Zoltan indicators: Turning the Mechanical Turk into Zoltan the Fortune Teller

IStock_000009276724XSmall Yesterday, Sara Winge yesterday told me about the mechanical turk.

The Mechanical Turk (as I understand it)

Let's say Amazon needs to catalog a new product.   It has the word "Java" in the title, and that's a problem.  The system can't tell whether Java refers to the place, the coffee, or the programming language.

Someone at Amazon could sit down and figure this out.  But this would use high priced talent to perform a relatively simple task.  So what Amazon does is reach out to their Mechanical Turk. 

The Amazon Mechanical Turk is, as Wikipedia puts it, "a crowdsourcing marketplace that enables computer programs to co-ordinate the use of human intelligence to perform tasks which computers are unable to do."  It consists of thousands of people who stand ready for tasks send them by Amazon or others who may wish to use Amazon's MTurk service.

MTurk "providers" work alone, often in their spare time.  Standing in line at a 7/11, they can  bang out a few turns.  They get paid a small fee for each decision.  No one gets rich working in a mechanical turk, but many find it interesting. 

That was probably not a perfect introduction to those just learning of the mechanical turk, but it sets things up for the theme of this post: how to detect change in our culture.

Change in our culture (when the forecasting was easy)

In the middle of the 20th century, when we were a more organized culture, it was pretty easy to detect and track cultural change.  Often it came from small elites (artists, editors, filmmakers, etc) in a handful of cities (New York, London, Paris, etc.).  It was picked up by ever less obscure venues (magazines, art galleries, etc.) and began its march towards the mainstream.  Many died along the way, but the ones that made it "all the way home" would shape the tastes and preferences of every consumer, and the ideas, sentiments and practices with which we see the world.  They became our culture, some for the short term, some for the long. 

The cultural changes were like breakers off Waikiki: big, fat, orderly, and easy to see from a long way off.  Even quite dim corporations could spot the future before it had installed itself in the present.  And make ready.  Even in the smoke filled haze of a Madison Avenue boardroom, even to ad men soaked in gin, someone was going to say, "you know, that bohemian thing…from Paris…you know, Henry Miller.  Bongos?  Berets?  Geez, you guys really need to get out more."  The corporation might be a little a slow off the mark, but it was never completely unprepared.

Now of course the future comes from many sources and all directions and the corporation lives as Peter Schwartz once put it in a "perpetual state of surprise."  There is always a blind side hit waiting to greet them…sometime…from somewhere.  Those breakers at Waikiki have turned into a perfect storm.  They thrash.  Veritably, they boil. 

Change in our culture now (now that it isn't)

We really are "crowdsourced" now.  In the place of small elites in a few places, cultural innovations can come from any number of people in any number of places.   One of the big effects of the digital regime is that it disintermediated the old media and players, and gave us freer access of one another.  Now some kid can invent a new kind of music or filmmaking, a new sensibility or point of view, and without the aid of elite players and channels, our culture can respond.  To be sure, for this innovation to find its way into your world, vast amounts of capital and influence will have to be spend on its behalf.  But that's the thing.  Capital and access are now much more responsive than before, responding to something more like the will of the crowd than the arbiter of taste. 

Of course, this makes it much harder to read the turbulent waters of our culture.  Corporations that were clumsy and a little late to the party now found themselves sometimes terribly out of touch.  What to do?

Where the Mechanical Turk comes in

It occurred to me that the Mechanical Turk could be pressed into service here.  What if we captured little innovations from the far margins of our culture, send them to our Turk providers, and asked questions like the following:

1.  Is X something you recognize as conforming to some genre, style, model, pattern?

2.  If yes, is it completely "true to form?"  [If yes, code accordingly and end task]

3.  If it is not true to form, is this because it is incompetently executed?  (The maker doesn't really grasp the genre, style, model, pattern.)   [If yes, code accordingly and end task]

4.  Does it depart from form because it contains another form?  (An example: It was usual to say of music in the late 1980s that it combined punk and heavy metal.) 

5.  Does it depart from form because it contains an element that is not formed (i.e., that is difficult or impossible to recognize?) 

6.  How new is this new element to you?  [rank on continuum: from "very" to "not very"]

7.  How disagreeable is this new element to you?  [rank on continuum; for "very" to "not very"]  [[My assumption is that real innovation is always a shocking, hard to think and therefore disagreeable.  This is because we don't yet have a form in our head for it.  It comes to us as noise.  Avant-garde types like noise.  But most of us find it disagreeable.  See my discussion of the "Kauffman continuum" in Flock and Flow for more on this.]]

