Duncan's Diurnal Diatribe: Learning Curves

Introduction

Learning Curves. I know from some of the work I do that many accountants and others need to have some knowledge of learning curves but they have little to no idea of how Excel can help them. Well, you came to the right place, then!

Using the Same Ten Numbers Every Time

I just set up my 9.5 year old daughter Abi, with ten randomly selected numbers, taken from a set of numbers from 1 to 75 inclusive. Her job was to place those numbers as efficiently as possible on a number template. She took 42.29 seconds to do that for the first time, then 40.49 seconds and then 32.42, 31.17, 36.00, 29.90, 29.63, 22.98 and 31.82 in a total of ten trials.

We now have enough data to assess Abi's learning rate. It turns out to be a creditable 91%. By the way, you will read that 80% is the standard learning rate for humans: I challenge that since that is a very steep learning rate that even a genius would find hard to attain.

Take a look at my worksheet, below. Any questions?

Abi's Results

The Learning Curve Blip

I gave Abi the task of placing number counters on a template grid. Did you notice that her results showed an improvement every step after trial 1? EXCEPT for Trial 6 where she went from 31.17 to 36: that happens every time I run experiments like this ... there is always a blip!

Using Ten Different Random Numbers for Each Trial

What about the learning curve results when the numbers to be sorted are different every time? Abi got the same numbers every time. I did the experiment next time but Abi randomly drew ten new numbers every time. How did I do? Did I learn? How could I learn?

Well, my times were: 39.94, 35.12, 34.40, 35.91, 34.32, 31.99, 31.67, 36.94, 29.18, 34.81 and my calculations show that I did learn, with a rate of 95.78% Not as good as Abi but I learned: what did I learn, though?

I learned the layout of the template, which was not as user friendly as you might think: it was a 5*15 grid, total of 75 numbers ... not as easy as a 5*10 or 10*10 ... Secondly, when I picked up a number counter, I turned it over and as I did so, I moved my hand to the centre of the template: minimising the distance my hand then had to travel whatever its value. Were there any blips in my results? Yes, at trials 4, 8 and 10. Anyone can run experiments like these!