Behavioural colour schemes

One of the seminal results of behavioural economics (a field I’m having less and less faith in as the days go by, especially once I learnt about ergodicity) is that by adding a choice to an existing list of choices, you can change people’s preferences.

For example, if you give people a choice between vanilla ice cream for ?70 and vanilla ice cream with chocolate sauce for ?110, most people will go for just the vanilla ice cream. However, when you add a third option, let’s say “vanilla ice cream with double chocolate sauce” for ?150, you will see more people choosing the vanilla ice cream with chocolate sauce (?110) over the plain vanilla ice cream (?70).

That example I pulled out of thin air, but trust me, this is the kind of examples you see in behavioural economics literature. In fact, a lot of behavioural economics research is about getting 24 undergrads to participate in an experiment (which undergrad doesn’t love free ice cream?) and giving them options like above. Then based on how their preferences change when the new option is added, a theory is concocted on how people choose.

The existence of “green jelly beans” (or p-value hunting, also called “p-hacking”) cannot be ruled out in such studies.

Anyway, enough bitching about behavioural economics, because while their methods may not be rigorous, and can sometimes be explained using conventional economics, some of their insights do sometimes apply in real life. Like the one where you add a choice and people start seeing the existing choices in a different way.

The other day, Nitin Pai asked me to product a district-wise map of Karnataka colour coded by the prevalence of Covid-19 (or the “Wuhan virus”) in each district. “We can colour them green, yellow, orange and red”, he said, “based on how quickly cases are growing in each district”.

After a few backs and forths, and using data from the excellent  , we agreed on a formula for how to classify districts by colour. And then I started drawing maps (R now has superb methods to draw maps using ggplot2).

For the first version, I took his colour recommendations at face value, and this is what came out. 

While the data is shown easily, there are two problems with this chart. Firstly, as my father might have put it, “the colours hit the eyes”. There are too many bright colours here and it’s hard to stare at the graph for too long. Secondly, the yellow and the orange appear a bit too similar. Not good.

So I started playing around. As a first step, I replaced “green” with “darkgreen”. I think I got lucky. This is what I got. 

Just this one change (OK i made one more change – made the borders black, so that the borders between contiguous dark green districts can be seen more clearly) made so much of a difference.

Firstly, the addition of the sober dark green (rather the bright green) means that the graph looks so much better on the eye now. The same yellow and orange and red don’t “hit the eyes” like they used to in green’s company.

And more importantly (like the behavioural economics theory), the orange and yellow look much more distinct from each other now (my apologies to readers who are colour blind). Rather than trying to change the clashing colours (the other day I’d tried changing yellow to other closer colours but nothing had worked), adding a darker shade alongside meant that the distinctions became much more visible.

Maybe there IS something to behavioural economics, at least when it comes to colour schemes.

Ordering in large groups

When you go out in a large group, ordering can sometimes become a pain. This is especially the case if you know each other well and want to collectively share a large number of dishes rather than each person ordering a dish for herself. Usually, you can end up either under ordering (I’ve seen cases where three curries have been ordered for a table of ten people) or over ordering (when lots gets left over). And someone or the other is usually left unsatisfied.

There are two extremes in which collective ordering for a large group can actually work. At one extreme, there is one “leader”, whom everyone else trusts to order. The leader finds out the group’s preferences and aggregates them and takes the decision on the group’s behalf. Usually the leader is someone who is trusted, so their decisions are generally followed. There might be some inefficiencies but the rest of the people can focus on the conversation while the leader can bask in the glory of power.

The other extreme that works is completely decentralized ordering, like we did last night when I met a bunch of relatives. People trickled in slowly, and we found it was not feasible (for the butterflies in our stomachs) to wait for the whole group to arrive before we started ordering. And so I ordered a pizza and a pitcher of sangria (when in a large group you don’t need to specifically target who is going to consume the pizza and each glass of sangria – it gets aggregated over). I took a slice of the pizza and a glass of the sangria, and the rest actually disappeared rather quickly.

As people came in, they got the hint, and we never had to waste any time in discussions of the “shall we order this” sort. People kept ordering what they wanted, and since we had an implicit agreement of “sharing”, everything presently got consumed. That we were collectively full was indicated by the point in time when no one was ordering. It turned out to be a fantastic dinner.

Now, there are some conditions that need to be met for this kind of ordering to work. Firstly, there should be no one in the group who is shy or hesitant to order by themselves or requires pampering – such people will end up hungry in this situation. Secondly, there should be some sort of implicit trust in the  group that people will be somewhat reasonable in their order. Finally, given that the only way to split the bill in such situations is equally (since who ate what is rather fuzzy) “tragedy of the commons” should not happen. All conditions were broadly satisfied last evening, and (in my opinion) things worked out.

What kind of ordering algorithms have you used in the past, and how has that fared? Do you think decentralized ordering actually works, or if there are other conditions that need to be satisfied for it to work? Do leave a note on your experiences with ordering!