Natural Selection in Design

Affinova.com is a company which has explicitly adopted an evolutionary approach to product design.  Essentially, products are broken down into genetic components which are then evolved by a group of participants.

Advertisements
This entry was posted in Aesthetics, Complex Systems, Design Process. Bookmark the permalink.

6 Responses to Natural Selection in Design

  1. Wow. What an odd thing. OK, so here is my question: Are they actually simulating evolution, or are they using evolution as a metaphor for an old school iterative design process? In other words, do we have a practice of evolution here, or merely a rhetoric of evolution?

    As you think about this question, don’t just weigh in with an opinion, but ask yourself how would one evaluate whether a system is truly (as opposed to merely metaphorically) evolutionary?

  2. jimmypierce says:

    At first I was skeptical, but after thinking about it, I now strongly suspect that what they are doing is in fact much closer to a simulation of evolution (rather than simply iterative design process).

    The reason I think this is the case is that, although they don’t make this explicit, I suspect they have FORMALIZED the following operations they mention, namely fitness-biased selection, crossover, mutation, and removal operations.

    They claim to use an Interactive evolutionary algorithm, which apparently means that the fitness function is outsourced to humans (hence is not completely formalized), because it’s difficult to quantify many fitness functions ahead of time.

    My understanding of evolution is that it IS an algorithm. The important aspect of formalizing these criteria (mutation, fitness function, etc.) is that you can simulate the algorithm. This means that eventually you will arrive at a good solution (good meaning it conforms to the criteria you defined) but the process of arriving at that solution is MINDLESS: there is no foresight taking place as candidate solutions combined and destroyed.

    This different from iterative design, where the designers reflect on what they learn at each iteration. In most design work, there is no well-defined ways for evaluating success or combining, killing, or saving candidate solutions, hence they cannot create an algorithm, hence they cannot transfer this algorithm to some medium (e.g. computer) so that it can QUICKLY and MINDLESSLY and CERTAINLY(with some probability) converge on a satisfactory solution.

    Essentially, in (formal) evolutionary design all of the mindful, creative, reflective, foresightful thinking goes into deciding the criteria and paramters to be optimized. After that, it’s mindless computation that does the ‘design’ work. It’s interesting to imagine different future scenarios for design in light of trends such as this. Will all design work involve evolutionary algorithmic tools? What will the role of the designer be in this world?

  3. davidroyer says:

    @Jimmy
    “Will all design work involve evolutionary algorithmic tools? What will the role of the designer be in this world?”

    That IIT math background has got you thinking crazy 🙂

    Just kidding, I see your point. I don’t think an algorithmic design tool will ever truly work in capitalist system because humans are crazy.

    We are crazy, we don’t always want what is best for us, what is best for us is relative and personal, we are irrational, and we don’t know what we want. We want to be unique, but not too unique. We are wicked problems.

    The diagram on the Affinova page looks alot like the itterative design process to me. Just formalized with fancy words.

  4. chmbrigg says:

    BTW, Affinova used to have a full interactive walkthrough of the system, but apparently it’s become a vestigial part of the site. If you go to slide 5 of the following slideshare presentation: http://www.slideshare.net/xllora/combating-user-fatigue-and-contradictions-in-subjectivebased-optimization-schemes, there is a diagram which describes their approach a little better. Essentially, the “genes” of a product result in different phenotypes that are then selected by potential customers.

  5. jimmypierce says:

    @ dave:

    “I don’t think an algorithmic design tool will ever truly work in capitalist system because humans are crazy.”

    I disagree. the links that christian shared are proof that it is working, at least a little bit. apparently, Affinova is using interactive genetic algorithms to design sexy looking beverage bottles. Let’s suppose they can produce sexier beverage bottles with these algorithms than old-school designers who refuse to use the algorithms… if sexier means more people buy them than how can the old-school anti-algorithm designers compete?

    i do agree with this, for the most part: “We are crazy, we don’t always want what is best for us, what is best for us is relative and personal, we are irrational, and we don’t know what we want. We want to be unique, but not too unique. We are wicked problems.”

    but i dont think evolutionary algorithms and the like necessarily threaten this. AND I dont think that even if they did, that we would necessarily let that stop us from allowing algorithmic design to take over (precisely because it probably would lead to increased $$$). let me propose one future scenario.

    With mass production, designers in the past had to give up production as part of the design profession. Nowadays, designers for the most part don’t produce the things that people buy physically with their own hands. rather, they design the templates, blueprints and processes so that machines can create the things.

    in the future, designers will have to give up search for a design within a space as part of the profession. instead, the designers role will be to define problem spaces. machines will then look for solutions. the designers will look at the solutions and maybe tweek them, or perhaps revise the definition of the design space…perhaps deciding that they were completely wrong and they didn’t define what was really desired. this doesnt imply that DESIGN is simply a search within a space (as herbie simon would suggest) but rather that it entails searches within spaces. it will still be an iterative design process, except the designer has become another step removed from the process of bringing a form into the world.

    i dont think this scenario is that far-fetched. AND, just as designers got used to not actually producing artifacts with their own hands, i think could get used to not honing in all the way on the final solution with their own intellect, but rather automating the final design work and then evaluating the solution. so i dont think this type of scenario necessarily threatens the role of the creative, unique, wicked-problem-solving designer any more than mass-production did…i dont think it destroys the need to work in the “swampy lowlands” of design that us designers love. (although, machines that are better at defining problems than humans definitely could undermine this, that is, machines that are better at telling us what we need and desire than we are.)

    btw, dont let my ideas and math background fool you into thinking that i think we SHOULD formalize design in this way. i dont necessarily agree with that. i think it’s important to carefully consider how the things we value (e.g. human creativity) will be affected before introducing technology that ‘solves problems’.

    @christian: thanks for posting those links to the slides…very interesting!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s