The Empty Seat
“Boarding Complete.”
Nothing is sweeter than that utterance if you’re sitting in economy class next to an empty seat, and everyone has already taken their places. There are instances when I’d gladly take an empty seat next to me over business, or three empty seats for a transoceanic flight. An empty seat next to you beats an exit row with an immovable armrest. Bulkhead might offer more leg room, but you have to stow your bag for take off and landing. Exit rows on some aircraft can also be colder than other areas.
You can easily start ranking hundreds of small variables against preferences for individuals to create a fingerprint of their unique tastes. The real power comes when you can do this fingerprinting for a larger group of customers and stakeholders.
It’s possible to start grouping individuals by preferences. Similar to rank voting, by optimizing for most people getting closest to their higher rankings, you a get overall happier group of individuals.
While this approach may sound effective, it’s difficult to implement because collecting and ranking preferences isn’t easy. Are you going to provide a long questionnaire to customers? Are you going to subject the equivalent of an optometrist exam… “which one is better… number one or number two? Number one… or number two?”
This is another opportunity to employ LLMs. With the proper prompting, we can extract from transcripts of interactions things people like versus things they don’t. This is a great exercise for us to perform on ourselves to surface our own tastes or to see how they shift over time. Feeding our chat histories and emails to these services, under proper privacy protections, might be enlightening (or frightening).
