A foray into pairwise comparisons where a user is asked to ask a series of questions comparing two options at a time to determine the ranking of a list of options.

Problem

The context for this subject was that pairwise comparisons came up as a solution when discussing how to choose a baby’s name. The idea is that human’s are much better at choosing between two options which one is better than choosing from a large list.

Initial Solution

The simplest solution is to generate all possible options, ask the user to compare each pair, and give a point to the winner.

I created a quick draft of what this could look like in this repo. There’s a main.py file which takes inputs from the user to generate all possible options. The core logic is in pairwise_comparison_session.py where the options and scores are stored. I store an indexing of an int to the string representation of the option and to the score.

A test class is added in test_pairwise_comparison_session.py for good measure.

Discussion of Improvements

• Algorithmic scaling
  • Constructing a directed graph. Space will look like a bunch of (potentially separate) trees.
  • Pruning: if you select option A as better than option B and option B as better than option C then you don’t need to compare optoin A and option C (and potentially just trim option C completely)
• Technical Scaling
  • the maps of index to string and score can be stored inside of a sql database for processing many options
• Architectural improvements
  • perhaps use data classes instead of basic string
  • consider something else besides python. For example web site or app. Python was the easiest/fastest option.
  • provide the above optimizations as options that can be used to finetune to specific use cases