Green Honey Problem

If you can't tell, you can't see.


Problem Statement

The green honey problem refers to the case when Alice has to rely on Bob to see the outside world:

  • The implication for Bob is pretty straight forward: Bob will be limited to what Alice can see, and on what she can express:

    • If Alice can't see the color pink, Bob will never hear the word pink.
    • If Alice don't know the word pink, Bob will never hear the word pink.
  • The real problem is on Alice:

    • Since Alice don't know the expression for the color pink, can she see the color? I will argue that she can't.

In Homer’s Odyssey, sea is wine dark, sky is bronze, and the color for honey and hair is green (thus, green honey). The color blue has never appeared during the tale. In ancient Greek, the language can only separate the gradient of two colors, but not the hue. The word kyaneos represents blue, dark green, violet, black and brown.

Complexity of Chess vs. Complexity of color

It seems absurd that the lack of vocabulary obstruct the physical ability to see, but consider Chess game: it has been confirmed that the difference between a master player and a beginner is not due to the fact that a former has better memory, or higher speed of processing [^1]. Instead, the master has developed domain specific model of Chess, that correctly abstract to current game, which can be further processed to the model has been internalized.

[^1]: If this was true, a great master in Chess should also be smarter than the beginner outside of Chess

Better abstraction helps you to see more detail, and this is what separate master from beginner: the abstraction they have developed.

The question then become: is the combination of color to be more complex than a game in Chess? I will like to think so, there is much more information: while the complete info on random Chess game can be recorded in 64^3 (black-white-space), one of the most common way to store a visual is to take a picture, a 500x500 pixel picture has the size of 153,119 byte.

This will explain why practicing sketching makes you see more details of the world: by practicing expression, one is developing a better model to decompose the world, when these models start to become more effective, one does see more.

I used to think color names as vocabulary, but it seems to be more like a rule for abstraction. In ancient Greek, they choose to abstract gradient. With such system, only gradient information is stored in their heads, which effectively is what they get from what they see.

Taken writing system as example, try to memorized:


What about:




If you have developed a model for Russian, Chinese and Klingon, you will need very little effort to store this word; otherwise it's nearly impossible to see the information, even though it's just in front of you.

That leads to some interesting question and insight:

  • If there are different way of abstracting information, how do we pick the/is there a best abstraction?
  • Publishing forced us to practice on our abstraction skills, which in turns enhanced our model of the world:
    • Consistent with the common belief that teaching make you better understand something
    • A practical advice on how to improve your work (data visualization), will be to publish often (i.e.. one chart per day)
  • This also yields an interesting view on the role of data visualization:
    • Intelligence can't be transmitted, but model can. We may not be as smart as Einstein, but since he transmitted his e = mc^2 model to us; we are somehow able to see the same thing he saw.
    • The role of Data Visualization is to transmit good model that was developed by domain experts; if there is no good model, there can't be good visualization. This seems pretty obvious, but it explains why most of the graph sucks: because the author didn't have any interesting model on the issue.
    • Some people still believe that we can have artist working on data that they don't understand, yet that somehow in the process of decorating, a graph can be made. It seems like the only way to make meaningful graph, will be to develop more knowledge.
The above are some ideas of the Green Honey talk; I plan to rewrite this article with more examples and graphs in the near future.

Derived from my talk at the Open Data Workshop 2013.