I would think determinism fits those.

    Do you understand that active information is maximal for deterministic processes?

    It shouldn’t be but if you have something to suggest otherwise present it so we can take a look.

  35. The idea of probability as a measure of uncertainty about unknown but deterministic quantities is an old one.

    Peter D. Hoff

  36. Mung: I do not, nor do I understand the relevance. But I hope you’ll get to it in your latest series. Or at the least I’ll look forward to seeing what the EvoInfo book has to say about active information.

    I don’t recall any mention of deterministic processes in the technical papers of Marks, Dembski, and Ewert. I do recall some places in which they should have mentioned deterministic processes, but did not. Dembski had some things to say about determinism in Being as Communion: A Metaphysics of Information, but I don’t recall any comment about the active information of deterministic processes.

    A deterministic process either hits or does not hit the prespecified target. That is, the probability that a deterministic process hits the target is either q=1 or q=0. The active information of the process is

        \[I_{+} = \log_2 \frac{q}{p}~\text{bits},\]

    where p is the probability that a baseline “undesigned” process hits the target. The denominator p is fixed, and the numerator q is either the greatest possible for any process, or the least possible for any process. That is, the active information of a deterministic process that hits the target is

        \[\log_2 \frac{1}{p} = -\log_2 p,\]

    and the active information of a deterministic process that does not hit the target is

        \[\log_2 \frac{0}{p} = -\infty.\]

  37. Mung:
    The idea of probability as a measure of uncertainty about unknown but deterministic quantities is an old one.

    Peter D. Hoff

    Not to be dismissive, but… Of course. Few scientists are saying with their probabilistic models that stuff just happens by chance (i.e., that unpredictability is due to quantum indeterminacy). Only in the last 15 years or so has it become clear that some mutations are the consequence of quantum-scale phenomena. Some mutations are caused by macroscopic processes (explained well, at least in principle, by classical physics).

  38. Frankie: Designs are planned, random is the antithesis of planned and stochastic processes are not planned.

    You can plan to incorporate the enventual unplanned occurrence into your design, flexibility . There are lots of ways to create. The Grand Canyon is designed . If like ID someone designed whatever somehow sometime, that ability would be very useful for the non omniscient designers.

    So if I planned to do cannonball in a pool then did it, is the shape of the splash a design?

  39. newton: You can plan to incorporate the enventual unplanned occurrence into your design, flexibility .

    It is still part of the PLAN. And who designed the grand canyon? It doesn’t look designed to me

  41. Designs are planned or does Richie think they happen by accident?

    design:

    a : to conceive and plan out in the mind

    Whoopsie, cupcake…

  42. newton: So if I planned to do cannonball in a pool then did it, is the shape of the splash a design?

    Does it spell out “created by newton”? If so, I would say yes, designed.

  43. Mung: Does it spell out “created by newton”? If so, I would say yes, designed.

    My Holy Book says that the waters of the pool were without form until Newton took the plunge. Newton in-formed the waters of the deep end.

  46. Richardthughes,

    I click.

    I see: “design is a stochastic process which eventually leads to a deterministic plan.”

    I recall: “It depends upon what the meaning of the word ‘is’ is.”

    The author presumably means that design is sensibly modeled as a stochastic process. That is is not the metaphysical is.

  47. Tom English,

    There’s potentially a metadesign, a design of the design. I don’t want to get too regressy (not a word) but at some point a collection of resources is brought to bear on a problem. These resources are likely constrained in some way, (time, budget, expertise, materials, methods) and so the inputs of the process are fluid. If we were to look at enough designs, we would see a distribution of inputs. These we could model either parametrically or nonparametrically if they were cardinal numbers, and future designs would have inputs drawn from those distributions.

    Edit to add: Categorical inputs could also be modelled as distributions.

  49. Frankie: It is still part
    of the PLAN. And who designed the grand canyon? It doesn’t look designed to me

    Correct, to plan to incorporate the unplanned. What designed the Grand Canyon. That’s ok.

  50. LoL! Richie quote mine strikes again! I bet you stayed up all night looking for that. Too bad that stochastic processes are unplanned whereas design processes are planned. That will never change.

Leave a Reply