Asking the right question

Profile picture for user gonzodaddy By Den Howlett February 21, 2016
Summary:
Asking the right question is much harder than we think.

braintwister

Here's a puzzle for you. It first turned up in my Facebook stream over the weekend and presents an interesting challenge. Curious though this may seem, there are three possible answers to the question, depending upon the logic you choose to apply. Can you work it out?

Here goes working along each line:

  1. 3r=60...r=20
  2. 20+2b=30...b=5
  3. 5-2y (or y)=3...y= 2 or 1
  4. 1 (or 2) +10 +5 (or 4 if you are counting petals)  gives us 25, 26 or 27

There is a logic to all of those answers. That's not the point although there has been plenty of discussion about which answer provides the best result. I prefer 26 because I view the number of petals is less relevant than the number of flowers but heh. Lucretia Madden Pruitt offers an interesting perspective on how this works. She argues:

....there is no single "logical" answer. The parameters are not clearly defined enough to exclude multiple answers. That's kind of why it's fun. But I personally think it rather reflects human communication in general. We all have our preferences about underlying assumptions and think other people should as well. Then, when they come to a differing conclusion? We argue for our own interpretation.

Of course. David Norfolk offers an alternative view:

Or we could count petals, or a blue flower with a different number of petals might be an entirely different symbol and the last line is insoluble, or 2 flowers together might mean flower plus flower (but why?) instead of flower times flower.

He concludes that the puzzle is silly because the terms are too ambiguous. This has huge implications about how we view the world of analytics and is one of the reasons I am often skeptical about technology's ability to provide good answers. Here's another example. Back in the day when I used to heft numbers for a living, there was a running joke that you could line up six accountants, give them the same set of balance sheets, ask them to provide an interpretation and you would get at least seven different answers. The real answer to that question, much as it is to this puzzle is: 'It all depends.' In the case of accounting analysis, it depends on what you want to know and the manner in which you go about solving the problem.

Fast forward to today where the expression du jour is 'big data analytics' - whatever that's supposed to mean. Regardless, my news feed is peppered with stories claiming variations on the theme of: analyzing big data is incredibly important because it provides insights and/or solutions into previously unimaginable problems like making strong predictions. That's all very well and good, but then I have a multipart question: What is the problem you're trying to solve, where are the reliable data sources and how is algorithmic bias managed?

Until we answer those questions with a fair degree of rigor then all the data in the world and all the beautiful visualizations we have available will not help us.