8.  Please review this innovation you looked at yesterday [or a week ago] and indicate whether it remains "new" and "disagreeable."  [rank on continuum from "yes, it remains disagreeable" to "no, I am beginning to like the look of it more."]  [[My assumption is that all culture novelty is disagreeable on first sight.  But some of this novelty begins to "win us over" over time.  The innovations that win interest and approval from Turk providers are hot spots for forecasting purposes.  In a perfect world, we would keep dropping innovations in front of our providers and watching for this "I am beginning to like the look of it more" movement.  How fast and in what volume this conversion takes place is a key indicator of innovation "with legs."

I don't doubt that this looks like a dog's breakfast, especially to people with training in questionnaire design and quantitative data collection.  And I know some will be alarmed at the freedom with which I have used terms like "form," "style," "pattern," etc.  At least here I can reassure you that these have substance and can be identified and defined particularly.

But I believe the Zoltan approach has the following virtues. 

1.  It allows us to canvass consumer reactions simply.  These are easy questions.  They allow the respondent to give us useful data even when they cannot necessary form coherent thoughts on innovations. We are asking them to say what they like, not why. 

2.  It allows us to canvass consumer reaction broadly.  Because this is crowd sourced, we can canvass reaction in every corner of the wired world.  This allows us to contend with the very distributed nature of innovation these days.  If something is stirring in Minneapolis, the mighty Zoltan will know about it.  Indeed, we can track innovations minted in Brazil, track them as they pass through the Caribbean and watch for landfall on the eastern seaboard.

3.  It allows us to organize our providers according to Kauffman and other identifiers, so we know the risk tolerances and innovation enthusiasm of the coders.  Some coders will be stand high on the Kauffman continuum.  We will know this from their other coding work and from our own diagnostics.  This will make them especially reliable in identifying innovation that are noisy but promising.  These people are good at handling noise.  They are not so much early adopters as "early oculars."  They can see the new in the noise.   Those who stand in the middle of the Kauffman continuum will be reliable indicators of the likely reaction of the mainstream.

4.  Data of this kind would give us a baseline from which to chart the movement of innovations.  This will allow us to track whether an innovation is meeting with acceptance, how fast it is doing so, and to whom it appeals.


Lorica, Ben.  2009.  Mechanical Turk Best Practices.  O'Reilly blogs.  June 11.  here.

McCracken, Grant. 2006. Flock and Flow: Predicting and Managing Change in a Dynamic Marketplace. Indiana University Press.

McCracken, Grant. 2009. Chief Culture Officer: How to Create a Living, Breathing Corporation. Basic Books. Preorder from Amazon here.

Schwartz, Peter. 1996. The Art of the Long View: Planning for the Future in an Uncertain World. Currency Doubleday.


John Kearon and Mark Earls are doing some fascinating work on new ways to canvass consumers.  I am indebted to them for several lively conversations on this topic.

4 thoughts on “Zoltan indicators: Turning the Mechanical Turk into Zoltan the Fortune Teller

  1. Steve Portigal

    I have this vague dis-ease that some of this might turn into online cool hunting (and I’m sure monitoring Twitter streams for brand mentions might be one necessary yet uninspiring manifestation). Is this revolutionary? Or is it creating a more dense hierarchy, where tastemakers are parsed into Dee Dee Gordon’s clients at the top, Dee Dee Gordon (she’s just an example, of course) near the top, and a broad swath of Dee Dee Gordon’s informants along the bottom. Do we have a hub-and-spoke alternative? I guess the magic/myth of viral is that someone can break through from another place, but it seems like as soon as that happens, the pyramid model returns and tries to replicate or exploit the hubs and spokes to suit their hierarchical (i.e., money-making) model.

    Sorry for the cynicism and I’m totally happy to have someone tell me I’m missing the point!!!

  2. Grant McCracken

    Steve, you make it sound like the Kauffman continuum is a popularity contest, another way to elites to muster, but it is about people relative tolerance for noise in the signal.  And this really is crowdsourcing.  We dont have to know who they are.  and we dont need to have brand names.  Thanks, Grant

  3. Patrick Pearce

    This is not unlike the approach adopted by trend spotting companies which use freelance folks from all over the world to find local examples of global trends and then rank them on their prevalence. Having been such a “provider” when I was living in Beijing, I can say that identifying trends takes some serious time investment to understand the nature and scope of each broad trend before being able to clearly spot local examples, so the simplicity of your approach has definite merits. But wouldn’t you want your providers to spot and submit innovations as well? This discussion could be facilitated with a practical example of ‘X’…

  4. Grant McCracken

    Patrick, I was thinking of something a little less difficult, a little more automatic, a little less taxing, not so much whats the idea, but is there an idea. But thanks! Grant

Comments are closed.